Handbook of Knowledge and Economics 1843764040, 9781843764045

Citation preview

HANDBOOK OF KNOWLEDGE AND ECONOMICS

Handbook of Knowledge and Economics

Edited by

Richard Arena Professor of Economics, University of Nice–Sophia Antipolis/ GREDEG – UMR 7321, France

Agnès Festré Professor of Economics, University of Picardie Jules Verne/ CRIISEA and University of Nice–Sophia Antipolis/ GREDEG – UMR 7321, France

Nathalie Lazaric Research Professor, CNRS, University of Nice–Sophia Antipolis/GREDEG – UMR 7321, France

Edward Elgar Cheltenham, UK • Northampton, MA, USA

© Richard Arena, Agnès Festré and Nathalie Lazaric 2012 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library Library of Congress Control Number: 2011934811

ISBN 978 1 84376 404 5 Typeset by Servis Filmsetting Ltd, Stockport, Cheshire Printed and bound by MPG Books Group, UK

Contents

List of contributors Acknowledgements 1

vii ix

Introduction Richard Arena, Agnès Festré and Nathalie Lazaric

PART I

1

KNOWLEDGE AND ECONOMICS: A HISTORICAL PERSPECTIVE

2

What Vilfredo Pareto brought to the economics of knowledge Ludovic Ragni

23

3

Knowledge in Marshall Brian J. Loasby

49

4

Carl Menger and Friedrich von Wieser on the role of knowledge and beliefs in the emergence and evolution of institutions Agnès Festré

5

6

7

The pragmatist view of knowledge and beliefs in institutional economics: the significance of habits of thought, transactions and institutions in the conception of economic behavior Véronique Dutraive Imagination and perception as gateways to knowledge: the unexplored affinity between Boulding and Hayek Roberta Patalano The knowledge–rationality connection in Herbert Simon Salvatore Rizzello and Anna Spada

PART II

8

73

99

121 144

ECONOMICS, KNOWLEDGE AND UNCERTAINTY

A note on information, knowledge and economic theory Giovanni Dosi

v

167

vi 9

10

Contents The cognitive explanation of economic behavior: from Simon to Kahneman Massimo Egidi

183

Towards a theoretical framework for the generation and utilization of knowledge Pier Paolo Saviotti

211

11

Models of adaptive learning in game theory Jacques Durieu and Philippe Solal

246

12

The fragility of experiential knowledge Dominique Foray

267

13

One knowledge base or many knowledge pools? Bengt-Åke Lundvall

285

14

Knowledge in finance: objective value versus convention André Orléan

313

PART III

ECONOMICS, KNOWLEDGE AND ORGANIZATION

15

Embodied cognition, organization and innovation Bart Nooteboom

16

Knowledge and its economic characteristics: a conceptual clarification Ulrich Witt, Tom Broekel and Thomas Brenner

339

369

17

Tacit knowledge Paul Nightingale

18

The firm as a ‘platform of communities’: a contribution to the knowledge-based approach of the firm Ash Amin and Patrick Cohendet

403

The architecture and management of knowledge in organizations Mie Augier and Thorbjørn Knudsen

435

19

20

Distributed knowledge and its coordination Markus C. Becker

21

Evolution of individual and organizational knowledge: exploring some motivational triggers enabling change Nathalie Lazaric

Index

383

458

488

517

Contributors

Ash Amin, Department of Geography, University of Cambridge, UK. Richard Arena, University of Nice–Sophia Antipolis/GREDEG, France. Mie Augier, Center for New Security Economics and Net Assessment, Naval Postgraduate School, USA. Markus C. Becker, University of Southern Denmark/Strategic Organization Design Unit – SOD and Danish Institute for Advanced Study – DIAS, Denmark. Thomas Brenner, Philipps University of Marburg, Germany. Tom Broekel, Institute of Economic and Cultural Geography, Leibniz University of Hanover, Germany. Patrick Cohendet, BETA, University of Strasbourg, France. Giovanni Dosi, LEM, Istituto di Economia, Sant’Anna School of Advanced Studies, Pisa, Italy. Jacques Durieu, University of Grenoble/CREG, France. Véronique Dutraive, University of Lyon 2/TRIANGLE, France. Massimo Egidi, University LUISS Guido Carli, Rome, Italy. Agnès Festré, University of Picardie Jules Verne/CRIISEA and University of Nice–Sophia Antipolis/GREDEG, France. Dominique Foray, Chair of Economics and Management of Innovation, MTEI, Swiss Federal Institute of Technology of Lausanne, Switzerland. Thorbjørn Knudsen, University of Southern Denmark/Strategic Organization Design Unit–SOD and Danish Institute for Advanced Study–DIAS, Denmark. Nathalie Lazaric, University of Nice–Sophia Antipolis, CNRS/ GREDEG, France. Brian J. Loasby, Division of Economics, University of Stirling, UK. Bengt-Åke Lundvall, Department of Business and Management, University of Aalborg, Denmark. vii

viii

Contributors

Paul Nightingale, Science Policy Research Unit, University of Sussex, Brighton, UK. Bart Nooteboom, University of Tilburg, Netherlands. André Orléan, Paris-Jourdan Sciences Economiques (EHESS), France. Roberta Patalano, Department of Institutions and Territorial Systems Studies, Parthenope University, Naples and Department of Dynamic and Clinical Psychology, La Sapienza University, Rome, Italy. Ludovic Ragni, University of Nice–Sophia Antipolis/GREDEG, France. Salvatore Rizzello, Center for Cognitive Economics, University of Piemonte Orientale, Italy. Pier Paolo Saviotti, University of Nice–Sophia Antipolis/GREDEG and INRA/GAEL, Grenoble, France. Philippe Solal, University of Saint-Etienne/GATE Lyon Saint-Etienne, France. Anna Spada, Center for Cognitive Economics, University of Piemonte Orientale, Italy. Ulrich Witt, Max Planck Institute of Economics, Jena, Germany.

Acknowledgements

Our thanks to Hella Guezguez, PhD colleague from the University of Nice–Sophia Antipolis, who provided excellent editorial assistance.

ix

1

Introduction Richard Arena, Agnès Festré and Nathalie Lazaric

In 1937, Friedrich von Hayek wrote what was to become a very famous article, which was published in Economica, on the relations between economics and knowledge. It was admired by the economics profession, but its direct influence on economic theory at the time was limited. Fifty years later, with the emergence of the so-called ‘knowledge-based economy’, many of von Hayek’s preoccupations were revisited, and this has given birth to a large literature dedicated to the role of knowledge within economic relations. The economic reality questions the economic theory. The concept of the knowledge-based economy has generated a new ‘economics of knowledge’ or ‘economics of science’. This has prompted greater reflection on the notion of knowledge in analytical areas such as game theory, innovation theory, organization theory, firm theory, spatial economics and growth theory. However, it is not certain whether the numerous contributions on these issues have contributed to a better understanding of the key questions related to the notion of knowledge in economics.

1.1 THE MICROECONOMICS OF INFORMATION, KNOWLEDGE AND GENERAL ECONOMIC EQUILIBRIUM THEORY The research programme that dominated economic analysis for more than one hundred years – general economic equilibrium theory (GEET) – did not pay attention to the notion of knowledge, and instead focused on information. The argument put forward to justify this focus was that information could be measured. Information theory (see Shannon, 1948) emphasizes that information can and must be codified in order to be transmitted through a digital system. Van Ha (1999, p. 1) notes: information has the property of reducing the uncertainty of a situation. The measurement of information is thus the measurement of the uncertainty. That measurement is called Entropy. If entropy is large, then a large amount of information is required to clarify the situation. If entropy is small, then only a small amount of information is required for clarification.

1

2

Handbook of knowledge and economics

Within this framework, computer scientists interested in measuring the volume or weight of information to be transmitted refer to the minimal number of ‘bits’ needed to transmit some piece of information (a bit being the measure of the smallest amount of computer information storage). Microeconomic theorists, on the other hand, tend to maintain that knowledge cannot be measured. They generally do not consider practical means for measuring information, although they do concede that information can be coded and is measurable while knowledge is not. However, it should also be emphasized that some economists do not consider the notion of quantitative information to be relevant. For instance Arrow (1974, p. 38, quoted in Garrouste, 2001) stated: this definition of information is qualitative, and so it will remain for the purposes of this volume. The quantitative definition which appears in information theory is probably of only limited value for economic analysis, for reasons pointed out by Marschak; different bits of information equal from the viewpoint of information theory, will usually have very different benefits or costs. Thus let A and B be any two statements about the world, for neither of which is its truth or falsity known a priori. Then a signal that A is true conveys exactly as much information, in the sense of Shannon, as the statement that B is true. But the value of knowing whether or not A is true may be vastly greater than the value of knowing B’s truth-value; or it may be that the resources needed to ascertain the truth-value of A are much greater than those for B. In either case, the information-theoretic equivalence of the two possible signals conceals their vast economic difference.

Another reason why GEET research preferred the concept of information over the notion of knowledge is related to the characterization of this concept within Walrasian economics. In such a theoretical context information was considered objective and symmetric, that is, the same for all economic agents. It was seen also as complete, implying that the agents agreed perfectly on a common characterization of all possible states of the world. It was assumed to be perfect because it was being defined in a world where all the data related to problems of agent-individual choices are known. And finally, the combination of these properties was the basis for making individual rational choices. Even after the GEET research programme was discontinued, information or its equivalent – coded or codified knowledge – continued for some economists to be more attractive than other forms of knowledge. The ‘new economics of science’ emerged in the 1990s (Dasgupta and David, 1994; David and Foray, 1995; Cowan and Foray, 1997), an approach that combined mainstream microeconomic analysis with contributions from new institutionalism, and identified information as codified knowledge and treated it as a commodity.

Introduction

3

1.2 PERSONAL AND TACIT KNOWLEDGE There were some, however, who could not accept this identification, on the grounds, first, that some knowledge is tacit. In Polanyi’s Personal Knowledge, tacit knowledge is described as ‘unarticulated’ knowledge, which underlies ‘the aim of a skilful performance’ (Polanyi, 1962, chs 4 and 5), which aim ‘is achieved by the observance of a set of rules which are not known as such to the person following them’ (ibid., p. 49). Tacit knowledge cannot be reduced to these ‘rules’: ‘Rules of art can be useful, but they do not determine the practice of an art; they are maxims, which can serve as a guide to an art only if they can be integrated into the practical knowledge of the art. They cannot replace this knowledge’ (ibid., p. 50). There is no clear dichotomy between tacit and explicit forms of knowledge in Polanyi’s approach. Polanyi maintains that articulated or explicit knowledge always requires focal awareness since it implies a fully conscious attitude. However, if tacit knowledge requires subsidiary awareness, this is not to imply entirely unconscious behaviour: ‘it [tacit knowledge] can exist at any level of consciousness, ranging from the subliminal to the fully conscious. What makes awareness subsidiary is its functional character’ (Polanyi, 1975, p. 39). This explains why tacit assessments and judgements are required at every step in the acquisition of – even codified – knowledge (ibid., p. 31). From this point of view, there is no purely explicit knowledge; in other words, knowledge is always personal knowledge. Second, for Polanyi, the introduction of tacit knowledge is strongly related to ‘personal knowledge’. According to Polanyi, knowledge can be seen as the product of subjectivity. ‘Personal knowledge’ refers to knowledge anchored in individuals and is the product of personal commitment. For example, before the scientist becomes committed to ‘pure’ research he or she has a personal vision and an intuition, which are constrained by the tradition of the particular discipline. Some of the assumptions made, according to Polanyi, were due largely to the ‘logic of tacit inference’: Upon examining the grounds on which science is pursued, I saw that its progress is determined at every stage by indefinable powers of thoughts. No rule can account for the way a good idea is found for starting an inquiry, and there are no firm rules either for the verification or the refutation of the proposed solution of a problem . . . It appears then that scientific discovery cannot be achieved by explicit inference, nor can its true claims be explicitly stated. Discovery must be arrived at by tacit powers of the mind and its content, so far as it is indeterminate, can be only tacitly known. (Polanyi, 1964, p. 138)

Tacit knowledge, therefore, is rooted in personal knowledge and is generated through the specific engagement of the scientific (or any other) agent

4

Handbook of knowledge and economics

with his or her daily activity. This kind of tacit (‘pre-verbal’) knowledge is difficult to articulate. The articulation or codification of knowledge has been the subject of intense debate among economists (see Cowan et al., 2000).

1.3 SITUATED AND DISTRIBUTED KNOWLEDGE The third reason for the refusal to identify knowledge with information or codified knowledge was the distinction between situated and distributed knowledge. The theory of ‘situated cognition’ states that cognitive resources in the environment complement the cognition of agents and are exploited by them. Knowledge is anchored not only in the mind, but also physically in the environment. This theoretical proposition was developed by Suchman (1987), who emphasized that cognition is rooted inherently in action: that is, the physical, technological or social environment is essential for building human knowledge. Suchman’s analysis suggests that the spatial arrangement of the environment (notably a specific division of labour and local division of tasks) is decisive for understanding human problem-solving capabilities (Lorenz, 2001). Nooteboom (Chapter 15 in this volume), demonstrates why situated cognition departs from the representational vision of knowledge described by Newell and Simon (1964). It suggests that cognitive structure is not fixed, but is built in action, and that knowledge is local in character because it can be understood fully only within a specific context. In Chapter 15 Nooteboom quotes Polanyi (1962) in arguing: ‘Situated action entails that knowledge and meaning are embedded in specific contexts of action, which yield background knowledge, as part of absorptive capacity, which cannot be fully articulated, and always retain a “tacit dimension”.’ This vision is shared by advocates of the notion of ‘community of practice’ (Brown and Duguid, 1991; Lave and Wenger, 1991, Wenger and Snyder, 2000), proposed by researchers at the Palo Alto Institute for Research on Learning in the 1980s. A community of practice is defined as a group of people bound by informal links, engaged and interested in a common practice. They develop knowledge in action through practice and a shared language and common understandings, which most of the time remain tacit and implicit for most of the community. In Chapter 18 of this volume Amin and Cohendet discuss why community provides some degree of coordination during knowledge creation: Communities are thus ‘suppliers’ of sense and collective beliefs for the agents and play a central role of coordination in the organization. The community

Introduction

5

framework provides the context within which are built the collective beliefs and the reference systems that structure individual choice. Adopting the idea that knowledge creation is primarily realized in contexts of action and that the action is always collective, the consideration of the intermediate level of communities is thus necessary to focus on the learning in the processes of action (Dupouët and Laguecir, 2001).

Situated and distributed cognition are separate and complementary. What distinguishes these two visions of knowledge is the role of the cultural determinant in the cognitive process (Lorenz, 2001). Edwin Hutchins, a famous American researcher in the field, subscribes to these views. Hutchins (1986) sees cognition as occurring via technological artefacts and social interactions, and human cognition as being mediated by technological artefacts that act as external memory (part of the cultural heritage of humankind). Individuals in interaction with their environment solve problems and perform particular tasks by exploiting these technological tools. Cognition is mediated through such tools and distributed via artefacts through a specific ‘agencement’ and social interaction (e.g. in the US navy the channels for the transmission of knowledge are mostly formal rules and organizational relations). External memory affects the process of routinization by introducing new knowledge and new tasks into the division of labour. Artefacts create new kinds of memory that facilitate cognitive activities, and enable the articulation of formerly tacit practices, through common references (Lazaric et al., 2003).

1.4 SUBJECTIVE AND DISPERSED KNOWLEDGE The division of labour and dispersion of knowledge chimes with the Hayekian vision of cognition. In this perspective, knowledge is conceived not only as being distributed relative to one’s sensory-motor system, but also as being distributed in time and space (Lazaric and Lorenz, 2003). According to Hayek (1945), the dispersed and locally contextualized nature of knowledge makes it quite impossible to centralize all economic decision making. Hayek provides a subjectivist interpretation of this dispersion of knowledge based on two main reasons. The first, which is cognitive, is discussed in The Sensory Order (Hayek, 1952), where Hayek champions the idea that the brain functions in a connectionist way. This means that the point of departure for a mental representation is not the physical order of things, as ‘scientistic objectivism’ (to use Hayek’s expression – cf. Hayek, 1952, ch. V), would have it, ‘but the product of abstractions which the mind must possess in order to be capable of experiencing that richness of the particular [of the reality]’ (Hayek, 1978, p. 44). The conscious

6

Handbook of knowledge and economics

experiences that individuals regard as relatively concrete and primary and that are attributed to the intrinsic properties of the physical order ‘are the product of a superimposition of many “classifications” of the events perceived according to their significance in many respects’ (ibid., p. 36). Thus there are as many subjective forms of knowledge as there are individual ‘nervous systems’, that is, as there are individual heterogeneous agents. The second justification for Hayekian subjectivism is found in what Hayek calls the ‘social division of knowledge’. For Hayek, as a civilization develops, the knowledge of its society becomes more complex and specialized. However, no single agent has access to all this knowledge: it is dispersed within and among the individuals constituting society, who have access to very small parts of this social knowledge and especially to the processes by which social and economic activity is regulated and reproduced globally. Hayek’s subjectivist methodological choice led him to investigate the features of a ‘cognitive’ individual rationality. The cognitive capacities that individual agents must mobilize refer to their ‘mental maps’. Hayek describes these ‘maps’ as a ‘semi-permanent apparatus of classification’, which ‘provides the different generic elements from which the models of particular situations are built’ (Hayek, 1952, p. 89). The notion of a mental map conveys the idea of cognitive limits to the mental considerations of individuals. Rather than ‘a sort of schematic picture of the environment’, mental maps act as ‘a sort of inventory of the kinds of things of which the world is built up, a theory of how the world works’ (ibid.).

1.5 KNOWLEDGE AND RATIONALITY Nooteboom (2006) suggests that various visions for considering learning and knowledge can be endorsed. The French philosopher Blaise Pascal, writing in the seventeenth century, made the distinction between ‘ésprit de géométrie’, ‘which abstracts drastically from reality to enable grip for rigorous formal reasoning and an “ésprit de finesse”, which stays closer to complex reality, that allows less for formal analysis’ (Nooteboom, 2006, p. 3). Simonian and Hayekian interpretations of knowledge and information differ. Alan Newell and Herbert Simon (1964) developed the perfect illustration of ‘ésprit de géométrie’, that is, a classic statement of the information-processing or physical symbol system view of human cognition and knowledge. The basic premises of this approach are that knowledge consists of rule-based representations or collections of abstract symbols that are stored in the mind, and that problem solving can be understood in terms of search procedures that select among means to transform the initial into the goal state. This view of human knowledge

Introduction

7

and problem solving underlies Newell and Simon’s (1964, pp. 282–3) claim that, at the level of information processing, the computer and the human mind are comparable. They justify their epistemological stance by suggesting that computer simulation techniques can be used to provide psychologically realistic characterizations of human problem-solving behaviours, which contrasts with Hayek’s vision of knowledge as more strongly rooted in a traditional ‘ésprit de finesse’, that is, a vision of knowledge that goes beyond its symbolic representation. However, it is the Simonian representation of knowledge and the procedural rationality it legitimizes that contributes most to improving our deliberations over decision making. This symbolic approach to cognition is adopted explicitly by several economists, notably Egidi (1992, p. 154), for whom ‘a problem is represented by means of a symbolic structure . . . and finding a solution means finding the program or procedure which leads to a solution’. Drawing on Newell and Simon’s (1972) classic discussion of human problem solving, Egidi argues that in searching for a solution individuals use conjectures to decompose a problem into a set of presumably solvable sub-problems. This conjectural division of problem solving gives rise to a division of knowledge that is efficient because it economizes on memory and thinking. Herbert Simon’s information-processing approach to human cognition naturally gives rise to an understanding of knowledge and learning as symbolic expressions stored within the minds of the organization’s members. This symbolic way of storing and representing knowledge at the individual level may explain interference in the decisionmaking process in a context of bounded rationality. It refers to decision making in a context of incomplete information. 1.5.1

The Frame Effect

In economics, framing effects emerged in relation to observed occurrences of fairness in subjects’ behaviour in experiments. Frey and Bohnet (1995) suggest that we need to examine institutionalist elements to observe the impact of fairness on economic outcomes. Framing effects are defined as ‘norms, perspectives, contexts and other social cultural elements’ (Elliot et al., 1998, p. 456) and refer more generally to the way decisions are presented and how they shape human judgements in specific settings. Kahneman and Tversky (1979) suggest that framing effects are a preliminary stage that precedes the decision problem, the second stage being the period of evaluation. They define framing effects as ‘the manner in which the choice problem is presented .  .  . [according to the] norms, habits, and expectancies of the decision maker’ (Kahneman and Tversky, 1981, p. 455). Thus framing effects represent the heuristics interplaying in

8

Handbook of knowledge and economics

the decision-making process before the problem is solved (Gabaix et al., 2001). This means that, underlying many of our intuitive inferences, are representativeness, availability and anchoring. This is not a new idea. The social sciences refer to it as cognitive frameworks, which result from internal processes and the local and cultural environment (Bandura, 1986; Witt, 1999). Cognitive frameworks are the outcome of the co-constitution of action and perception, proposed in the constructivist approach (see notably Weick, 1979, on this dimension). For Boulding (1956), images play this role of intermediation between the perception of raw data and the internal value system. Every human action is induced by the person’s image, which, in turn, may be revised by the action. Images provide a way to interpret information and make sense of the environment. They create temporarily stable cognitive frameworks with individual and collective regularities. For instance, in Chapter 6 of this volume, Patalano says: ‘individual imagery has a relevant social function because it enables collective sharing of values and meanings . . . the image has cohesive power that may exert a strategic function in both organizational contexts and cooperative interaction.’

1.6 KNOWLEDGE, LEARNING AND ROUTINES In the historical evolutionary economics debate, collective learning rests on individual habits, routines and other types of more or less formalized practices (Commons, 1934; Veblen, 1914). Veblen developed an anthropological approach to capitalism and believed that it evolved with technical and social changes (Veblen, 1904, 1914). From this perspective, the question is not how a set of behaviours or actions becomes stable and balanced over time, but how it evolves (Veblen, 1919, p. 8). Individuals have certain habits and behaviours that are conditioned by experience (ibid., p. 79), which is why the cumulative and self-reinforcing process of a set of routines and habits on which the economic order rests needs to be depicted. These habits and propensities, embedded in social structures, tend to reproduce themselves, hence the potential for inertia. Interest in the notion of routines was reawakened by Nelson and Winter’s (1982) work, which highlights the relative permanence of firm behaviours, but also the capacity of firms to innovate. The notion of routine is increasingly used to analyse microeconomic change (Becker et al., 2005). Therefore a re-examination of the role of institutions would allow us to identify and understand the forces behind these changes, which are not related exclusively to cognitive contingencies (Nelson and Sampat, 2001). The interplay of the individual and the collective levels of action is

Introduction

9

far from neutral (Dopfer, 2007). For instance, entrepreneurs shape their judgements, beliefs and acts by themselves, but also in interaction with others. These micro interactions can produce ‘recurrent interacting patterns’ that need to be observed carefully (Cohen et al., 1996). Commons (1934, 1950) proposed an interactions taxonomy based on the type of knowledge involved (see Dutraive, Chapter 13 this volume). ‘Routine transactions’ are related to habitual activities involving stabilized knowledge (embodied in rules); ‘strategic transactions’ are those related to novel situations requiring new practices and implying new opportunities, for which there is no stabilized knowledge or rule of thumb. In other words, routine transactions are stabilized procedures that are deeply entrenched in the entrepreneur’s procedural memory, while strategic transactions are related to new ways of doing things, not yet classified by the human mind. For Commons, the processes of deliberation and calculation are not always mobilized, but may rely on past habits when they are appropriate. In certain circumstances, the mind may reveal ‘a creative agency looking towards the future and manipulating the external world and other people in view of expected consequences’ (Commons, 1934, p. 7; see also Hodgson, 1988). Thus institutions must be understood as the working rules of collective action that may restrain individual deliberation and can play a cognitive role by creating ‘institutionalized minds’ and ‘institutionalized personalities’ (Commons, 1934, p. 874). Both Commons and Veblen invite us to scrutinize the mechanisms of change brought about by the individuals (the ‘upward causation’ that has an impact on the organization), and the changes within the organization (the ‘reconstitutive downward causation’ that affects the individual) (Hodgson, 2007, p. 108). Routines lie between these two levels of analysis because they are enacted by individuals in a social context, which regulates the relative level of autonomy (Becker et al., 2005; see also Giddens, 1984). This interplay of the individual and collective dimensions is described in the literature in terms of entrepreneurs not always able to take the ‘best’ decision because of the amount of unreliable information. They may need to employ heuristics derived from other contexts in order to analyse the competitive structure of the environment (Porac and Thomas, 1990). The entrepreneur’s images are framed by collective actions within the local environment, which may ‘tie’ them, not because of the entrepreneur’s own cognitive limits, but because of the vast quantity of information available that may not be relevant to the decision involved. This may promote the adoption of mimetic behaviours to deal with the uncertainty in forming personal judgements (Greve, 1998). Mimetic local behaviour, in some circumstances, may avoid the necessity of weighing up all the possible actions (Kahneman, 2003), based on voluntary ignorance of some facts and data

10

Handbook of knowledge and economics

and also on the willingness to reduce the level of learning and the information search costs (Kirzner, 1979). This localized learning – induced by various vicarious learning processes – occurs at the level of industry and also at the local level (Maskell and Malmberg, 2007). It can induce a deliberate unwillingness to absorb new knowledge in order to avoid redefinition of deeply entrenched procedural knowledge to match the current vision (see Chapter 7 by Lazaric in this volume, on the discussion and definition of declarative and procedural knowledge). This willingness to continue with ‘routine transactions’ and steer clear of creating new ‘strategic transactions’ is exemplified by the famous exploration/exploitation dilemma (Levinthal and March, 1993; Greve, 2007). The compromise required shows that exploitation not only increases the probability of repeating organizational routines, but simultaneously avoids exploration by reducing the resources available for research. Innovation may arise from an innovator modifying current thinking on the economic activity thanks to the emergence of less stereotyped images in some specific context (see Chapter 6 by Patalano, this volume). Regularities are rooted in ‘cognitive automatisms’, which are generated by the stabilization of the ‘procedural knowledge’ that allows faster memorization in circumstances that appear to be similar (Bargh, 1997; Cohen and Bacdayan, 1994). These potential automatisms, which are rooted also in ‘declarative knowledge’, that is, the representational level, help human beings to identify predictable behaviour in dynamic environments and to integrate some plasticity into the solving of new problems not yet memorized (see Lazaric, 2008, for a longer discussion). Images are part of this system because they produce regularities inside the procedural knowledge as well as new insights in the declarative knowledge that are not always put into practice – that is, transformed by the mind into a purposeful cognitive act. Mindful reflexivity (Langer and Moldoveanu, 2000) and motivation related to organizational change are necessary, but not always sufficient, to overcome these obstacles (Howard-Grenville, 2005). This implies that motivational factors within current practices should accord with the change introduced at the cognitive level. The perception and image of change are crucial and relate to both the declarative and procedural forms of knowledge, that is, to the representation of change and its effective implementation. In this perspective, changes to routines should not be seen as fateful coincidences related to external and disruptive factors, but as ingredients crucial for the revitalization of individuals and organizations. This echoes recent research on organizations, about mindfulness, or attention to weak cues and learning from rare events (Rerup, 2005), and the place of mindful and less mindful attitudes as necessary for

Introduction

11

organizations to evolve and survive. Indeed, ‘learning to be surprised’, that is, reflection in action in a context of high-reliability organizations, seems vital for their adaptation (Jordan, 2010). In this latter approach, sense making is important at the individual and collective levels and is not always opposed to organizational routines (Rerup and Feldman, 2011). Thus attention becomes critical and induces some reflexivity on routines. Mindfulness matters and materializes by the intention and capacity to absorb change – at both the motivational and cognitive levels (Huet and Lazaric, 2008; Lazaric et al., 2008). In the field of interest here, a mindful attitude can be defined as the capacity to go beyond routine transactions in order to change the procedural knowledge embedded in entrepreneurs’ minds and ways of doing things. A mindful attitude is the explorative behaviour that must be adopted to generate a strategic transaction, that is, a transaction that is not always known in advance and that may trigger unpredictable change inside the organization.

1.7 CONTENT OF THE VOLUME Part I of this Handbook provides a historical perspective on how knowledge is dealt with in various economic traditions. Chapter 2 by Ludovic Ragni re-evaluates Pareto’s contribution to economics and sociology in the light of the current literature on the role of knowledge and beliefs in economic relations. More precisely, Pareto’s action theory is described as pioneering work in the field now referred to as cognitive or behavioural economics, and focused on how people acquire and treat information, and elaborate beliefs or ways of thinking by interacting with each other. It is interesting that Pareto’s focus on human behaviour is the result of a methodological perspective that tries to integrate other disciplines such as psychology or sociology into economic analysis. Brian Loasby’s contribution (Chapter 3) is devoted to Marshall’s view of knowledge and its centrality in his explanation of how economic systems work. Beyond the traditional reasons why Marshall focused on knowledge (observation of the remarkable industrial developments that occurred during his lifetime and that rested on the organization of the growth and application of knowledge, and his desire for improvements in the condition of the people), Loasby refers to Marshall’s ‘kind of intellectual crisis’ in discussing the sources and reliability of human knowledge. This crisis, which is documented by Tiziano Raffaelli (2003), led Marshall to develop his own model of an evolutionary, contingent and fallible process by which the human brain could develop classification systems

12

Handbook of knowledge and economics

for interpreting phenomena and planning action. Like Hayek, Marshall emphasized the limitations to human cognition and the importance of organization in the coordination of economic activities. In Chapter 4, Agnès Festré discusses the role of knowledge in the Austrian tradition. While it has become common now to refer to Hayek’s pioneering work on the relation between economics and knowledge (see Section 1.1 above), other Austrian economists have not endured to the same extent, although their contributions to our understanding of how knowledge moulds behaviour and helps to coordinate economic activities are far from being negligible. It is especially interesting that the work of the founding father of the Austrian School, Carl Menger, paved the way to various attempts to deal with the role of knowledge in economic activities. This chapter contrasts Menger’s conception of knowledge with that of his direct successor, Friedrich von Wieser, and shows that, although they had a shared interest in how institutions emerge in an environment characterized by individual heterogeneity, time and spatial constraints, they developed divergent perspectives of institutional dynamics. Along similar lines, Véronique Dutraive argues in Chapter 5 that old American institutionalism, in particular Veblen and Commons, anticipated some of the trends of contemporary economic analysis in dealing with the interactions between knowledge, cognition and institutions (e.g. Denzau and North, 1994). Dutraive stresses that Veblen and Commons, building on American pragmatist philosophy, pioneered the focus on the importance of the interaction with institutions and mental processes for our understanding of the dynamics of economic phenomena in modern societies, and made early claims that economics must interact  with  other  sciences – and particularly with the psychological sciences. Roberta Patalano’s contribution (Chapter 6) analyses more deeply how mental representations and knowledge interfere, by focusing on Kenneth Boulding’s theory of action in The Image (Boulding, 1956), a work often neglected by economists, which is based on perception and imagination. Patalano makes a comparison with Hayek’s theory of knowledge developed in The Sensory Order (1952), to show that both authors anticipated some modern developments in economics (cognitive economics, neuroeconomics), and to emphasize the relevance of the neuro-psychological and psychic underpinnings of economic behaviour that emerged over half a century ago. While some of the ideas proposed by Boulding, and especially by Hayek, have been developed using modern instruments, others have been neglected and would be worth rediscovery. Roberta Patalano argues that the most significant one might be imagination, that is, the attitude of mind involved in framing situations and developing images of

Introduction

13

what might happen in the future. She argues that this notion should be investigated in order to clarify its influence, in economic terms, on individual and collective behaviour. In Chapter 7 Salvatore Rizzello and Anna Spada focus on the debates of the 1950s that opposed the defenders of rational choice theory and those interested in developing new tools, which constituted the ground for a psychology-based theory of decision. Rizzello and Spada argue that the problem of uncertainty in decision making is at the core of the debate on economic decision making, while knowledge is often regarded as implicit, that is, as a method of facing uncertainty, or is neglected. Here, Simon’s work is illuminating because he regarded knowledge as connected to procedural rationality. Unlike those GET economists who focused on information and risk (see Section 1.1 above), Simon took up the challenge to address the problem of uncertainty and developed a theory of decision making grounded on the concept of ‘pragmatic’ rationality, which is built around human knowledge. Part II of the book deals with the conceptions, role and use of knowledge in economics in general. Chapter 8 by Giovanni Dosi provides an overview of the contribution of economic theory to the understanding of knowledge-based economies, observing that all economies that we know were profoundly knowledgebased as much as a century ago and are still so today. However, he also argues that there is a need to develop an adequate toolkit in order to identify what distinguishes the contemporary role of knowledge (in relation to basic economic mechanisms of demand formation, accumulation, employment generation etc.) from what Marshall and Schumpeter were observing a century earlier. Although recent developments in the economics of information and of innovation have brought important insights into the processes of generation and diffusion of knowledge, and their economic consequences, many streams of macroeconomic analysis are being very slow to adopt them. Massimo Egidi’s contribution (Chapter 9) discusses the many attempts to provide cognitive foundations to the limitations of (conscious) rationality. He points out that this trend has been hampered by Harrod’s (1939) evolutionary justification of marginalist economic rationality, and Friedman’s (1953) positivist methodology (his ‘as if ’ hypothesis), which completely disregards the psychological aspects of decision making since, in this view, even low individual awareness is not supposed to be incompatible with full rationality. This ‘cognitive gap’ could be reduced, even in an evolutionary perspective, through the provision of cognitive foundations to a bounded-rationality approach to decision making. The solution offered along these lines is related to the question of consciousness

14

Handbook of knowledge and economics

and, ultimately, to the distinction between tacit and explicit knowledge. This distinction may be cognitively grounded through reference to the distinction between conscious and effortful and automatic and effortless reasoning, as described by many cognitive psychologists such as Daniel Kahneman. Chapter 10 by Pier Paolo Saviotti provides a characterization of the processes of knowledge generation and utilization, something that is missing from the economic literature despite the growing interest in knowledge for innovation and economic development in the so-called knowledge-based societies. This description uses a theoretical framework to represent, model and measure knowledge. It is based on two properties of knowledge: (a) as a co-relational structure, meaning that knowledge generally involves connections between variables making it possible to deduce the value of unknown variable from the value of known or linked variables; (b) as a retrieval/interpretative structure. This is not a complete representation of knowledge, but is intended to help interpret the collective processes of knowledge creation and utilization involving different types of organizations (firms, public research institutes, universities etc.) and taking place in knowledge-based economies. Chapter 11 by Jacques Durieu and Philippe Solal provides an overview of the literature on learning in evolutionary game theory. The hypotheses commonly related to game-theory models are bounded rationality justified by lack of information about the game structure (payoff functions or the rationality of other players), and a stationary environment in order to simplify the decision task. The authors distinguish between two kinds of models depending on hypotheses concerning the behaviour of the other players: one family of models assumes that agents do not elaborate their beliefs about their opponents’ behaviour; the other considers that agents form (naive) expectations about their opponents’ future play. For each category of adaptive learning models, the authors show that repetition of physical interaction among agents can overcome the problems of lack of strategic information and the limitations of rationality, suggesting that knowledge acquisition is embedded in interindividual interactions. Chapter 12 by Dominique Foray deals with experiential knowledge defined as a kind of knowledge that springs from the experience of individuals and organizations, which is local and specific, sound, rational and effective, although it does not have the status of scientific knowledge. Knowing how wind flows vary can help to avoid forest fires is an instance of experiential knowledge. The properties of experienced knowledge (it is local, disturbing and disruptive) lead to specific problems (deterioration, disinvention and deactivation) that can jeopardize the community, in

Introduction

15

particular when experienced knowledge is collective (to be distinguished from private experienced knowledge, like a trade secret). Bengt-Åke Lundvall (Chapter 13) provides a critical assessment of the policy maker’s concept of knowledge base, and proposes the idea of numerous separate knowledge pools that constitute ‘community resources’ that are not easily transformed into private property. This concept of knowledge pools could inspire innovation policy in both the more and less developed parts of the world. For example, in developing countries there is a need to build absorptive capacity in order to access the knowledge pools in the richer parts of the world, and an accompanying general need to reconsider the rules of the game related to intellectual property rights. In the rich countries, finding ways to connect specific separate pools of knowledge could be seen as key to stimulating radical innovation by exploiting knowledge diversity. André Orléan’s contribution (Chapter 14) is a tribute to Keynes’s outstanding work on collective knowledge or beliefs and their criticality for economic decision making in conditions of uncertainty. This approach is at odds with economic and financial theory that relies on objective values (e.g. the fundamental value of securities) and deriving efficiency theorems (e.g. the informational efficiency hypothesis of Fama), and takes it for granted that those objective values exist. This oversimplification explains why many financial models cannot provide satisfactory explanations for phenomena such as financial bubbles. It is also not consistent with the idea that (financial) markets improve or even transform the functioning of the economy. Orléan advocates for a conventionalist approach to finance, which differs from the standard approach that considers the knowledge that agents are capable of producing in relation to the future development of the economy. Rather than considering finance as an a priori fact resulting from an objectively defined future, we should conceive it as the contingent product of opinion-based reasoning. Part III extends the contributions in Part II, focusing more closely on the role of knowledge in organizations. Bart Nooteboom’s contribution (Chapter 15) is centred on the debate over the notion of ‘embodied cognition’ (cf. Section 1.3 above). The notion of embodied cognition refers to Polanyi’s concept of ‘personal knowledge’ defined as knowledge rooted in an agent’s body that is physically positioned and interacting with the world. Consequently, the embodiment of cognition entails a continuum rather than a Cartesian duality between rational evaluation, feelings and underlying physiological processes in the body. This perspective has far-reaching implications for economics and management, and enables improved understanding of the ‘knowledge

16

Handbook of knowledge and economics

economy’ and the ‘network economy’. Clearly, if knowledge arises from interaction with others, in a ‘knowledge economy’ interaction between firms in networks becomes crucial. In Chapter 16, Ulrich Witt, Tom Broekel and Thomas Brenner offer a conceptual clarification of the economic properties of knowledge. They evaluate the many distinctions that economists have used to describe the characteristics of knowledge, starting from Polanyi’s distinction between implicit and explicit knowledge, through the difference between encoded and non-encoded knowledge, to the different categorizations of knowledge as a public good, a locally public good or a private good. In accordance with evolutionary approaches, the authors show that the characteristics of knowledge should not be viewed as intrinsic since they depend strongly on the state of the knowledge technology, that is, on how knowledge can be acquired, stored, used and communicated. Paul Nightingale focuses in Chapter 17 on the notion of tacit knowledge, discussing its topicality and relevance, mentioning the many findings in the neurosciences that support it, exemplified by experiences led by Damasio (1994) on ‘somatic markers’ and Edelman (1992) on learning. Nightingale argues, in line with Polanyi’s reasoning, that the concept of tacit knowledge should be used to move explanations outwards from agents, and to become a foundation for understanding the more complex causal processes at work, rather than a variable that explains everything. He concludes that tacit knowledge is a useful concept when used properly, but its flexibility means that it can be used to explain ‘anything’ or to justify any policy position. In Chapter 18 Ash Amin and Patrick Cohendet discuss knowledge shared by communities of practice (Brown and Duguid, 1991), defined as a group of people bound together by informal links engaged in a common practice. The community framework provides the context within which collective beliefs and reference systems that structure individual choice are built. This supports the conception of knowledge creation anchored in collective action, and its consideration permits a better understanding of the learning processes at work at intermediate levels in the firm’s organizational structure. This has strong implications for knowledge management. For instance, it could mean that firms should devote great attention and energy in order to benefit from the ‘spontaneous or intentional emergence of the cognitive platforms’ that are vital for innovation and creativity inside the firm. Mie Augier and Thorbjørn Knudsen (Chapter 19) take up the challenge of modelling knowledge organization by introducing a new, unifying way of thinking about the organization of knowledge. The organization of knowledge is conceived as an architecture whose design

Introduction

17

requires consideration of the cognitive skills of potential employees, the distribution of alternatives available in the business environment, and the costs associated with alternative modes of employment. Like Simon and March, Augier and Knudsen view organizations as helping economic agents to take decisions by constraining the set of alternatives. In a context of low communication costs and increased connectivity of various media, phenomena commonly associated with the knowledge economy, an investigation of what kind of architecture is the most appropriate to help boundedly rational agents to make better choices and avoid costly, irreversible decisions is a crucial theoretical and empirical issue. Chapter 20 by Markus Becker offers an in-depth investigation of the implications of distributed knowledge for organizations, distinguishing between different types of architectures, that is, different ways of linking people depending on the degree of specialization and the degree of overlap in the knowledge held by agents. A serious difficulty arises in the attempt to disentangle the effects due to the relation between the dispersion of knowledge and the division of labour from the whole set of non-ambiguous effects of the dispersion of knowledge on organizations. Although the dispersion of knowledge has been acknowledged by many economists (in particular Smith and Hayek) as a fundamental theoretical issue, and has given rise to many applications in management (e.g. Taylorism), the problem of its coordination (including its coordination in time) is deserving of more attention. The Handbook concludes with a chapter by Nathalie Lazaric that explores what drives change in knowledge. Lazaric draws on Anderson’s (1983) distinction between declarative and procedural memory, and recent findings in the cognitive sciences. She tries to disentangle the cognitive mechanisms by which declarative memory, that is, a form of memory that is focused mainly on the recollection of facts or events, can be converted into procedural memory, that is, a form of memory that concerns how things are done or the knowledge that is put to use. Both kinds of memory are subject to change and are intertwined, corroborated by the work of Shiffrin and Schneider (1977) and Kahneman (2003) on the relation between automatic and deliberately controlled forms of cognitive processes. Lazaric draws an analogy between the individual and the organizational levels. Although individual and organizational forms of memorization are distinct, their theorization involves similar difficulties: how are representations made to change? How can a repertoire of knowledge that is used daily be changed and improved? And how can new knowledge be created?

18

Handbook of knowledge and economics

REFERENCES Anderson, J.R. (1983), The Architecture of Cognition, Cambridge, MA: Harvard University Press. Arrow, K. (1974), The Limits of Organization, New York and London: W.W. Norton & Company. Bandura, A. (1986), Social Foundations of Thought and Action: A Social Cognitive Theory, Englewood Cliffs, NJ: Prentice-Hall. Bargh, J. (1997), ‘The automaticity of everyday life’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 1–61. Becker, M., N. Lazaric, R.R. Nelson and S.G. Winter (2005), ‘Toward an operationalisation of the routines concept’, Industrial and Corporate Change, 14 (5), 775–91. Boulding, K.E. (1956), The Image. Knowledge in Life and Society, Ann Arbor, MI: University of Michigan Press. Brown, J.S. and P. Duguid (1991), ‘Organizational learning and communities of practice: towards a unified view of working, learning and innovation’, Organization Science, 2 (1), 40–57. Cohen, M.D. and P. Bacdayan (1994), ‘Organisational routines are stored as procedural memory: evidence from a laboratory study’, Organization Science, 5 (4), 554–68. Cohen, M.D., R. Burkhart, G. Dosi, M. Egidi, L. Marengo, M. Warglien and S. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5 (3), 653–97. Commons, J.R. (1934), Institutional Economics: Its Place in the Political Economy, New Brunswick, NJ: Transaction Publishers. Commons, J.R. (1950), The Economics of Collective Action, New York: Macmillan. Cowan, R. and D. Foray (1997), ‘The economics of codification and the diffusion of knowledge’, Industrial and Corporate Change, 6, 595–622. Cowan R., P.A. David and D. Foray (2000), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 9 (2), 211–53. Damasio, A. (1994), Descartes’ Error: Emotion, Reason, and the Human Brain, New York: Putnam’s. Dasgupta, P. and P. David (1994), ‘Toward a new economics of science’, Research Policy, 23, 487–521. David, P.A. and D. Foray (1995), ‘Accessing and expanding the science and technology knowledge base’, STI Reviews: OECD – Science, Technology, Industry, No. 16, 13–68. Denzau, A. and D.C. North (1994), ‘Shared mental models: ideologies and institutions’, Kyklos, 47 (1), 3–31. Dopfer, K. (2007), ‘The evolutionary foundations of behavioural economics: Leibenstein’s legacy’, in R. Frantz (ed.), Renaissance in Behavioral Economics, Abingdon and New York: Routledge, pp. 59–91. Edelman, G. (1992), Bright Light, Brilliant Fire: On the Matter of the Mind, New York: Basic Books. Egidi, M. (1992), ‘Organizational learning, problem solving and the division of labour’, in Herbert Simon, Massimo Egidi, Robin Marris and Ricardo Viale (eds), Economics, Bounded Rationality and the Cognitive Revolution, Aldershot, UK and Brookfield, US: Edward Elgar, pp. 148–73. Elliott, C.S., D.M. Hayward and S. Canon (1998), ‘Institutional framing: some experimental evidence’, Journal of Economic Behavior and Organization, 35, 455–64. Frey, B.S. and I. Bohnet (1995), ‘Institutions affect fairness: experimental investigations’, Journal of Institutional and Theoretical Economics, 151, 286–303. Friedman, M. (1953), Essays in Positive Economics, Chicago, IL: University of Chicago Press. Gabaix, X., D. Laibson, G. Moloche and S. Weinberg (2001), The Allocation of Attention: Theory and Evidence, MIT: mimeo. Garrouste, P. (2001), ‘What economics borrows from the statistical theory of information?’,

Introduction

19

in P. Petit (ed.), Economics and Information, Boston, MA: Kluwer Academic Press, pp. 33–48. Giddens, A. (1984), The Constitution of Society: Outline of the Theory of Structuration, Berkeley and Los Angeles, CA: University of California Press. Giddens, A. (1994), Beyond Left and Right, Cambridge: Polity Press. Greve, H.R. (1998), ‘Performance, aspirations, and risky organizational change’, Administrative Science Quarterly, 44, 58–86. Greve, H.R. (2007), ‘Exploration and exploitation in product innovation’, Industrial and Corporate Change, 1 (5), 945–75. Harrod, R.F. (1939), ‘Price and cost in entrepreneurs’ policy’, Oxford Economic Papers, 2, 1–11. Hayek, F.A. (1945), ‘The use of knowledge in society’, American Economic Review, XXV (4), 519–30. Hayek, F.A. (1952), The Sensory Order. An Inquiry into the Foundations of Theoretical Psychology, London: Routledge and Kegan Paul. Hayek, F.A. (1978), ‘The primacy of the abstract’, in New Studies in Philosophy, Politics, Economics and the History of Ideas, London and Henley: Routledge & Kegan Paul, pp. 35–49. Hodgson, G.M. (1988), Economics and Institutions: A Manifesto for a Modern Institutional Economics, Philadelphia, PA: University of Pennsylvania Press. Hodgson, G.M. (2007), ‘Institutions and individuals: interaction and evolution’, Organization Studies, 28 (1), 95–116. Howard-Grenville, J.A. (2005), ‘The persistence of flexible organizational routines; the role of agency and organizational context’, Organization Science, 6, 618–36. Huet, F. and N. Lazaric (2008), ‘Capacité d’absorption et d’interaction: une étude de la coopération dans les PME française’, Revue d’économie Industrielle, No. 121, 60–75. Hutchins, E. (1995), ‘How a cockpit remembers its speeds’, Cognitive Science, 19 (3), 265–88. Jordan, S. (2010), ‘Learning to be surprised: how to foster reflective practice in a highreliability context’, Management Learning, 41 (4), 390–412. Kahneman, D. (2003), ‘Maps of bounded rationality: psychology for behavioral economics’, American Economic Review, 93 (5), 1449–75. Kahneman, D. and A. Tversky (1979), ‘Prospect theory: an analysis of decision under risk’, Econometrica, 47 (2), 263–92. Kahneman, D. and A. Tversky (1981). ‘The framing of decisions and the psychology of choice’, Science, 211, 453–8. Kirzner, I.M. (1979), Perception, Opportunity, and Profit: Studies in the Theory of Entrepreneurship, Chicago, IL: University of Chicago Press. Langer, E.J. and M. Moldoveanu (2000), ‘The construct of mindfulness’, Journal of Social Issues, 56 (1), 1–9. Lave, J. and E. Wenger (1991), Situated Learning: Legitimate Peripheral Participation, Cambridge: Cambridge University Press. Lazaric, N. (2008), ‘Routines and routinization: an exploration of some micro-cognitive foundations’, in M. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 205–27. Lazaric, N. and E. Lorenz (eds) (2003), ‘Introduction’, in Knowledge, Learning and Routines, Critical Studies in Institutions, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. ix–xxiii. Lazaric, N., P.A. Mangolte and M.L. Massue (2003), ‘Articulation and codification of collective know-how in the steel industry: some evidence in the French blast furnace’, Research Policy, 32, 1829–47. Lazaric, N., C. Longhi and C. Thomas (2008), ‘Gatekeepers of knowledge versus platforms of knowledge: an illustration with the case of a high tech cluster’, Regional Studies, 42 (16), 837–52. Levinthal, D.A. and J.G. March (1993), ‘The myopia of learning’, Strategic Management Journal, 14, 95–112.

20

Handbook of knowledge and economics

Lorenz, E. (2001), ‘Models of cognition, the contextualisation of knowledge and organisational theory’, Journal of Management and Governance, 5 (34), 203–31. Maskell, P. and A. Malmberg (2007), ‘Myopia, knowledge development and cluster evolution’, Journal of Economic Geography, 7 (5), 603–18. Nelson, R. and S. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Belknap Press of Harvard University Press. Nelson, R. and B. Sampat (2001), ‘Making sense of institutions as a factor shaping economic performance’, Journal of Economic Behavior and Organization, 44, 31–54. Newell, Allen and H.A. Simon (1964), ‘Simulation of cognitive processes’, Proceedings of the 17th Congress of Psychology, Washington, DC, pp. 298–9. Newell, A. and H. Simon (1972), Human Problem-Solving, Englewood Cliffs, NJ: Prentice-Hall. Nooteboom, B. (ed.) (2006), Knowledge and Learning in Organisations, part II, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Polanyi, M. (1962), Personal Knowledge, Chicago, IL: University of Chicago Press. Polanyi, M. (1964), Personal Knowledge: Toward a Post-Critical Philosophy, New York: Harper and Row. Polanyi, M. (1975), ‘Personal knowledge’, in M. Polanyi and H. Prosch (eds), Meaning, Chicago, IL: University of Chicago Press, pp. 22–45. Porac, J.F. and H. Thomas (1990), ‘Taxonomic mental models in competitor categorization’, Academy of Management Review, 15, 224–40. Raffaelli, T. (2003), Marshall’s Evolutionary Economics, London and New York: Routledge. Rerup, Claus (2005), ‘Learning from past experience: footnotes on mindfulness and habitual entrepreneurship’, Scandinavian Journal of Management, 21, 451–72. Rerup, C. and Feldman, M. (2011), ‘Routines as a source of change in organizational schemata: the role of trial-and-error learning’, Academy of Management Journal, 54 (3), 577–610. Shannon, Claude (1948), ‘Communication theory of secrecy systems’, Bell System Technical Journal, 28 (4), 656–715. Shiffrin, R.M. and W. Schneider (1977), ‘Controlled and automatic human information processing: II Perceptual learning, automatic attending and a general theory’, Psychological Review, 84 (2): 127–90. Suchman, L.A. (1987), Plans and Situated Actions: The Problem of Human–Machine Communications, Cambridge, UK: Cambridge University Press. Van Ha, K. (1999), ‘Information measurement’, Høgskolen i Østfold, No. 1, 1–99. Veblen, T. (1904), The Theory of Business Enterprise, New York: Charles Scribner’s Sons. Veblen, T. (1914), The Instinct of Workmanship, New York: Macmillan. Veblen, T. (1919), The Place of Science in Modern Civilisation and Other Essays, New York: Huebsch. Weick, K. (1979), The Social Psychology of Organizing, New York: McGraw-Hill. Wenger, E. and W. Snyder (2000), ‘Communities of practice: the organizational frontier’, Harvard Business Review, January–February, 139–45. Witt, Ulrich (1999), ‘Do entrepreneurs need firms? A contribution to a missing chapter in Austrian economics’, The Review of Austrian Economics, 11 (1–2), 99–109.

PART I KNOWLEDGE AND ECONOMICS: A HISTORICAL PERSPECTIVE

2

What Vilfredo Pareto brought to the economics of knowledge Ludovic Ragni

What economists generally remember about Vilfredo Pareto are his contributions to general equilibrium theory, his definition of optimality and the law that bears his name. Economists refer to Pareto much less as a sociologist. This can be seen as a deficiency to the extent that his sociology actually determines his contributions to the fields of pure and applied economics from a methodological point of view. Indeed, according to Pareto, economics must answer to the method of ‘successive approximations’. This consists in grasping economic phenomena first inductively, then deductively, so as to be able to offer a mathematically logical explanation, as is done in the experimental sciences. Then the explanation is completed by taking into account whatever the other social sciences have contributed, especially sociology. A synthesis of these differing approaches aims at producing an action theory that would help explain the way economic and social agents think. Pareto sets out to explain how the actors reason either from an individual point of view or from that of social group dynamics. The reach of his study of this problem can be seen in the theory of logical and non-logical actions developed by the author, especially in his Courses (1896–97), Manuale (1909) and Treatise (1916).1 Paretian theory can thus be seen as a reflection on the knowledge economy considered as a field of thought relating to the way agents think, treat, and acquire information and elaborate ways of thinking and belief.2,3 It is the result of the interdependence of disciplines as varied as economics, psychology, logic and sociology, to which we would today add the cognitive sciences. Overall the Paretian action theory constitutes a pioneering framework of the ‘economics of the mind’,4 the importance of which has too often been underestimated. This chapter aims to revisit this dimension of Pareto’s work so as to evaluate its reach in the field of economics. We will show that his works already announce some of his thinking in an effort to try to explain the forms of rationality suitable to economic agents. To do so, we shall assess the epistemological foundations of the hypotheses put forward by Pareto to explain the behavior of economic agents as they set up economic models. Four complementary themes will be touched upon. 23

24

Handbook of knowledge and economics

The first will try to show that Paretian action theory, and therefore knowledge, rests on a methodology that implies that economics as a discipline depends upon the other social sciences. In the second section, we shall indicate that Paretian knowledge theory involves the different concepts of rationality that were systematized later by Simon (Simon, 1976, 1978, 1986, 1990, 1996). Thus we shall show why Pareto’s action theory can be interpreted in the light of concepts such as substantive, limited, procedural or even cognitive rationality. A third stage will lead us to examine how the complex categories that Pareto defines to explain the behaviors of social agents – ‘interests’, ‘derivations’, ‘residues’ and ‘elites’ – represent the ways by which they acquire knowledge in a complex interactive situation. We shall end by showing that Pareto’s representation of action goes far beyond that which is normally associated with homo œconomicus; in fact, it provides the basis of his analysis of the group dynamics needed to reach equilibrium. We shall illustrate this point by analyzing how Pareto explains the alternative between free trade and protectionism, taking as a starting point the process of ‘equilibration’ the economic, social and institutional forms knowledge takes.

2.1 PARETIAN METHODOLOGY AND ACTION THEORY Being able to evaluate Paretian action theory in terms of knowledge economics implies understanding its methodological foundations. Shedding light on this method will then allow us to investigate his action theory from the perspective of the forms of rationality that are usually discussed in today’s economic analysis. Pareto’s methodology can be described by what he calls the logicoexperimental approach, and applies to both economics and sociology (Pareto, 1981, p. 16). In an effort to grasp social phenomena, the method involves three steps. The first is inductive: it consists in the observation of behavioral patterns or facts that display some regularities. The second is deductive: it is concerned with the working out of specific assumptions from which models explaining the laws gathered from previous observations are deduced. These models belong to the fields of economics, sociology and other social sciences. The third step synthesizes the various models so as to reach an overall explanation. This process is best illustrated by Pareto’s analysis of action theory, which consists in the investigation of the way economic and social agents

What Vilfredo Pareto brought to the economics of knowledge 25 act when trying to reach a goal. Pareto’s contention is that economics should be concerned especially with logical actions, whereas sociology and psychology should concentrate on non-logical ones.5 His Treaty on General Sociology brings these different points of view together. Pareto’s methodology is also characterized by the use of what he calls ‘the method of successive approximations’. Regarding action theory, it corresponds to a method that grasps individuals’ behavioral patterns and facts from the twofold perspective of subjective and objective knowledge, that is, respectively, from the perspective of how people see and how people produce knowledge when acting. The first stage consists in observing enough facts in order to ‘purify’ them. On the basis of these purified elements, scientists can then inductively formulate hypotheses or discover regularities in economic phenomena. They are then able to deduce a theory that can later be confirmed on the basis of these hypotheses or uniformity principles. The suggested models, when dealing with pure economics, are conceived as a first approximation of reality. At this stage, the agents’ behavior meets the principles of pure rationality. During the second stage, the same phenomena are studied in all their complexity, especially allowing for their social, cultural and emotional dimensions. Economic and sociological laws are then considered to be tendentious regularities. At this point, Pareto completes the pure economic or mathematical models by enriching them with input from the other social sciences, thereby making a synthesis of those points of view that should be kept (Freund, 1974; McLure, 2001; Bruni, 2002). This method assumes that the results of pure economics fit in with and depend upon the sociological ones. Logical actions that are confined to pure economics must therefore be made consistent with non-logical actions that concern both economics and sociology (Pareto, 1968, pp. 44, 63). Such a global perspective then enables us to show, for instance, that free trade or protectionism results from both features of pure economics, such as logical actions and pure rationality, and sociological features such as social stratification, illogical actions based on ethical or religious beliefs or bounded or satisficing rationality. Paretian ‘knowledge economics’ thus aims at making synthetic models of logical and illogical actions so as to explain the cognitive processes of social actors at work when pursuing a specific aim. Consequently, Pareto’s method is first of all both inductive and a priori. It is then deductive to the extent that Pareto seeks to set general laws and tendencies based on observations from which results or forecasts are then deductible. Moreover, Pareto’s methodology relates to the logico-experimental method. This method is experimental in the sense that it is based on prior

26

Handbook of knowledge and economics

observations and is combined with models that are tested on the basis of further observation. It is logical in the sense that it follows the usual scientific procedure, that is, to deduce logically from assumptions the models used to make predictions. Finally, it takes a global (‘synthetic’) perspective because, by using a series of approximations, it permits us to deal with both economic and sociological models in a unified framework focused on the driving forces of human action. Pareto’s action theory therefore goes much further than the theory of homo œconomicus as described by standard economic theory. It indeed seeks to take into account complex behavioral patterns, explaining them by adopting the complementary angles of homo œconomicus and homo sociologicus. It is obviously seeking to account for individual rational acts, agents’ limited cognitive skills as well as their individual and collective beliefs. Within this context, logical actions more specifically concern pure economics, be it mathematical or applied, while non-logical actions are the concern of sociology: ‘the study of non-logical actions thus belongs mainly to sociology’ (Pareto, 1967b, p. 8). It is not, however, always possible to discriminate precisely between those actions that fall within the province of economics and those that concern sociology. The author states in the Manuale that logical actions mainly concern pure economics, whereas in the Treaty he stresses that some economic actions are actually non-logical and some sociological non-logical actions must be taken into account to better explain economic activity.

2.2 LOGICAL AND NON-LOGICAL ACTIONS AS CONVEYING DIFFERENT KINDS OF KNOWLEDGE According to Pareto, for an action to be logical, two conditions must be met. The first is that it must be logical from both the objective scientific point of view and that of the agent carrying it out. Objectivity is thus defined according to current available scientific knowledge, while subjectivity relates to the perception social actors have of the phenomena under consideration. The second condition implies that, for an action to be logical, both the objective and subjective goals must correspond to each other, or in other words, the means and the reasoning process used to attain the goal must be objectively and subjectively identical. Obviously, logical actions deal with instrumental and substantive rationality as Simon (1945, 1976, 1978 and 1990) defined it. According

What Vilfredo Pareto brought to the economics of knowledge 27 to this author, appropriate behavior in order to achieve a goal under a known system of constraints is substantively rational. Simon, in his Administrative Behavior (1945), goes so far as to define substantive rationality as all those actions to which the principle of objective rationality can be applied from both the agent’s and the scientist’s perspectives. How very close the definitions suggested by Simon as related to substantive rationality and those given by Pareto concerning logical actions are thus becomes obvious. More precisely, and to insist even more on the relationship with Pareto, we need to recall that Simon, in his Administrative Behavior, uses the term ‘objective rationality’ for both the agent and the scientist when speaking about substantive rationality. In his Manuale, Pareto still specified that ‘non-logical action does not mean illogical; a non-logical action may be one which a person could see, after observing the fact and the logic as the best way to adapt the means to the end; but that adaptation has been obtained by a procedure other than that of logical reasoning’ (1981, p. 41). In the Treatise an action is non-logical when ‘the objective end differs from the subjective purpose’ (Pareto, 1935, p. 78). Pareto distinguishes four kinds (genera) of non-logical actions. 1. 2. 3. 4.

There is no logical end, subjectively or objectively. There is a logical end subjectively but not objectively. The end is objectively logical but not subjectively. There is a logical end subjectively and objectively.

Pareto subdivides the third and fourth types (genera) into two sub-types (species): sub-types 3a and 4a, which are characterized by the fact that the objective end could be accepted by the subject if he knew it; and sub-types 3b and 4b, where the objective end would be rejected by the subject if he knew it. Table 2.1 summarizes the typology Pareto developed in his Treatise in 1916. Pareto’s action theory is thus of an intentional nature because it assumes that the individual seeks to reach a goal even if at the outset such a goal is neither totally defined nor conscious (Dennett, 1981, 1987). This concerns mainly the logical and non-logical actions of types 2 and 4. The intentional nature is less clear regarding the non-logical actions of types 1 and 3. The non-logical actions of the first and third types indeed assume that the subjects are not seeking a subjective goal. Are they therefore less interesting in economics? They are either precepts or forbidden acts having no explanations for the individual or for science. Type 1, for example, concerns politeness or customary behavior. Type 3 covers principally innate

28

Handbook of knowledge and economics

Table 2.1

Pareto’s typology of actions

Genera and species

Have the actions logical ends and purposes  Objectively?

Subjectively?

Class I: Logical actions The objective end and the subjective purpose are identical YES YES Class II: Non-logical actions The objective end differs from the subjective purpose NO NO Genus 1 YES NO Genus 2 NO YES Genus 3 YES YES Genus 4 Species of the genera 3 and 4 The objective end would be accepted by the subject 3a, 4a if he knew it The objective end would be rejected by the subject 3b, 4b if he knew it

acts or reflex behavior. These actions are indeed efficient, often reaching the desired goal without the subject having thought it out or been aware of it. Type 2 and 4 non-logical actions lead to a discussion of the nature of the means used, that is, to the knowledge and the reasoning processes put into play by individuals to reach an objective. Indeed, if the objective and subjective goals do not correspond, the existence of a subjective aim can help evaluate the type of rationality the subjects used. Rightly or wrongly, the subjects rationalize their actions because human nature has a very conspicuous tendency to put a logical varnish on their conduct (Pareto, 1935, pp. 1120–21). Logical actions, and non-logical ones, as in types 2 and 4, are carried out because subjects have developed reasons that they consider either as having been well thought out or simply as good. While it seems obvious that logical actions are of the instrumental rationality type, Pareto points out, as concerns the non-logical actions of types 2 and 4, that they represent ‘what the subjects found to be the most appropriate based on their observation of the facts and on logical reasoning, to adapt the means to the end’ (Pareto, 1981, p. 41). These non-logical actions can thereby be understood as resulting from an imperfect reasoning mechanism, satisfying the criterion according to which ‘the subject has many good reasons

What Vilfredo Pareto brought to the economics of knowledge 29 for believing in such and such a theory X . . . even though these reasons are logically wrong’ (Boudon, 1990, p. 71). In other words, the reasons behind the acts that a subject has can be objectively wrong as they are based on bad reasoning, or unjustifiable beliefs; but they do result from a reasoning process that he considers adequate. To sum up, Pareto’s action theory rests on either objectively good reasons behind logical actions or on reasons that the subjects themselves consider as sufficient, concerning type 2 and 4 non-logical actions. At this point, it would be useful to discuss what such a behavioral analysis implies in terms of the various forms of rationality economics accepts. Type 2 non-logical actions relate to acts having a subjective goal without being based on any scientific theory. The subject is incapable of giving a satisfactory rational explanation from a scientific point of view. The objective and subjective goals differ in that the means used cannot reach the desired end. In sociology, these actions concern ritual activities such as those of the ‘rainmaker’. According to Pareto, they involve situations wherein the subjects make a mistake but can and do justify their behavior by reasoning, offering explanations that they subjectively judge to be sufficient. This kind of action can admittedly be seen as responding to the principles of cognitive rationality in the sense that it is the historical and social context at the level of scientific knowledge that brings the subjects to justify the means used to reach their ends. Thus, according to Pareto, Greek sailors offered sacrifices to Poseidon to be saved from storms and because they felt they had enough good reasons to do so. The same can be said for those scientists who believed that the earth was flat or refused to believe that Galileo’s arguments were good enough to disprove this. These kinds of action likewise correspond to any number of situations as studied by cognitive psychology within the framework of ‘problemsolving analysis’ to which Simon dedicated so much of his work. They correspond to the principle of cognitive bias whereby in any number of situations, subjects believe and explain their mistakes by reasoning in the very same way. The reasoning and the knowledge they use are wrong, which leads to non-logical actions. To give an example, many cognitive psychologists (following Simon, 1986) have shown that most subjects give the wrong answers when given tasks where inferential probability reasoning comes into play. They all give the same explanations for their mistakes. In this sense, type 2 non-logical actions correspond to those principles that define limited rationality in its primary sense because the subjects have limited computational capacity making them unable to display satisfactory behavior as seen from an outside, objective knowledge point of view. Cognitive psychologists explain this kind of behavior wherein subjects systematically make the same kinds of mistake and give the same explanation

30

Handbook of knowledge and economics

for their errors by pointing out the inherent limits to human reasoning powers (George, 1997).6 Type 4 non-logical actions cover situations where the economic actor, without mastering any theory or having access to all the information, sets up reasoning processes that he feels, right from the beginning, are all he requires to reach his goal, or not to do so. These illogical reasons are likewise related to the behavior categorized by sociologists as ‘fire makers’ because the subjects start an activity – more or less efficiently – without having a good idea of all the scientific knowledge underlying it. In this sense, too, the subjects’ rationality is limited. For example, non-logical actions of the 4b type are those situations wherein the economic agent would not have started his action if he had known its objective end right from the start. Pareto illustrates this by describing the activity of a producer in an open competition situation. In the sphere of political economy, certain measures (for example wage-cutting) of businessmen (entrepreneurs) working under conditions of free competition are to some extent non-logical actions of our 4b type, that is, the objective end does not coincide with the subjective purpose. On the other hand, if they enjoy a monopoly, the same measures (wage-cutting) become logical actions . . . while the businessman aims at reducing costs of production, involuntarily he achieves the further effect of reducing selling prices .  .  . competition always restoring parity between two prices . . . So competing enterprises get to a point where they had no intention to going. Each of them has been looking strictly to profits and thinking of the consumer only in so far as he can be exploited; but owing to the successive adjustments and readjustments required by competition their combined exertion turns out to the advantage of the consumer. (Pareto, 1935, §159, p. 86)

This situation is depicted by the ‘pursuit curve’ in the Manuale (see Figure 2.1). The curve depicts a situation where a producer at point a maximizes profit on the basis of his understanding of the market – located at m – from the information he has at his disposal at that time. However, when the producer is at b, the market has turned into m9, leading the entrepreneur to reconsider and make a new set of calculations on the basis of the new information he has received. This mechanism goes on until it reaches position M where no entrepreneur makes either profits or losses. Situation M would not have been accepted had the entrepreneur been able to know right from the start of the process that it corresponded to a non-profit situation. Two interpretations of rationality can be given on the basis of this curve. First of all, one can admit that the actions are non-logical over the long term. During the competitive process, entrepreneurs have no way of knowing that profits and losses will cancel each other out. They would not

What Vilfredo Pareto brought to the economics of knowledge 31 m

m'

m"

M

d c b

a Figure 2.1

Pareto’s pursuit curve

have accepted the long-term objective if they had been aware of this, but they do accept the subjective one, which is making short-term profits. The two objectives clearly do not correspond to each other. It is market activity sustained by agents acting through ‘trial and error’ that brings about short-term profit while canceling it over the long term. This outcome can be seen as limited rationality to the extent that the agents discover their real objective as they adapt to market circumstances given that they are able to obtain enough information to be able to do so. Pareto’s analysis illustrates the split that March (1978) points to when he makes the distinction between those rationality models that base behavior on calculated reasoning and those that consider that subjects’ behaviors can be explained as a result of a trend largely independent of the choices or calculations of a subject considered as independent. In this case, the example of pure competition that Pareto refers to can be classified among the models of ‘systemic rationality’. Paretian entrepreneurs can, at each step of the competitive process, calculate for the best according to the information they have at hand. Yet the results of their calculations, or the objective goal of their acts – to use Pareto’s terminology – lies beyond their grasp. This line of reasoning can be put in perspective with Hayek’s approach: rationality is systemic because the objective or the results come from adjustments partially due to the system governing the market and the prices at each step of the way and not only from agents’ individual behavior. Rationality in this case involves both calculations and systems (Laville, 1998, 2000). Pareto has here demonstrated how impossible it is, whether for an individual or a state, to determine pure competition equilibrium. Admittedly Pareto sees competition as a real mechanism wherein a series of adjustments takes place, partially beyond the agents’ grasp, but

32

Handbook of knowledge and economics

within a historical time frame. To put it another way, competition cannot be reduced to an atemporal trial-and-error mechanism as Walras suggested, implying that only logical acts must be taken into account. As far as Pareto is concerned, a representation of free competition may allow for logical short-term actions and non-logical long-term ones. A second interpretation of the ‘pursuit curve’ can be given. If it is held that substantive rationality deals with situations in which the economic agents are omniscient and therefore capable of reaching the best solutions, it must then also be recognized that reducing production costs in a pure competition situation is a logical action in the short term. On the other hand, this cannot be the case in the long term when agents are unable to know the results of the competitive model (if they did, their actions would be logical). To put it otherwise, not only are the subjects not unaware of the objective goal resulting from the competitive marketing process that their activities spontaneously bring about, but their subjective goal of price reduction does not fit in with the objective one, that is, the fact that the theoretical result of the competitive process implies long-term zero profits. In the short term and throughout the competitive process, they are nevertheless able to make sophisticated calculations in an effort to adapt, as shown in the pursuit curve: ‘Economic questions bring up questions analogous to those studied in mathematics under the name pursuit curve’ (Pareto, 1964, §41, t.1). This quotation suggests that Pareto is concerned about how agents adapt and how they correct their mistakes step by step. Figure 2.1 corresponds exactly to that of a differential equation solved by the tangential method, which Pareto, being the good mathematician that he was, could not have known well. The implications of the situation as it is described is therefore that the producers can themselves make these calculations and that at each step of the process, they have access to the available information ex ante, thus conforming to the principles of substantial rationality. Although their actions and expectations can be correctly described by the model of substantive rationality, they are not to be held as logical actions. Taken as a whole, that is, in both the short and long terms, Pareto’s example implies that the economic subject has extensive calculating abilities but faces an environment that is so complex that he can neither know nor control it totally and perfectly. From what has been discussed so far, it seems that Pareto’s action theory involves the articulation of different kinds or models of rationality. As we shall try to show, Pareto’s approach is an attempt to provide a coherent overall explanation of how different types of models of rationality interfere in the building of knowledge. To a certain extent, Pareto’s action theory can be seen as an explanation of how subjects acquire knowledge. In an attempt to bring together economics and sociology, Pareto tries to

What Vilfredo Pareto brought to the economics of knowledge 33 show how dynamic economic and social processes of equilibrium organize and determine each other. Within this interactive process, specific categories are involved. We shall now discuss the nature of these categories and the ways in which they work together in structuring the actors’ knowledge.

2.3 ACTION THEORY AND PARETO’S CATEGORIES IN THE BUILDING OF KNOWLEDGE The forms of thought and cognition that underlie Pareto’s action theory can be grasped from the perspective of two complementary approaches. The first interpretation is ‘cognitivist’; the second ‘emotionalist’ (Bouvier, 1999c). On the one hand, Paretian action theory can be seen as cognitivist in that it postulates that subjects act according to objective reasons or on the basis of those they consider good. These reasons come from a process of conscious objectification. The cognitivist interpretation assumes that agents determine their choices on the basis of mental activities that they set into play in order to transform information into knowledge that they can then better use to achieve their purposes or to adopt the proper behavior. The ‘emotionalist’ interpretation, on the other hand, insists on the fact that subjects are guided by their passions and feelings. It places much more emphasis on the unconscious mechanisms that cannot be controlled by reason. The categories ‘residues’, ‘derivations’, ‘interests’ and the ‘elites or heterogeneity of society’, which Pareto refers to when describing the evolution of economic and social equilibrium, provide many examples of both the ‘cognitivist’ and ‘emotionalist’ interpretations. Studying how these categories underlie logical and non-logical actions can help to better understand how Pareto conceives the process of knowledge acquisition by subjects. According to him, agents, as they make more and more contacts, acquire knowledge that they will then use and manipulate, often in an opportunistic fashion, when trying to reach a goal that was not clearly defined at the outset. Therefore Paretian action theory also corresponds to a theory of knowledge building which, as we shall discuss, can be placed within the theory of procedural rationality. In order to provide a better understanding of the connections between Pareto’s knowledge theory, action theory and types of rationality, we propose to revisit the four categories the author refers to, as given above. The ‘residues’ are a manifestation of feelings, emotions and human instincts. They are the mirrors of the psychological state of individuals. Though unconscious, they provide individuals with ready-made logical paths to follow. According to Busino (1967, 1999), their composition

34

Handbook of knowledge and economics

comes from a spontaneous, balanced, and self-regulated7 organization of mental processes expressed by logical and non-logical actions. The logical part of the residues is made up of a series of spontaneous cognitive functions. For example, identification, representation, naming, classification and serializing are some of the functions that can be found. Taken this way, the residues strengthen the idea that Pareto developed a cognitivist model. Yet residues are not objective entities that can be observed in themselves. Only their effects can be observed, thus giving certain validity to the interpretation of Pareto as an emotionalist in the sense that some of the residues may be unconscious. This being the case, admittedly Pareto’s explanation of action and reflection partially rests on a kind of black box made up of residues that can only be observed when they manifest themselves. On the other hand, a cognitivist interpretation of action assumes that an act comes from reason itself even if the subjects can make mistakes or believe in something for the wrong reasons. They can always be rationalized or defended from an epistemological point of view. This is typically the case for type 2 and 4 non-logical actions. The emotionalist interpretation, however, gives plenty of room to the passionate determinants of actions found especially in type 3 and 4 non-logical actions. Likewise, residues are the very source of that share of the constant forces that lead men to act individually or collectively. Among the six classes of residues that Pareto describes, two are particularly interesting to economists. They are, on the one hand, the ‘instinct for combination’ residues that concern all those arrangements aiming at putting knowledge or beliefs together in order to innovate. On the other hand, they are the ‘persistence of aggregate’ residues that allow social actors to formulate the arguments they need to justify an ideology or a certain number of historical institutions. The persistence of aggregate residues thus corresponds to all the reasons put forward to resist those changes and innovations that will give an advantage to certain social classes. ‘Derivations’ come from conscious and intentional activities, that is, logical and non-logical reasoning. They deal with all those actions started by the subject or a group of subjects to convince others without the latter being always considered as scientifically accurate.8 In some cases, the reasoning is valid (logical actions); in others, they are partially wrong (nonlogical actions.) Concerning derivations, Pareto indicates that even though ‘men are led by their feelings, passions and interests; they like to think that they only follow their reason’ (Pareto, 1967b, p. 9). Derivations allow us to develop explanations and give the explanations that everyone will agree to whether or not the subjects believe them or they do not appear to be logical. Derivations correspond to ‘pre-constructed’ elements that both underlie knowledge and

What Vilfredo Pareto brought to the economics of knowledge 35 are knowledge known as common sense aiming at producing social rules or norms, and are the conditions required to structure symbolic meaning. In this sense, they can be considered as routines. They aim at establishing social rules, norms or the conditions needed to structure symbolic meaning. Their function is to put together deductive and inductive acts that may be logical or pseudo-logical. Obviously, they are particularly relevant for a cognitivist interpretation of Paretian action theory. They concern valid or invalid inferences as well as whatever the subjects consider to be the best in justifying or self-justifying their acts and theories, or in convincing others. In the latter case, derivations might lend support to a logical argument that might turn out to be purely strategic or the result of deliberation. The latter are then carried out through interactions of different agents tending towards an identical objective that they gradually understand through the interactivity itself. De facto, derivations belong to the area of limited rationality and situated cognition. It can be admitted that derivations produce knowledge in an environment where the agents are interacting and that they then understand during the discussions leading them to make decisions. Thus, for Pareto, action is not so much the result of rigid deduction as it is a process of coordination between groups of agents with often conflicting interests. Subjects can improve the ways they reach their goals by interacting (situated cognition) and by using their environment (situated action). Derivations are thus concerned with interactive or strategic rationalities in so far as Pareto uses them for describing how certain groups of subjects fight with others to impose upon them this or that reason or goal. Indeed, according to Pareto, a great number of economic and social situations involve groups of subjects with antagonistic interests, which use derivations to convince others that their opinion is the right one. From this perspective, derivations could be understood as beliefs, which takes place alongside valid syllogistic reasoning, and are utilized in designing strategic behavior. The kind of rationality they thereby support is likewise argumentative as well as being limited. In fact, most of the examples Pareto gives do not distinguish the means used to attain a goal, that is, the arguments and heuristic models that support his arguments about derivations as is the case for those models based on the hypothesis of substantive rationality. ‘Interests’ are defined as all those economic and social desires aimed at obtaining possession of economic goods (Pareto, 1968, pp. 3 and 1406). Interests correspond to all those elements that push individuals to act in order to obtain those goods that they find useful. Both interests and residues are often described as belonging to the broader instinct category. Pareto does his best not to identify interests with residues, unlike many of his commentators. Interests are indeed objectified elements because they deal with material goods that can be subjectively assessed in terms of

36

Handbook of knowledge and economics

economic satisfaction, while residues are based on mental activities whose object is not perfectly defined by the individual: Interest. Individuals and communities are spurred by instinct and reason to acquire possession of material goods that are useful – or merely pleasurable – for purpose of living, as well as to seek consideration and honors. Such impulses, which may be called ‘interest’, play on the whole a very important part in determining the social equilibrium. (Pareto, 1935, p. 1406)

The ‘elites’ or ‘heterogeneity of society and circulation’ correspond to categories of individuals defined according to the qualities they possess. More precisely, Pareto uses a rating for each individual so as to measure his qualities at a given time or for a specific activity. Pareto comes to see it as a fighting process to obtain power or to defend an ideology or an institution (Pareto, 1968, pp. 2025–57). Social change involves the replacement of the elite, which is very often associated with a renewal of the economic institutions or the laws that organize power. In this case, social change emerges from a struggle between elites generally, belonging to two different leadership groups. They are characterized by differing interests, derivations and residues. Social equilibrium can then be defined as a set of rules brought about by the ‘circulation of the elite’ who have the ability to build logical and non-logical arguments by using derivations to impose their interests. Hence, for Pareto, the distribution of wealth and the efficiency of the various forms of economic actions are the reflection, on average, of individual qualities. The more capable individuals benefit from the best resources because they know how to impose their version of political, institutional and economic actions on others. The choice between free trade and protectionism, which Pareto discusses at length in the Manual and the Treatise, provides a good example of this kind of social dynamics. As we shall show in the following section, this example also permits us to investigate the dynamics of social change from the perspective of the cognitive mechanisms and the forms of knowledge they involve.

2.4 PARETO’S KNOWLEDGE ECONOMICS AND SOCIOLOGY: THE FREE TRADE/ PROTECTIONISM CHOICE In the Manual as well as in the Treatise, the choice between free trade and protectionism is based on the articulation between residues and derivations on the one hand, and between logical and non-logical actions or reasoning on the other.

What Vilfredo Pareto brought to the economics of knowledge 37 Pareto sees protectionism as the result of antagonistic relationships among the socially best organized and most convincing subjects able to make their opinions win the day even if they are false from the point of view of mathematical economics. Pareto’s analysis points out how certain political coalitions or interest groups aim at influencing institutions or laws to lean either towards protectionism or free trade in line with their own interests: The major reasons why coalitions succeed are the direct or indirect help provided by the law; in other words, it is because the public powers exercise pressure on their behalf that they succeed. (Pareto, 1964, p. 253)

Pareto emphasizes the practices, rational procedures and strategies that bring interest groups to defend their opinions even if the latter go against the common welfare, as envisaged from the perspective of pure economic theory, that is, on the basis of the mathematical demonstration of the superiority of free trade over protectionism. What Pareto shows is that this theoretical superiority is not necessarily confirmed by facts and effective individual or group behavior, but is context dependent: free trade, just like protectionism, comes from a process of knowledge that instigates both logical and non-logical actions manifested by the categories, that is, residues, derivations, interests and the elites. Thus Pareto is in favor of free trade because the mathematical demonstrations prove it to be superior in terms of social welfare. However, according to the logico-experimental method, he tries to explain why protectionism so often imposes itself as the result of interactions of logical and non-logical actions or as the outcome of interest group struggles. Pareto distinguishes two types of situation in the Manual and in the Treatise. First, there are those wherein certain protectionist or free trade actions benefit every member of a society; second, those wherein these actions put some members at a distinct disadvantage. There may be an infinite number of possibilities, but protectionism always favors one class over another. Pareto hereby stresses the dynamic nature of the social elite, and puts aside the weight of the mathematical demonstrations of the superiority of free trade, and consequently, of objective knowledge or logical action. Even if we were to demonstrate beyond any doubt that protectionism always leads to the destruction of wealth; even if this had been taught to every citizen in the same way as we teach the alphabet, protectionism would lose so few proponents and free trade would gain so few that the effect can almost be put aside, even completely so. (Pareto, 1981, p. 520)

38

Handbook of knowledge and economics

The four major reasons that explain why protectionism survives are to be found in the logical and non-logical action theory or in the strategies reasoning comes up with to maintain international trade. First, some subjects, such as persons of independent means, have a vested interest in defending protectionism in order to increase their own welfare (means of livelihood). Their actions are subjectively logical even though they do not correspond to the objective logic of science, which shows that protectionism is harmful to the community. Second, Pareto reminds us that many politicians defend protectionism to bring in more public revenues. Once again the subjective and objective goals do not correspond. Third, Pareto points out that nationalists are traditionally and ideologically in favor of protectionism. This belief comes from axiological rationality, as defined by Boudon (1997), and is therefore illogical in Pareto’s eyes. Protectionism is generally set up by a league whose most important members are . . . those so convinced of their nationalist feelings as to make them believe that protectionism serves to defend the motherland against foreigners .  .  . they imagine or pretend to believe that these measures are in tune with their ethics. They are a special breed (nationalist partisans); when of good faith, one can show them the moon in a well; when of bad faith, they show it to others. (Pareto, 1981, p. 521)

This passage refers to good faith as supporting axiological beliefs and to the fact that social groups acting out of hypocrisy will readily develop pseudo-logical reasons to justify reaching their goals. Fourth and lastly, Pareto points to those who, even when they have the intellectual faculties but are lacking in courage, will choose actions based on derivations to defend their interests. Pareto’s analysis of the choice between free trade and protectionism depends on an overall perspective and the successive approximation method in an attempt to bring together the economic theory viewpoint (on the basis of logical actions and objective knowledge) and the sociological one (on the basis of non-logical actions and subjective knowledge). What indeed requires explanation is how individuals cognitively define and justify their actions to promote either one of these two theories of international trade. Pareto’s analysis is based on a dialectical process between economic theoretical knowledge and sociological knowledge, be it individual or collective, subjective or objective. In his Treatise, Pareto illustrates the choice between free trade and protectionism by means of a chart (Figure 2.2), which shows how the different types of actions are ordered in relation to the categories underlying them.

What Vilfredo Pareto brought to the economics of knowledge 39 • The C element is either free trade theory or protectionism. • The C element is at the root of the mathematical demonstration of free trade’s superiority. • Knowledge comes from logical actions if free trade wins. • Knowledge, in this case, is objective. • If free trade is not set up, the C element is the root of protectionism. • In this case, protectionism comes from reasoning backed up by derivations. • Actions are likewise in this case non-logical. C

B

A

D

• The A element represents the actor’s psychological state. • The residues are the main category determining the A elements. • Knowledge is collectively or individually subjective. • Agents are guided by their residues, and to a lesser extent by their interests. • Non-logical actions are essential and overtake logical ones. • Agents build up collective and individual strategic reasoning to reach their goals using derivations to do so. • They use argumentative logic and limited rationality. • Agents are mistaken or think that their arguments are valid. • To a lesser extent, the agents are able to construct logical reasoning acts in order to build up their knowledge and arguments.

Figure 2.2

• The B element results mainly from collective or individual actions and subjective knowledge. • To a lesser extent, it is the result of collective or individual actions and objective knowledge born from the A, C and D elements. • C represents all concrete acts within free trade or protectionism. • Actions are logical but mostly illogical.

• The D element represents the practical adoption of free trade or protectionism as an institution. • The D element is a combination of all rules, regulations and ideologies defining the institution. • The D element is the combination of all defining the institution. • Derivations and interests underlie the D element. • Actions are mainly non-logical and reasoning is based on procedure. • Actions are only logical when free trade exists.

Actions ordered by underlying categories

40

Handbook of knowledge and economics

Four elements designated by the letters A, B, C and D are set up as a square. They each contain varying degrees of logical and non-logical actions. They are also made up of varying degrees of residues, interests, derivations and heterogeneity of society or elites. The elements A, B, C and D are connected to each other according to three major relationships and three minor ones. The latter are shown as dotted lines; the former as solid lines. Single-headed arrows show the direction of the minor relationships, and double-headed arrows those Pareto considers as highly intense (Pareto, 1968, §167). We have added to the author’s original chart the categories that are most involved in the elements A, B, C and D to make it easier to see how logical and non-logical actions interact in a way which is more favorable to either free trade or protectionism. The squares indicate what kind of knowledge is mobilized so that either type of international trade takes place. In order to understand the arguments behind the various relationships, we need to make a few comments about what lies behind the letters A, B, C and D. Element A represents the psychological states of the subjects resulting from economic, political and social interests that organize their social lives. It is essentially the residues, where non-logical actions take shape. It is also where, to a lesser extent, certain logical actions may be used to define or justify (either rightly or wrongly, but always for good reasons) free trade or protectionism theories. By element D Pareto designates those factors that favor the adoption of a theory, an ideology, or a religious or philosophical morality. In this case, both residues and interests underlie this item. Interests act as all the factors expressing the subjects’ desires for wealth. Hence the letter D corresponds to the legal rules and the institutions that are associated with free trade or protectionism. It is the result of clashes between interest groups or any other elite category capable of making the legal rules and the institutions go in their favor. Pareto gives many clear examples from history to show how this or that interest group was able to bend the laws which at times govern free trade and at others protectionism. Element C contains the mathematical theory behind either free trade or protectionism. These theories not only belong to the domain of logical actions, but also derive from interests and residues whenever the latter are not a logical cover-up but actually do underlie valid logical reasoning. Knowledge is objective whenever free trade predominates because it is based on a mathematical demonstration where only logical actions are at work. On the other hand, if protectionism predominates, it is the result of non-logical actions that satisfy some who, guided by their interests or

What Vilfredo Pareto brought to the economics of knowledge 41 false beliefs, have successfully imposed their point of view through either sufficiently reasonable arguments or slyness. Non-logical actions are really out of the question when pure economic theory is being dealt with, but are, on the other hand, at the very heart of the changes in theory whenever the fact of abandoning free trade in favor of protectionism requires justification. In this case, residues are rationalized through the derivations, whose function is to give a logical varnish to the arguments put forward by the interest group to enable it to impose its point of view on the lawmaker in the pursuit of its own advantage, pushing the latter to adopt either free trade or protectionism. Pareto thinks that logical actions predominate when a theory is being formulated and when it comes to finding ways to impose it on others. The category interests, made tangible through the use of indifference curves and logical demonstrations, then plays the main role. In this sense, it is seen as all the ‘reasoned or instinctive tendencies’ (Pareto, 1968, §2211) that come into play when formulating a theory. On the other hand, in every complex situation where there are divergent interests or groups acting strategically, non-logical actions come to the forefront. Here derivations play the lead role. This category underpins the ruses and reasons that a subject, or a group, can set up to convince others that the international trade theory he defends is the best one even if it turns out to be suboptimal in the Paretian sense. In most of the situations Pareto describes, social dynamics is based on the meeting of the derivations category and that of social heterogeneity. This meeting also concerns interests, but to a lesser degree. The elite in favor of protectionism oppose those in favor of free trade within this framework. This is indeed a struggle between the strongest and the weakest that in no way assumes that a logical scientific demonstration might lead a particular category of elite to emerge. Protectionism, seen as the result of a battle among the elite, is an objective that would not have been accepted had the social actors seen that their actions would lead to a worse situation than under free trade (type 4a non-logical action). Pareto’s endeavor is to explain the facts as shown from daily experience. He therefore sets out to explain why protectionism has often triumphed throughout history. The argument relies on the relative weight of nonlogical actions and the related influence of residues and derivations at work in the interest game played by the different entities fighting to attain their goal. For example, as far as the elite in favor of protectionism is concerned, Pareto emphasizes that whenever it is being questioned, its members are moved by second-class residues known as the ‘persistence of aggregates’. The residues will then be described as increasingly inefficient routines.

42

Handbook of knowledge and economics

Those members of the elite in favor of free trade will be pushed to act by the first-class residues – ‘instinct of combinations’, such as political innovation combined with opportunism or lobbying, and efficiency. The elite in favor of protectionism and the elite in favor of free trade will come up with both pseudo-logical and logical reasons to justify their respective positions. Going from protectionism theory to that of free trade happens whenever the second type of residues win the day. Finally, element B corresponds to the concrete manifestations, in action, of the psychological states described in elements A, C or D. Concerning free trade or protectionism, B synthesizes the effective economic actions of agents involved in real international trade relations. What was said about D also applies to B. In other words, actual exchanges themselves have more influence on the rules and institutions than the free trade mathematical theory. Yet D relies directly on A and indirectly on B, so that once again the residues – and to a lesser extent the interests and derivations – come together to modify the economic agents’ actions, leading to new international trade regulations over time. It is clearly the residues, non-logical actions, supported by the elite with superior qualities that influences social dynamics, bringing about the appearance of a certain equilibrium over time. Logical actions are only used by some of the elite to help them improve their situation. For example, C is the theory of free trade; D, the concrete adoption of free trade by country; A, a psychic state that is in great part the product of individual interests, economic, political and social, and of the circumstances under which people live. Direct directions between C and D are generally very tenuous. To work upon C in order to modify D leads to insignificant results. But any modification in A may react upon C and D. D and C will be seen to change simultaneously, and a superficial observer may think that D has changed because C has changed, whereas closer examination will reveal that D and C are not directly correlated, but depend both upon a common cause, A. (Pareto, 1935, p. 91)

Once again, residues and non-logical actions play the main role in adopting free trade (element B) or in setting up the rules that organize it (element D), while the perfect rational theory about free trade appears as secondary (element C). This is so true that understanding how social actions organize themselves as a result of the interdependence of the residues raises the problem of how to accede to knowledge, a hurdle Pareto finds it difficult to jump. Element A can only be perfectly known by studying psychology. Only the logical and the non-logical justifications in element C and the concrete actions of international trade placed in element B come to light: But C and D are simply consequences of a certain psychic state, A. There is nothing therefore to require perfect logical correspondence between them. We

What Vilfredo Pareto brought to the economics of knowledge 43 shall always be in the wrong, accordingly, when we imagine that we can infer B from C by establishing that correspondence logically. We are obliged, rather, to start with C and determine A, and then find a way to infer B from A. In doing that, very serious difficulties are encountered; and unfortunately they have to be overcome before we can hope to attain scientific knowledge of social phenomena. (Pareto, 1935, p. 92)

Using Simon’s perspective, we may, in the light of the example of the choice between free trade and protectionism, summarize Pareto’s action theory and his related conception of knowledge as follows. First, rational behavior cannot be conceived without taking into account the mental deliberative process that precedes choice. This behavior can be linked to the satisfacing principle as defined by Simon. Second, the subjects’ behavior can be considered appropriate to the subjective aim sought (objective in the case of logical actions) even if it is the result of an imperfect reasoning process. The latter is described as a deliberative mechanism preceding the selection of the action. It is founded on logical reasoning (perfect or imperfect), as can be seen in the derivations. It is likewise based on pseudo-logical reasoning under the influence of the residues. The reasoning and deliberating acts are carried out in situ so that the rationality Pareto gives the subjects implies that they can use their environment to improve their reasoning, actions or the strategies set up to reach their goals. More often than not, these activities are strategic due to the fact that they are the result of struggles between opposing groups with conflicting interests. Third, Pareto’s action theory not only takes into account reasoning processes but also the process that generates the subjective representation of the choices to be made. This process is both cognitive and sociologically situated, for it not only depends on having thought things through using intuitive mechanisms but also on identification to the particular group or set of values that the world’s complexity brings out. The latter gives rise to axiological reasoning processes.9

2.5 CONCLUSION What has been developed above highlights to what degree Pareto’s economics depends on and is contained in his sociology. Notwithstanding that the two disciplines are still separate, Pareto’s action theory continues to structure all of the social sciences (including more specifically economic sociology) and remains the first general attempt at understanding how economic subjects reason, rationalize their decisions, and acquire or use knowledge.

44

Handbook of knowledge and economics

Our argument led us to put Pareto’s typology of action (logical and nonlogical actions) and conception of rationality in perspective with more recent approaches as varied as those of Simon or Boudon. Setting this perspective has allowed us to explain how the kinds of knowledge mobilized by economic subjects in order to reach a goal are organized when economists seek to take complex social relationships into account. Far from reducing Pareto’s rationality theory to a substantive concept such as can be found in his construction of general equilibrium, his sociological categories help design a knowledge theory wherein cognitive rationalization activities are naturally limited, intentional, procedural, situated, strategic and possibly axiological. Pareto’s endeavor was to construct a logico-experimental theory of logical action and knowledge that aims at demonstrating how social equilibrium emerges and evolves. It is also an action theory where homo rationalis meets homo œconomicus, homo sociologicus and homo ethicus. Although these two aspects are often left behind in economic theory, there are however grounds for comparing Pareto with Hayek’s later work as developed in particular in his famous 1937 article ‘Economics and knowledge’. Hayek’s contention is indeed that it is not possible for economists to analyze social processes on the basis only of a priori hypotheses or pure individual choices. Putting aside the assumption of perfect rationality implicitly means placing Pareto’s logical action theory in the background. In the same text, Hayek sets the problem of the epistemological status of those hypotheses that guarantee the general equilibrium theory so that he can better point out how they fail to take into account interactions taking place among social actors. From this perspective, it is possible to establish a parallel between Hayek’s conception of knowledge and Pareto’s typology of action, but a precise account of those possible connections is beyond the scope of this contribution.

NOTES 1. The expression ‘non-logical action’ is to be found for the first time in a letter to Panteloni dated 17 May 1897 (Bobbio, 1961 and 1964). 2. Regarding this, Pareto’s Treaty of General Sociology was translated into English as Mind and Society: A Treatise on General Sociology. It is the first of a long list of studies on the sociology of knowledge from a great variety of different authors, from Perrin (1966), Berger (1967), Aron (1967), Freund (1974), Maniscalco (1994), Gislain and Steiner (1995), Passeron (1995), Bouvier (1999a, 1999b) or Boudon (1990, 1998a, 1998b, 1998c, 2000). In a complementary way, economists have paid particular attention to the border that Paretian action theory introduced between economics and sociology (Steiner, 1995, 1998, 1999; Legris and Ragni, 1999, 2004, 2005; Bruni and Guala, 2001) by putting it within the framework of knowledge economics.

What Vilfredo Pareto brought to the economics of knowledge 45 3. This aspect of Pareto’s work continues to determine how the various social sciences are structured (Boudon, 1999a). 4. Cf. Rizzello (1997) on this theme. 5. The logical–experimental method applies to economics and sociology the methods used by nineteenth-century physicists and engineers. This has often led to Pareto’s being considered as belonging to the classical inductivist school of thought (Marchionatti and Gambino, 1997; Marchionatti, 1999; McLure, 2001; Bruni, 2002). This opinion should be seen somewhat differently to the extent that Pareto uses examples or observations (Aron, 1967) rather than the experimental method in its strictest sense simply because it is seldom possible, in either economics or sociology, to make repeated and codified experiments. Pareto uses a great many examples taken from history as the source of his hypotheses and as ways of confirming the models developed using the inductive method. 6. It is quite difficult to quote all the authors or schools of psychology that use the cognitivist paradigm when treating problem solution. Still, the works of Wason (1969), Wason and Johnson-Laird (1969, 1971), Wason and Golding (1974) and George (1997) concerning conditional reasoning are based on modus ponens or modus tollens, which assume that agents’ errors have at least four sources. First, error comes from the general atmosphere or the context wherein the test is being administered (Woodworth and Sells, 1935; Wetherick and Gilhooly, 1990). Second, error comes from a more or less abstract formulation of the problem. Third, error depends on how familiar the subject is with the task being tested (Chen and Holyoak, 1985; Didierjean and Cauzinille-Marmèche, 1997). Fourth, it depends on the mental models the subject has built for himself in an effort to solve problems. These mirror his limited cognitive abilities when he attempts to come up with an adequate representation of the problem he faces (Jonson-Lair and Byrne, 1991; Bara et al., 1995). 7. This obviously bears on F. von Hayek’s work in psychology as found in The Sensory Order: An Inquiry into the Foundations of Theoretical Psychology (1952) and in his commentaries on the way Pareto looks at rationality in Scientism and the Social Sciences. 8. Most authors (Monnerot, 1978; Busino, 1999) would agree that non-logical actions are mostly guided by the feelings category and especially by the residues that leave greater room for the ‘pulsional’ magma than to the logical or pseudo-logical reasons an individual could give. In this case, Pareto’s explanation of the causes behind an act would belong to the ‘hot theories’ group in sociology. 9. ‘As creatures of bounded rationality, incapable of dealing with the world in all of its complexity, we form a simplified picture of the world, viewing it from our particular organizational vantage point and our organization’s interests and goals’ (Simon, 1996, p. 43).

REFERENCES Aron, R. (1967), Les étapes de la pensée sociologique, Paris: Tel – Gallimard. Bara, B.G., Bucciarelli, M. and Johnson, L.P.N. (1995), ‘Development of syllogistic reasoning’, American Journal of Research, 63, 69–93. Berger, B. (1967), ‘Vilfredo Pareto and the sociology of knowledge’, Social Research, 34 (2). Bobbio, N. (1961), ‘Vilfredo Pareto’s sociology in his letter to Maffeo Pantaleoni’, Banca Nazionale del Lavoro Quarterly Review, 14 (58), 269–316. Bobbio, N. (1964), ‘Introduction to Pareto’s sociology’, Banca Nazionale del Lavoro Quarterly Review, 17 (69), 180–96. Boudon, R. (1990), L’Art de se persuader des idées douteuses, fragiles ou fausses, Paris: Fayard.

46

Handbook of knowledge and economics

Boudon, R. (1997), ‘L’explication cognitiviste des croyances collectives’, in R. Boudon, A. Bouvier and F. Chazel (eds), Cognition et sciences sociales, Paris: PUF, pp. 19–54. Boudon, R. (1998a), Etudes sur les sociologues classiques, tome I, Paris: PUF. Boudon, R. (1998b), ‘Le phénomène idéologique: en marge ďune lecture de Pareto’, in Boudon (1998a). Boudon, R. (1998c), ‘Au-delà du “modèle du choix rational”’, in Saint-Sermin et al. (1998). Boudon, R. (1999a), ‘L’actualité de la distinction parétienne entre actions logiques et actions non logiques’, in R. Bouvier (ed.), Pareto aujourd’hui, Paris: PUF, pp. 35–70. Boudon, R. (1999b), Le sens des valeurs, Paris: Quadrige/PUF. Boudon, R. (2000), Etudes sur les sociologues classiques, tome II, Paris: PUF. Bouvier, A. (1999a), Pareto aujourd’hui, Paris: PUF. Bouvier, A. (1999b), ‘Naturalisme et actionnisme chez Pareto. Pertinence des problèmes parétiens en sociologie cognitive’, in A. Bouvier (ed.), Pareto aujourd’hui, Paris: PUF, pp. 273–92. Bouvier, A. (1999c), ‘La théorie de l’équilibre social chez Pareto: une théorie paralléliste. Versant causal et versant intentional de l’équilibre social’, Revue Européenne des sciences sociales, XXXVII (116), 245–54. Bruni, L. (2002), Vilfredo Pareto and the Birth of Modern Microeconomics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Bruni, L. and Guala, F. (2001), ‘Vilfredo Pareto and the epistemological foundations of choice theory’, History of Political Economy, 33 (1), 21–48. Busino, G. (1967), ‘Introduction à une histoire de la sociologie de Pareto’, Cahier Vilfredo Pareto, Vol. XII, Genève: Droz. Busino, G. (1999), ‘L’actualité des travaux de Pareto’, Revue Européenne des Sciences Sociales, XXXVII (116), 359–80. Chen, P.W and Holyoak, K.J. (1985), ‘Pragmatic reasoning schemas’, Cognitive Psychology, 17 (4), 391–416. Dennett, D. (1981), ‘Three kinds of intentional psychology’, in R. Healey (ed.), Reduction, Time, and Reality, Cambridge: Cambridge University Press, pp. 37–62. Dennett, D. (1987), The Intentional Stance, Cambridge, MA: MIT Press. Didierjean, A. and Cauzinille-Marmèche, E. (1997), ‘Eliciting self-explanations improves problem solving: what processes are involved?’, Current Psychology of Cognition, 16, 235–351. Freund, J. (1974), Pareto: la théorie de l’équilibre, Paris: Seghers. George, C. (1997), Polymorphisme du raisonnement humain, Paris: PUF. Gislain, J.-J. and Steiner, P. (1995), La sociologie économique 1890–1920, Paris: PUF. Hayek, F. (1952), The Sensory Order: An Inquiry into the Foundations of Theoretical Psychology, Chicago, IL: University of Chicago Press. Hayek, F. (1983), Scientisme et sciences sociales, Paris: PUF. Jonson-Lair, P.N. and Byrne, R.M.J. (1991), Deduction, Hove, UK: Lawrence Erlbaum. Laville, F. (1998), ‘Modélisation de la rationalité: de quels outils dispose-t-on?’, Revue Economique, 49 (2), 335–65. Laville, F. (2000), ‘La cognition située. Une nouvelle approche de la rationalité limitée’, Revue Economique, 1301–40. Legris, A. and Ragni, L. (1999), ‘Recouvrement du champ de l’économie dans l’œuvre de Vilfredo Pareto’, Revue européenne des Sciences sociales, XXXVII (116), 325–46. Legris, A. and Ragni, L. (2004), ‘La représentation de la rationalité des acteurs dans l’œuvre de Pareto: une tentative de mise en ordre’, Revue d’économie politique, 371–92. Legris, A. and Ragni, L. (2005), ‘Théorie de l’action, rationalité et conception de l’individu chez Pareto’, Cahiers d’Economie Politique, No. 49, 103–26. McLure, M. (2001), Pareto, Economics and Society: The mechanical analogy, London: Routledge. Maniscalco, M. (1994), La Sociologia de Vilfredo Pareto e il Senso della Modernità, Milano: Franco Angeli.

What Vilfredo Pareto brought to the economics of knowledge 47 March, J.G. (1978), ‘Boundary rationality, ambiguity, and the engineering of choice’, Bell Journal of Economics, 9, 587–608. Marchionatti, R. and Gambino, E. (1997), ‘Pareto and political economy as a science: methodological revolution and analytical advances in economic theory in the 1890s’, Journal of Political Economy, 105 (6), 1322–48. Marchionatti, R. (1999), ‘The methodological foundation of pure and applied economics in Pareto’, Revue européenne des Sciences sociales, XXXVII (116), 277–94. Monnerot, J. (1978), ‘Pareto-Freud ou l’introduction à la doxanalyse’, in Intelligence et politique, 2 tomes, Paris: Gauthier-Villars. Pareto, V. (1896–97/1964), Cours d’économie politique, Genève: Droz. Pareto, V. (1909/1981), Manuel d’économie politique, Genève: Droz. First Italian edn Manuale di Economia Politica con una Introduzione alla Scienza Sociale, Milan: Società Editrice Libraria. Pareto, V. (1916/1968), Traité de Sociologie Générale, Genève: Droz. Pareto, V. (1935), The Mind and Sociology (Traité de Sociologie Générale), Harcourt, Brace Einaudi editore. Pareto, V. (1966a), Marxisme et économie pure, Genève: Droz. Pareto, V. (1966b), Mythes et Idéologies, Textes réunis par G. Busino, in Oeuvres complètes, tome VI, Genève: Droz. Pareto, V. (1967a), ‘Lettres d’Italie’, in Oeuvres complètes, tome X, Genève: Droz. Pareto, V. (1967b), ‘Sommaire du Cours de sociologie suivi de Mon Journal’, in Oeuvres complètes, sous la direction de G. Busino, tome XI, Genève: Droz. Passeron, J.-C. (1995), ‘Weber et Pareto: La rencontre de la rationalité dans l’analyse sociologique’, in L.-A. Gérard-Varet and J.-C. Passeron (eds), Le modèle et l’enquête, Paris: L’Ecole des Hautes Etudes en Sciences Sociales. Perrin, G. (1966), La sociologie de Pareto, Paris: PUF. Rizzello, S. (1997), The Economics of the Mind, Cheltenham, UK and Lyme, USA: Edward Elgar. Saint-Sermin, B., Picavet, E., Filleule, R. and Demeurlenaere, P. (1988), Les modèles de l’action, Paris: PUF. Simon, H. (1945), Administrative Behavior. A Study of Decision-Making Processes in Administrative Organization, New York: Macmillan. Simon, H. (1955), ‘A behavioral model of rational choice’, Quarterly Journal of Economics, 60 (1), 99–118. Simon, H. (1976), ‘From substantive to procedural rationality’, in S. Lassis (ed.), Method and Appraisal in Economics, Cambridge: Cambridge University Press, pp. 129–48. Simon, H. (1978), ‘Rationality as process and as product of thought’, American Economic Review, 1–16. Simon, H. (1986), ‘Rationality in psychology and economics’, Journal of Business, 59, 209–24. Simon, H. (1990), ‘Invariants of human behavior’, Annual Review of Psychology, 41, 1–19. Simon, H. (1996), ‘The Sciences of the Artificial, Cambridge, MA: MIT press. Steiner, P. (1995), ‘Pareto et le protectionnisme: l’économie appliquée, la sociologie générale et quelques paradoxes’, Revue Economique, 1240–62. Steiner, P. (1998), ‘Sociologie et économie: la théorie parétienne de l’action économique, Colloque C. Gide d’Histoire de la pensée économique. Steiner, P. (1999), ‘L’entrepreneur parétien et la théorie de l’action’, Revue européenne des Sciences Sociales, XXXVII (116), 103–18. Wason, P.C. (1969), ‘Regression in reasoning?’, British Journal of Psychology, 60 (4), 537–46. Wason, P.C. and Johnson-Laird, P.N. (1969), ‘Natural and contrived experience in a reasoning problem’, Quarterly Journal of Experimental Psychology: General, 104 (1), 5–29. Wason, P.C. and Johnson-Laird, P.N. (1970), ‘A conflict between selecting and evaluating information in an inferential task’, British Journal of Experimental Psychology, 61 (1), 63–71.

48

Handbook of knowledge and economics

Wason, P.C. and Golding (1974), ‘The language of inconsistency’, British Journal of Experimental Psychology, 21 (1), 14–20. Wetherick, N.E and Gilhooly, K.J. (1990), ‘Syllogistic reasoning: effects of premise order’, in K.J. Gilhooly, M.T.G. Keane, R.H. Logie and G. Erdos (eds), Lines of Thinking, Vol. 1, New York: John Wiley, pp. 99–108. Woodworth, R.S. and Sells, S.B. (1935), ‘An atmosphere effect in formal syllogistic reasoning’, Journal of Experimental Psychology, 18 (451), 451–60.

3

Knowledge in Marshall Brian J. Loasby

3.1 OVERVIEW It has often been remarked that Alfred Marshall – in contrast to Keynes – wrote in a style so bland as to conceal the distinctiveness of what he was saying. Yet there is nothing bland about his presentation of capital as ‘the main stock of wealth regarded as an agent of production’ at the very beginning of Book IV of the Principles. Capital consists in a great part of knowledge and organization: and of this some part is private property and other part is not. Knowledge is our most powerful engine of production; it enables us to subdue nature and force her to satisfy our wants. Organization aids knowledge; it has many forms, e.g. that of a single business, that of various businesses in the same trade, that of various trades relatively to one another, and that of the State providing security for all and help for many. (Marshall, 1920, pp. 138–9)

Knowledge is clearly central to Marshall’s conception of an economic system, and to his view of how that system works; it is also central to Marshall’s conception of how economists should attempt to understand and analyse such systems. Since these concerns with process have been marginalized in much of subsequent economics, they need to be examined; and this examination should be based on an understanding of the reasons for Marshall’s distinctive attitude. The two reasons most widely recognized are his observation of the remarkable industrial developments taking place during his lifetime, which rested on the organization of the growth and application of knowledge, and his desire for improvement in the condition of the people, the surest means to which he believed lay in greater knowledge – including, it should be noted, knowledge applied to consumption, which will receive specific attention later. These issues of understanding and policy are sufficient to justify a focus on explaining how knowledge grows and how it can be encouraged to grow. However, there are other important reasons, long neglected, that illuminate the content and method of Marshall’s economics, and allow both to be seen in an unfamiliar perspective. The recently developed study of Marshall’s life before he turned to economics, and of his few surviving writings from that time, has revealed 49

50

Handbook of knowledge and economics

a kind of intellectual crisis (similar to that experienced by many of his contemporaries) about the sources and reliability of human knowledge. Marshall responded by developing his own model of an evolutionary, contingent and fallible process by which the human brain could develop classification systems for interpreting phenomena and planning action; and in the latest stage of a remarkable research programme Tiziano Raffaelli (2003) has argued in detail that the way of thinking about knowledge exemplified by this early model pervades not only Marshall’s substantive economics but also his method of developing and presenting it – notably his caution in using mathematics or any other form of logical argument. In contrast to Schumpeter’s later position, Marshall did not accept the theoretical separation of growth and equilibrium; this was a consequence of his understanding of how human knowledge necessarily grows, because of the nature of the human mind. Though not explicitly declaring, as Knight (1921, p. 313) and Hayek (1945, p. 523) were later to do, that economic problems are always problems of change, he nevertheless conceived of equilibrium as the outcome of a process by which interacting forces come into balance for a time, and the equilibrium models in the Principles are always supported by an account of how each particular kind of equilibrium could be achieved. All these accounts invoke the knowledge that is available to the relevant economic agents, and this knowledge is always a product of a particular kind of economic organization. An economy generates the patterns of knowledge to which it responds. Every process requires a structure that is substantially invariant with respect to that process; change is always localized, being dependent on the absence of change in the remainder of the system. If everything can change, then nothing can be relied on as a basis for individual decisions; thus even a general equilibrium, which as a completely specified set of relationships in which all future change is suppressed by defining goods and technologies in terms of location, date and contingency, cannot be attained by a process in which every variable is free to move in relation to every other variable. Moreover, a general equilibrium deduced from the original data is liable to be invalidated by any trades at non-equilibrium prices. This particular implication was not spelt out by Marshall; it was developed by Richardson (1960), who took care to indicate its Marshallian character. If the concept of equilibrium is to be used in explaining movement, even movement towards equilibrium, then the equilibrium must be partial; and it is the requirement of contextual stability, which may be represented as an equilibrium in the old sense of a balance of forces rather than the contemporary interpretation of internal coherence (see Giocoli, 2003), that gives some credence to the partial equilibrium method.

Knowledge in Marshall 51

3.2 THE PROBLEM OF KNOWLEDGE Marshall (1920, p. 240) emphasized the power of Adam Smith’s principle of the division of labour in explaining the course of economic development, and drew attention to the fundamental theme of organizational differentiation, both in the variety of species and the variety of function within each organism, which evolutionary biology had derived from Smith’s work. However, he seems to have been unaware of Smith’s earliest surviving account of the effects of the division of labour, which is not presented in an economic context but as an application of his psychological theory of the growth of knowledge (Smith, 1795/1980); otherwise he would surely have commented on the similarity between his own theory of structured knowledge and activities and Smith’s central idea of the invention of ‘connecting principles’ that could be applied to classes of phenomena in order to resolve disturbing failures of understanding. Because of their exclusive concentration on the Wealth of Nations among Smith’s works, most economists have been no more aware; and so even economists who drew attention to Marshall’s theory of growth and properly located it in the Smithian tradition failed to recognize the theory of knowledge that underlay that tradition. Moreover, as a natural consequence the closeness of the link between Smith and Darwin was not perceived; and this in turn encouraged the general dismissal of Marshall’s evolutionary ideas as decoration or, at best, hopes for future analytical development. However, within the last two decades there has been a revolution in Marshallian scholarship, founded primarily on the serious study of Marshall’s long-neglected papers in the context of the intellectual excitement of the 1850s and 1860s. John Whitaker (1975) had pioneered this study and, though restricting himself to the origin and early development of Marshall’s economic ideas, had drawn attention to the surviving evidence of Marshall’s non-economic interests and his path to economics; his subsequent edition of Marshall’s correspondence (Whitaker, 1996) provided further significant information about these non-economic interests and their effects on his work. Giacomo Becattini (1975) invited Italians to read Marshall afresh, and encouraged Tiziano Raffaelli, who had been educated as a philosopher, to apply that education to the interpretation of the neglected evidence in Marshall’s surviving papers, while Peter Groenewegen (1995) produced a comprehensive biography in which Marshall, very properly, escapes any simple classification. In relation to this chapter, the most significant result of this revolution is the recognition that Marshall, like Smith, confronted early in his life the question of how one could establish empirical truth and that his response to this confrontation had lasting effects on his attitude to economics. Like

52

Handbook of knowledge and economics

Smith, he concluded that proof of empirical propositions was not possible, and turned to the practical question of the processes by which individuals established usable empirical knowledge; and like Smith, he argued that these processes entailed interaction with particular phenomena that were collected into serviceable categories. These resemblances are so striking that it is hard to believe that Marshall would not have commented on them had he encountered Smith’s (1795/1980) History of Astronomy. However, there are some differences that should be noted. Smith’s analysis is in the contemporary tradition of seeking psychological answers to philosophical questions, without reference to any physical counterpart to mental operations; but Marshall draws explicitly on Alexander Bain’s (1864, 1865) elaborate theory of consciousness, thought and action, in which psychology and physiology were closely linked – though, having been trained in physics-oriented mathematics and impressed by Babbage (1864), he was particularly interested in exploring how far one could go in representing mental operations by pure mechanism. This led him to devise a conceptual model that was more elaborate than any that he subsequently developed for any economic phenomena; but though ‘Ye machine’ (Marshall, 1994c) has rightly been recognized as a key document for interpreting the development of Marshall’s ideas, it should be clearly located, as Raffaelli (2003) has done, in the context of the transformation of psychology that was brought about by Bain and the impact of Darwinian ideas, with substantial contributions from others. In taking up his Fellowship at St John’s College, Marshall’s status as second Wrangler in the Mathematics Tripos gave him a good deal of freedom in his choice of both study and teaching. However, these were troubled years, of which Groenewegen (1995) provides a comprehensive account. Marshall had been thoroughly instructed by his father in the principles of evangelical Christianity, notably the importance of demonstrated truths on which logical reasoning could be based; and mathematics was a second sphere of knowledge in which logic operated on unassailable foundations. Although Cambridge was not then a centre of scholarly innovation, the Mathematics Tripos provided a sound training in method, as O’Brien has shown by studying the examination papers of the time (Creedy and O’Brien, 1990); its reputation as the finest training in the university rested, as Groenewegen (1995, p. 116) emphasizes, on the power of mathematics to deliver ‘necessary and inevitable truth, derived axiomatically’ (see also Butler, 1991, pp. 271–4). Moreover, definitive proof that mathematical logic gave direct access to empirical truth was manifest in the theorems of Euclid, which were deemed to constitute the only conceivable geometry and had a prominent place in the Cambridge programme. So important was this proof of the empirical power of axiomatic reasoning

Knowledge in Marshall 53 that Euclidean geometry was quite commonly evoked in support of the possibility of a reasoned demonstration of the truths of religion (ibid., p. 274; Groenewegen, 1995, p. 116). Thus Marshall’s enthusiasm for mathematics seemed to reinforce his religious belief – as was true for some of his contemporaries. However, by the time that Marshall arrived in Cambridge this link between mathematics and religion had already been questioned. In the Bampton Lectures of 1858, the lecturer of that year, Henry Mansel, who was a distinguished theologian and philosopher, had explored in some detail the limits to the possibilities of human knowledge, and had concluded that the attempt to establish religious truth by reason was fundamentally misguided; the bases for religious belief must be revelation and faith. These lectures provoked a double debate, about the validity of Mansel’s argument and the implications of accepting it. Marshall discovered this debate after completing his degree, when his own religious beliefs were wavering, and he realized its significance, not only for religious belief but also for the status of axiomatic reasoning. He noted, for example, that John Stuart Mill’s dismissal of any axiomatic basis for Christian truth seemed incompatible with the unquestioned power of Euclidean geometry to deliver necessary truths about the universe (Butler, 1991, p. 276). Then, in late 1869 or early 1870, Marshall’s closest friend, Clifford, introduced him to non-Euclidean geometry (ibid., p. 282), which demonstrated that axiomatic reasoning could not be relied on to ensure the empirical truth of its conclusions; and both of the major sources of certainty in Marshall’s life were undermined. Marshall might have derived a similar sceptical conclusion from David Hume, but there is no evidence that he did. Nevertheless, even before his acquaintance with non-Euclidean geometry, his deep uneasiness about the possibility of attaining true knowledge had prompted a reaction similar to Hume’s: since there was no route to assured empirical truth, it was better to turn to the question of how people developed particular beliefs. The key was to be found not in axiomatics but psychology. For some years, Marshall contemplated a specialization in psychology, and at the end of his life wondered whether he had been right to prefer economics; and if we read Marshall with this strongly motivated early interest in mind, it is not difficult to identify substantial psychological elements in his writing, as Raffaelli (2003) has done. Moreover, this interest coincided with the impact of Charles Darwin’s ideas, and so Marshall’s psychology is evolutionary, in a double sense: the evolution of the human brain makes possible a different kind of evolution of connections within the brain of each individual. The human mind has evolved into a complex structure that is capable of imposing patterns on events, and the particular patterns that

54

Handbook of knowledge and economics

are preserved within each mind result from trial and error in a particular environment; different environments tend to lead to different knowledge, thus providing the basis for the knowledge-enhancing effects of the division of labour that was the key to Adam Smith’s theory of endogenous growth. An enquiry into the growth of knowledge entails a fundamental shift of analytical attention from proof to process; for economists it implies a requirement for theories that are ‘in time’, not just theories in which time is a parameter. (For a superb exposition of this issue, see Hicks’s (1976/1982) ‘Time in economics’.) What has not been noticed until Raffaelli’s recent work is that Darwinian ideas not only increasingly served, for Marshall as for many of his contemporaries, as a substitute for religious beliefs, but they also suggested a resolution of the controversy between the extreme empiricist view of all mental phenomena as mechanical processes of association and the claims for consciousness as a prior condition of such associations – or, in Kant’s formulation, the claim that associations are guided by certain a priori beliefs. In retrospect, Marshall’s choice of topic for the first paper that he presented to the Grote Club, a small Cambridge discussion group, in 1867 may be thought significant. Its title was ‘The law of parcimony’ (thus spelt) – the proposition that new explanations should never be sought for phenomena that could be encompassed by the extension of known causes; Marshall argued that although this law was a guide of great value, it was for that very reason liable to be applied almost as a metaphysical principle without seeking for evidence of its validity in any particular instance. He noted that Darwin had encouraged this tendency by his own speculations; and although he agreed that these speculations might be supported by the apparent homogeneity of biological phenomena, he declared that such homogeneity did not extend to mental phenomena (Marshall, 1994a, p. 99). Consciousness, in particular, could not simply be treated as the product of experience, because – as Herbert Spencer (1855, pp. 578–80) had argued – experience always requires some premises, such as consciousness may provide. Marshall concludes his paper by suggesting that only evolutionary theory can resolve this difficulty (Marshall, 1994a, p. 101). Marshall’s second paper for the Grote Club continued the argument for consciousness as a distinct phenomenon, and indicated (most clearly in a subsequent note) that a general theory of psychology could be developed by giving proper attention to both consciousness and mechanism as products of biological evolution; such a development could be based primarily on Bain’s extensive work, while accommodating substantial elements of the subjectivist position in explaining the development and behaviour of the individual. What remained to be done was to explain the mechanism that could produce, not only different mental phenomena, but

Knowledge in Marshall 55 also the perception of such differences by the conscious agent (Marshall, 1994b, pp. 104–15). This Marshall attempted in the third of his papers for the Grote Club – very different in its austere formality from its predecessors and the most elaborate model to be found anywhere in his writings (Marshall, 1994c, pp. 116–32). Marshall’s ‘machine’ is first conceived as a combination of a ‘body’, which is capable of receiving impressions from its environment and performing actions in that environment, and a ‘brain’, which has no direct connection with the environment but is capable of linking ideas that are produced by impressions with ideas that generate actions, and also of linking the latter with ideas of the impressions that appear to be a consequence of those actions. If the latter linkage produces a pleasurable sensation, then the linkage from initial impression to action is strengthened, and if the sensation is unpleasant, it is weakened. The mechanism envisaged by Marshall, possibly inspired by Babbage’s conceptions of analytic engines and automata, is of wheels connected by bands, which may become tighter or looser in response to the sensation experienced. (Electrical connections are suggested as another possibility.) This process of forming associations of contiguity or similarity through cumulative trial and error is consistent with Bain’s account of the physiology of mental phenomena; and Marshall shows how such a process could produce complex patterns of relationships. It should be noted that action is essential to the formation or dissolution of associations; this was to become an important element in Marshall’s theory of economic development. Over time such a machine may develop a range of closely connected ideas of impressions and actions, which we might now call routines; these routines are not the result of anticipatory choice but of selection among actions that, by Marshall’s intentional specification of his model, cannot originate in consequential reasoning. To be precise – and this may sometimes be very important, although Marshall does not say so – selection depends on the machine’s perception of the effects of those actions. He shows that machines operating on similar principles may develop different connections, because of differences in initial perceptions and initial actions and the selective reinforcement of what appears to work; thus a population of machines constructed to a uniform design may generate the variety that is essential for any evolutionary process. Marshall then postulates a second level of control, which uses similar mechanisms for different purposes (an early example of exaptation as a postulated evolutionary mechanism). Ideas of impressions that have not been linked to any satisfactory idea of action can now be referred to this higher level, which may generate the idea of a novel action and associate it with the idea of an impression of its effects, thus conducting

56

Handbook of knowledge and economics

a mental experiment; a pleasurable linkage of ideas is transferred to the lower level for action, and if this action produces a pleasurable impression, the link between ideas of impression and action at this level forms a new routine. This is a crucial development: it introduces imagination and expectations within an evolutionary process that does not conform to modern neo-Darwinian principles, first because the variations generated at this level are not random but oriented to problems, and second because potential variations are subject to trial and error within the mind, which may improve the chances of success in the environment. Nevertheless the variations are still conjectural, and the internal testing fallible, leaving the decisive selection to the environment, and so the process is evolutionary in the broad Darwinian sense. This formulation allows Marshall to bypass the philosophical issue of free will versus determinism; what novel connections are created in this upper level cannot be deduced from the initial conditions. It also combines evolution, organization and economics. The second level of the brain requires the prior development of the first, but then demonstrates the economic advantages of a division of labour within the brain through the separation of levels. The first level develops effective behavioural routines that are maintained at a low cost in mental energy, and the higher energy costs of generating and checking ideas at the second level are incurred only when the existing set of routines has proved inadequate, without disturbing those elements within the set that appear to work well; any improvements in performance are stored at the first level, and thus cease to require active supervision. This combination of routine and highly selective and oriented search was to become a feature of Marshall’s explanation of the working of the contemporary economic system. It is an efficient mechanism for making adjustments within limited domains, a precursor of Marshall’s partial equilibrium analysis. Although any elaboration would exceed our present remit, it is worth drawing attention to the work of the third economist who began by seeking psychological answers to the problem of human knowledge, Friedrich Hayek (1952). Like Smith and Marshall, Hayek explained the growth of knowledge by the formation of specific mental connections, which were instantiated in the physiology of the brain through a process of trial and error within particular environments. Hayek recognized that this trial-and-error process might, in principle, be located within the development of the human species or of each individual, but carefully evaded any discussion of the issue. Although Marshall does not address it directly – a clear formulation had to await an understanding of genetic transmission, subject to recombination and mutation – he clearly distinguishes between the capabilities that are built into each ‘machine’ and the particular set of

Knowledge in Marshall 57 connections that each machine develops; and he sketches out the range of possibilities that is inherent in a division of labour between machines of identical design. Moreover, he notes that machines might make other machines, with accidental variations that would be subject to natural selection (Marshall, 1994c, p. 119). It is also a feature of Marshall’s model that a substantial evolution, in the biological sense, of the machine’s brain is necessary before it becomes capable of trying out ideas prior to putting them into action: thus the potential for consciousness, if not consciousness itself, would seem to depend on development of the species. Marshall appears to have been particularly influenced by Spencer’s (1855) work on psychology, which is clearly and approvingly referenced. Perhaps of particular interest is his observation: With Kant ‘a priori’ means ‘of which the origin is unknown’; with H. Spencer it means ‘of which the origin probably dates from a long time back’. I often wonder what Kant would have said if he had had H. Spencer’s interpretation of the words shewn to him. (Marshall, 1994d, p. 135)

In Marshall’s model of how the mind works, the Spencerian ‘a priori’ has an evolutionary explanation and itself helps to explain the evolution of human knowledge within what Thomas Kuhn (1962, 1970) was to call a succession of paradigms.

3.3 THE GROWTH OF KNOWLEDGE Recent explorations of the relationship between Marshall’s early studies and his economic writings, culminating in Raffaelli’s (2003) interpretation, have revealed how extensively Marshall’s conception of the human mind is reflected in his account of the working of the economic system that he knew. Marshall’s primary reason for preferring economics to psychology was the ‘increasing urgency of economic studies, as a means towards human well-being . . . not so much in relation to growth as to the quality of life’ (Whitaker, 1996, vol. 2, p. 285); and the quality of life was crucially dependent on mental as well as physical factors. Better knowledge was a primary source not only of increased productivity (as mainstream economists have rediscovered), but also of better patterns of consumption (which is still neglected); for Marshall, preferences, like capabilities, were products of the economic system. However, to understand behaviour it was necessary to go beyond knowledge to what Marshall called character. In illustrating how his ‘machine’ might choose between alternatives by postulating a chess automaton, he observes that ‘[w]hen a man is playing

58

Handbook of knowledge and economics

at chess, just as when he is doing anything else, his character is displayed in the way in which he grasps at immediate advantages or, on the other hand, tries to look further’ (Marshall, 1994c, p. 122). Raffaelli argues that it was because Marshall was both so concerned with improving the quality of life for most people and impressed with the effects of the economic system on this quality, and especially on individual character, that he was not prepared to follow John Stuart Mill’s prescription that economists should deduce the necessary general laws of economic activity and leave to psychologists the study of individual variations and their consequences. Economics was on the one side a study of wealth; and on the other, and more important side, part of the study of man. For man’s character has been moulded by his everyday work and the material resources which he thereby procures, more than by any other influence unless it be that of his religious ideals . . . his character is being formed by the way in which he uses his faculties in his work, by the thoughts and the feelings which it suggests, and by the relations to his associates at work. (Marshall, 1920, pp. 1–2)

Marshall therefore gave particular attention to the effects of economic organization on the use of talents and the development of character, which he believed were aspects of the same process; and since progress depended on variation the differences between individuals should not be suppressed either in economic analysis or in economic systems. The implications for demand analysis are clear. Although at any instant the pattern of demand may be taken as data (even if the data are not always easy to establish), ‘each new step upwards is to be regarded as the development of new activities giving rise to new wants, rather than of new wants giving rise to new activities’ (Marshall, 1920, p. 61); therefore the study of activities must precede the study of wants; and so, in a volume on foundations, Book IV on the relationships between organization and productive knowledge is much more substantial than Book III on wants. Marshall’s analysis of the development of wants is not the least important casualty of his failure to produce the second volume that was an essential part of his original plan. We may also note that in explaining the process of consumer choice by reference to the marginal utility of money, Marshall was adhering to his context-oriented view of human cognition. Marshall’s method has significant practical and policy implications. Hicks points out that the replacement of marginal utility by an ordinal preference system was not so clear an advance as is usually supposed .  .  . The marginal utility of money, on which Marshall relies, is much more than the mere Lagrange

Knowledge in Marshall 59 multiplier, the role to which it has been degraded. It is the means by which the consumer is enabled to make his separate decisions, and to make them fairly rationally, without being obliged to take the trouble to weigh all conceivable alternatives. (Hicks, 1976/1982, pp. 285–6)

Not only does Marshall’s formulation provide a basis for believing that people will usually be able to make reasoned decisions of the kind that will support equilibrium; it also suggests, as Hicks observes, that reasonable consumer behaviour will be disrupted ‘when all prices, or nearly all prices, have broken loose from their moorings’. The routine trivialization of the costs of inflation as ‘shoe-leather costs’ by many economists would have been much harder to sustain if Marshall’s consumer theory, along with other applications of Marshall’s method, had not been discarded. It seems highly plausible that Marshall’s model of the ‘machine’, the evolved skills of which were shaped by its particular environment, made it easy for him to appreciate the link between Darwinian variation and Smith’s (1776/1976) division of labour as evolutionary mechanisms; he had a similar conception of progress, in both productivity and character, that rested on the human ability to make sense by making patterns, rather than on logical skills. Therefore the simplifying assumption of ‘economic man’ (not yet refined into a model of the rational optimizer) was not adequate, although Marshall recognized the necessity for a general presumption that people would prefer a larger gain to a smaller – as these alternatives were perceived. Rather than competition, Marshall (1920, p. 5) emphasized ‘self-reliance, independence, deliberate choice and forethought’ as ‘the fundamental characteristics of modern industrial life’ – aspects of human character that encouraged the search for new knowledge and practical skills and thereby contributed to human progress, which was his fundamental motivation. By the time that he came to economics, Marshall already had a model of how the mind works: it is an organized system of connections, developed through interactions between internal structure and external phenomena and events. Humans are neither atomistic agents nor are they constant; they co-evolve with their fellow humans, the organizational structures within which they operate, and the wider economy and society. Marshall’s model of the mind reflected both the operational constraints and the productive potential of human mental processes and of any organizational structure that was intended to make use of them; and it was based on the universal scarcity of cognition, which in most standard economic theory is the only resource that is never assumed to be scarce. Long before the ‘planning debate’ Marshall had already identified the limitations of human cognition and the consequent necessary dispersion

60

Handbook of knowledge and economics

of knowledge as the crucial weaknesses of socialist schemes. Science itself, the activity most crucially dependent on cognitive skills, is undertaken by a dispersed community that relies on a wide-ranging ceteris paribus clause in order to focus on closely defined problems, which it attempts to reduce to repetitive patterns (Ziman, 2000); and the enterprising businessman must likewise be selective in his focus and rely on many established regularities in order to devise and implement new patterns. Marshall’s recognition of this is exemplified by his ‘principle of substitution’, which is a guide to selective experimentation against a baseline of established practices (Loasby, 1990), as in scientific procedures. To an economist trained in current theory he appears to confuse changes of input combinations within well-defined production sets in response to changes in relative prices with modifications of these sets; but if performance skills and knowledge about production possibilities are both formed in the process of production, including the experiments that it suggests (as in Marshall’s ‘machine’), then the productive potential of any firm is never well defined. This insight was to become the basis of Penrose’s (1959) distinction between resources and productive services and of her account of the development of resources and the perception of productive opportunities in the course of a firm’s activity. Marshall and Penrose both implicitly reject the notion that the production function contains all the knowledge necessary to operate production processes in favour of an emphasis on ‘knowledge how’ (Ryle, 1949), which emerges from practice and becomes embodied in routines that do not require to be controlled by rational choice; both also describe the co-evolution of the firm and its environment, which corresponds to the co-evolution of the human mind and its environment in Marshall’s theory of the mind. In summarizing the qualities required of ‘the manufacturer who makes goods not to meet special orders but for the general market’, Marshall links ‘ a thorough knowledge of things in his own trade’ with ‘the power of forecasting the broad movements of production and consumption, of seeing where there is an opportunity for supplying a new commodity that will meet a real want or improving the plan of producing an old commodity’, and connects both with the encouragement of ‘whatever enterprise and power of origination’ his employees may possess (Marshall, 1920, pp. 297–8). Innovation requires both imagination and existing procedures, each of which is represented by a level in the ‘brain’ of Marshall’s ‘machine’. This sequence of creativity against a background of routines, leading to new routines that provide a more advanced basis for further creativity, is a dialectical process. (Marshall was impressed by Hegel.) When many people or many organizations pursue this sequence, their somewhat

Knowledge in Marshall 61 differently organized mental structures, which are the product of different histories, generate a variety of products and processes to be winnowed by competition. Unlike biological variety, however, this economic variety is the product of freedom as well as of chance; and the selection among this variety depends on its compatibility with existing patterns, and to some extent on conscious choices. The dialectics of evolution are summarized by Marshall in his discussion of custom. If custom had been absolutely rigid, it would have been an almost unmixed evil . . . But . . . stagnant social conditions do not crush out of everyone the desire to humour his own fancy, or his love of novelty, or his inclination to save trouble by a rather better adjustment of implements to the work done: and . . . the solidity of custom has rendered the supreme service of perpetuating any such change as found general approval. (Marshall, 1919, p. 197)

This ‘limited but effective control over natural development by forecasting the future and preparing the way for the next step’ (Marshall, 1920, p. 248) may be reasonably compared with Darwin’s recognition of the significant success of artificial breeding; in both, purposeful though fallible activities, the results of human selection, are subject to the selection processes of the wider environment, and the favoured activities become embodied in routines. The overall results of this process, Marshall believed, was a tendency, emphasized by Herbert Spencer, towards ever greater differentiation of function, matched by closer coordination (Marshall, 1920, p. 241). This closer coordination was exhibited in the various forms of organization that aid the utilization and expansion of knowledge, which for Marshall were joint products of the human mind and the systems that it supported.

3.4 ORGANIZATION Since human action is directed by human brains, the successful organization of human activity must respect the particular powers and limitations of those brains; and Marshall’s treatment of organization matches his early model of mental activity (Raffaelli, 2003). This correspondence, together with his insistence that productive activity is itself a contributor to the knowledge that enhances its effectiveness, may underlie his suggestion that organization might be recognized, alongside land, labour and capital, as a fourth factor of production (Marshall, 1920, p. 139). Indeed, Marshall’s discussion of organization begins in Chapter 9 of the Principles with an account (corresponding to his early model) of the multi-level structure of the brain, in which conscious attention is reserved for problem solving or

62

Handbook of knowledge and economics

the introduction of novelty. The application of solutions or the repetition of new actions ‘develops new connections between different parts of the brain’ (ibid., p. 252), and these connections gradually take over the maintenance of these activities, leaving the conscious brain free for new initiatives, including those that utilize these now-automatic connections. The process is illustrated by learning to skate; acquired skills no longer require thought, which may then be devoted to devising and controlling particular sequences of skating (ibid., p. 251). Order makes room for creativity, which is stabilized in a new order that combines newly established connections between perceptions and actions into a patterned performance. This performance is governed by the logic of appropriateness rather than the logic of consequences; exploring and appraising potential outcomes of action is very expensive in time and cognitive powers, while responding to circumstances is quick and simple. Knowledge does not support rational expectations but is coded into routines, without which individuals cannot function; as Whitehead (1948, pp. 41–2) pointed out, ‘[c]ivilisation advances by extending the number of important operations which we can perform without thinking about them’. All this applies to organized groups of humans. Directed action within a group relies on pre-existing routines within which no choices, in the normal sense, are exercised; but if directed action fails to achieve its objective, the recognition of failure leads either to a modification of existing routines or to experimentation resulting in new routines. This is the principle of management by exception, which may take many forms, and of the aspiration–achievement analysis developed by the Carnegie School (Cyert and March, 1963). Thus knowledge that is already organized into routines facilitates the creation of new knowledge – especially that which builds on the old; and new knowledge that is corroborated by apparently successful application is consolidated into new routines. Whitehead’s principle is equally applicable to the management of human interaction. It is not then surprising that experimentation should be at one or other of the margins of knowledge; and these margins will differ according to the past history of the growth of knowledge within each organization, because this history influences its developed capabilities and its knowledge about its environment, which together provide the baseline for experiment and for ideas about novel applications. This process is influenced by the structure of relationships within the firm and its relationships with other firms. Penrose (1959, p. 149) took care to define a firm as ‘a pool of resources the utilisation of which is organised in an administrative framework’, and showed how this form of organization aids and directs knowledge. Since resources and administrative frameworks differ (especially when we allow for informal organization), the generation of variety is a natural consequence; and

Knowledge in Marshall 63 this may be considered an effective response to the underlying and pervasive uncertainty about the likely directions of progress. We may therefore expect organizational change to be an essential factor in economic development; and so it was for Marshall. ‘[T]he part which man plays [in production] shows a tendency to increasing return. The law of increasing return may be worded thus: – An increase of labour and capital leads generally to improved organization, which increases the efficiency of the work of labour and capital’ (Marshall, 1920, p. 318). A reorganized division of labour stimulates reorganization within the brains of those affected; increasing return is not a property of any production function, but the outcome of a cognitive process in which new productive arrangements are created. In the brain as in the economy, the increasing return is to be attributed not to the elements but to the organization of more productive connections between them. Among the difficulties that naturally arise from this conception of progress is that of finding an appropriate balance between orderly activity, which is both efficient and facilitates innovation by providing a secure baseline and releasing cognitive resources but may engender hostility to any change, and creativity, which offers prospects of improvement but may be discouraged in the interests of maintaining efficiency because it consumes cognitive resources and may impose additional cognitive demands through unexpected disruption. Marshall saw this as a particular problem with large firms, in which routines are prime supporters of organizational coherence, and especially dangerous because of the valid claims that large firms could achieve greater efficiency through more carefully planned and larger-scale routines: the means of achieving this efficiency may repress ‘elasticity and initiative’ (Marshall, 1919, p. 324), and the changes in both mental and formal organization that aid knowledge. Moreover, larger firms necessarily implies fewer firms, and therefore a reduction in variety. In standard economics fewer firms may reduce welfare because they reduce allocative efficiency; that they may reduce welfare because they reduce the range of experiments is not compatible with the assumptions that are necessary to sustain the standard analyses of rational choice equilibria. This, however, is a direct implication of Marshall’s theoretical system, in which economies of scale should not be confused with increasing returns. It is perhaps because of this double threat to initiative and variety that Marshall was so impressed with the virtues of an industrial district, which seemed to ensure the ‘automatic organization’ (Marshall, 1919, p. 600) of highly specialized activities while facilitating both the generation and the active discussion of novel ideas, including ideas for constructing new patterns of relationships between firms. Appropriate routines and

64

Handbook of knowledge and economics

appropriate patterns of thought are learnt within the family and the local community, and provide a basis for efficient local markets for both labour and the specialisms of the trade and also for experiments in production methods and business organization. Moreover, those who are dissatisfied with the practices of their present employers or their attitude to new ideas can easily move elsewhere or even, in many cases, establish new firms. Continuous local interactions also promote business morality, which Marshall thought very important as a promoter of both economic and personal development. However, there is a danger that local coherence, especially if associated with success, may diminish the incentive and the mental capacity to conceive of significant innovations. Marshall (1920, pp. 197–8) noted the advantages of ‘a shifting of places’ in providing a variety of ideas and contexts; the (generally beneficial) effects of mobility on American industry, together with some less favourable effects on moral character, were the principal themes of his report on his American tour in 1875 (Whitaker, 1975, vol. 2, pp. 355–77). In view of the more recent history of many British industrial districts, it is worth recording Marshall’s (1919, pp. 135–7) warning that a network of well-proven routines could impede a major reordering of productive systems, which would then be undertaken by newcomers. Confidence in the continued validity and sufficiency of well-established knowledge provides the assurance to act; but this confidence may prevent the timely revision of that knowledge. The industrial district organizes most of the external knowledge on which each firm within it relies. But Marshall insisted that every firm required some form of external organization: a set of linkages to customers, suppliers and (perhaps indirectly through trade associations and trade journals) to other firms in the same trade. In Quéré’s (2000, p. 59) words, ‘relationships among producers and consumers are fairly particular in the sense that they require specific and mutual knowledge between the partners involved’. Such particular knowledge requires the development and maintenance of connections in which production and exchange become closely associated; and the consequence is an interfirm organization of appropriately localized knowledge that makes ‘feasible an evolution of productive activities’. The development of an appropriate and reliable set of linkages is necessarily a lengthy business (Marshall, 1920, p. 500), and requires much conscious attention before it can be taken sufficiently for granted to provide the expectations on which both regular business and experimentation can be based. (Remember that the two are closely linked.) This is the kind of competition, Darwinian though not neo-Darwinian, that is a major instrument of economic progress. It does not meet the requirements of rational choice as that is now defined. Instead it relies on

Knowledge in Marshall 65 created knowledge, assumptions that are not explicitly recognized unless something goes wrong, and fallible beliefs that may need to be revised – perhaps by newcomers – in order to conceive productive opportunities, some of which will prove, often in an amended form, to be genuine.

3.5 EQUILIBRIUM AND EQUILIBRATION Marshall sought to explain how human action, in a context of human organization and market relations, generated economic progress. This seemed to require a definition of ‘normal’ that was explicitly differentiated from ‘competitive’ (Marshall, 1920, pp. 347–8), and a range of partial equilibria, with expectations considered in corresponding time perspectives and organizational contexts. Moreover, since Marshall adopted the physical concept of equilibrium as a balance of forces (as economists generally did until fairly recently), he believed that the use of equilibrium theorizing was questionable, and even misleading, without some account of how these forces came into balance; no doubt this seemed especially desirable if equilibrium were to be used in a volume that was ‘concerned throughout with the forces that cause movement’ (Marshall, 1920, p. xiv). When the total amount of knowledge in the economy is being continually expanded by the joint effects of the division of labour and the ‘tendency to variation in any trade’ (ibid., p. 355), the knowledge of each person is necessarily incomplete, and the ability of the human brain to process knowledge is severely restricted (this was a central theme of Simon’s work); therefore any account of movement towards equilibrium must rely – just like accounts of progress – on local interpretation of local knowledge. The standard practice in current theory of deriving equilibria directly from the basic data of the model – goods, preferences and production sets – requires knowledge to be relevantly complete. Probability distributions are admissible, and even asymmetric information, if the asymmetries are themselves known; but the data must allow closure if proof of equilibrium, or of multiple equilibria, is to be possible. In Marshall’s system, however, knowledge is never complete, for that would exclude the possibility of generating new knowledge. On the other hand every agent has a history, which has left that agent with a cluster of productive and decision-making skills, including a set of expectations that provide a baseline for conscious thought, and a cluster of connections to other agents and to institutions that may be expected to guide behaviour within groups. Not least among these institutions is the current set of prices, which depend on the use of money as a unit of account, as noted earlier.

66

Handbook of knowledge and economics

From the agents’ perspective, therefore, what economists see as the equilibration problem is not a task to be solved de novo, but always a question of adjustment – which may be difficult – from an established position, like any spontaneous change; moreover this established position has a history, although the interpretation of that history may vary across agents. Equilibration is therefore not a unique issue, but a particular instance of the general problem of intelligent decision making. What makes such problems soluble in principle, most of the time, is the familiarity of the setting and the boundedness of the problem space, which allows individuals and firms to operate in ways that Herbert Simon was later to describe as ‘procedural rationality’. Marshall avoided general equilibrium because change could be handled only in a partial equilibrium framework – by economists and by those seeking to cope with or achieve change within an economy, because both had to rely on the human mind. Marshall’s principle of continuity links movement with stability, a combination that is essential to Darwinian evolution. Darwinian selection is part of Marshall’s scheme, but it operates not only through market competition but also through the evolution and application of patterns within individual brains; thus human agency is a direct (though fallible) contributor to the process of equilibration. In Book V of the Principles the most elaborate demonstration of the importance of a familiar setting in reaching equilibrium is provided in Marshall’s (1920, pp. 332–6) treatment of temporary equilibrium; this is appropriate because the most localized adjustments provide the most secure basis both for the operation of routines and for their timely and appropriate amendment. He begins in what became the orthodox fashion, by deducing the equilibrium price from supply and demand curves, but then, noting that out-of-equilibrium transactions might generate income effects that could affect both the demand and supply schedules and therefore lead to a different price, enquires whether this is a likely outcome. Taking as his setting the corn market in a country town, he argues that in a local and regular market the traders will be equally matched and well informed; their routines will be derived from appropriate experience and normally compatible enough to lead to a price close to that which might be calculated by an omniscient analyst. Uniqueness of equilibria is not a requirement of Marshall’s economic theory. Short-run prices are also affected by the history of repeated transactions between buyer and seller, and the expectation that these will continue; this leads to the prediction, contrary to modern theory, that short-run prices, even in times of depressed trade, will ‘be generally very much above’ shortrun cost (Marshall, 1920, p. 374). This is a consequence of the shared desire to preserve the continuing relationships that contribute to each

Knowledge in Marshall 67 party’s ‘external organization’, thus (in modern terms) reducing transaction costs – which are essentially knowledge costs, and therefore particularly important when knowledge is being changed by the economic process itself. His presentation of long-run equilibrium is less well supported by an examination of the setting, perhaps because the long run allows for much greater changes in relationships and in expectations; indeed the creation of an external organization that guides short-run decisions is specified by Marshall as a long-run phenomenon. Overall, we may conclude that the analysis of Book IV logically precedes Book V not only because it deals with supply but also because it provides the institutional and cognitive setting for explaining the formation of prices – which has never been satisfactorily explained within a Walrasian system. That this setting may sometimes be inadequate to ensure full employment is noted by Marshall (1920, pp. 710–12) in a few paragraphs carried over from Economics of Industry (Marshall and Marshall, 1879). Equilibration depends on beliefs, and it will fail if confidence is lacking. The human brain cannot invariably imagine a plausible new course of action when familiar routines no longer work; and when a cluster of familiar routines, based on established relationships, seems inadequate, businessmen may simply not know what to do, and so do nothing. (This was Schumpeter’s (1934) explanation for the Depression, which, he claimed, followed radical innovation.) It is significant that Marshall described this situation as ‘commercial disorganization’; the structure is inadequate to support expectations. Ceteris paribus no longer applies. This is not a Walrasian concept; it does, however, hint at Keynes’s explanation of unemployment.

3.6 MARSHALL’S METHOD Marshall’s encounter with the problem of knowledge had lasting effects on his views of the appropriate method of doing economics. With his assertion that ‘[k]nowledge is our most powerful engine of production’ Marshall (1920, p. 138) implies that the knowledge available to any person or organization at any time is inevitably incomplete; but he never sought to develop the economics of uncertainty in its usual modern sense as an extension of rational choice, because he never regarded optimization on the basis of perfect knowledge as a credible reference model, as such modern treatments do. Both his overriding desire to understand the mechanisms of human progress and his evolutionary psychology led him in another direction; and they were mutually reinforcing. It is becoming clearer, almost day by day, that the characteristic features of Marshall’s subsequent economic

68

Handbook of knowledge and economics

theorizing are natural applications of his developed understanding of the human mind. Learning processes are domain-limited, more appropriate to partial than to general equilibrium models. Operations are organizationally separated from the formation of expectations, the planning of experiments and the creation of new knowledge through the establishment of new connections; and yet they are intimately linked in the individual brain just as management is linked to the introduction of new products and processes in his theory of economic development. The reasons are that in both individuals and organizations satisfactory routines are prior to reasoning, the failure of a hitherto satisfactory routine prompts a search for a better pattern, and better patterns become embodied in new routines. Marshall also recognized that economists faced similar problems to those of economic agents; but his response was very different from the imposition of the modern principle of internal coherence, as portrayed by Giocoli (2003). ‘Every year my consciousness of the narrow limitations of all the knowledge in the world becomes more oppressive’, he wrote in 1916 (Groenewegen, 1995, p. 662). Economic knowledge was inevitably incomplete, and could be improved only by recognizing the characteristics and limitations of human cognition. The implications of non-Euclidean geometry were profound, and their relevance for economics were impressed on Marshall (1961, p. 521) by his discovery that Cournot’s analysis (which had a great initial appeal) when applied to increasing returns led directly to empirical falsehood. The applicability of axiomatic reasoning depends crucially on the premises; and when reasoning about anything as complex as an economy it is impossible to ensure that all the necessary premises have been correctly specified or even recognized. Long chains of reasoning were therefore to be avoided, and theory, though indispensable in order to organize thought, was ‘a very small part of economics proper: and by itself sometimes even – well, not a very good occupation of time’; much the greater part was ‘a wide and thorough study of facts’, which for Marshall were to be found both in documents and by direct experience (Whitaker, 1996, II, p. 393). Instead of seeking to reduce his exposition to a theoretical core, Marshall strove to keep it as realistic as possible. This was not simply a rhetorical preference: the theoretical core could not be relied on unless it could be shown to fit reality. His use of equilibrium is the most striking illustration. Schumpeter (1934, p. 80) asserted that it was the slow emergence of routine, and not rationality, that made Walrasian equilibrium an acceptable fiction, but saw no need to study this emergence; his focus was on the importance of disruptive entrepreneurship as the generator of change. But, as Raffaelli (2003, pp. 43–7) argues, for Marshall equilibrium was the end-state of a process, and should not be used as a theoretical

Knowledge in Marshall 69 concept unless some adequate account of this process and its setting could be supplied. This adequate account, in Marshall’s view, must be related to a partial, not a general, equilibrium; and Marshall’s partial equilibrium is dynamics in disguise (Dardi, 2002). Indeed, Marshall told J.B. Clark that ‘my statical state is a mere phase in my dynamical state’ (Whitaker, 1996, II, p. 419); and in the Preface to the Fifth Edition of the Principles he wrote that ‘while a description of structure need not concern itself with causes or forces, an explanation of structure must deal with the forces that brought that structure into existence; and therefore it must be in the physical sense of the term “dynamical”’ (Marshall, 1961, p. 48). Moreover, ‘time must be allowed for causes to produce their effects [and] meanwhile the material on which they work, and perhaps the causes themselves, may have changed’ (Marshall, 1920, p. 36); indeed in an early theoretical exposition he observed that ‘in economics every event causes permanent alterations in the conditions under which future events can occur’, because ‘economic forces .  .  . depend upon human habits and affections, upon man’s knowledge and industrial skill’ (Whitaker, 1975, 2, p. 163), and every event or action has consequences both for the environment and for the internal processes of human minds. We should not therefore be surprised that he identified time as ‘the centre of the chief difficulty of almost every economic problem’ (Marshall, 1920, p. vii) – nor that the domestication of time as a dimension of analysis instead of its context has accompanied economists’ ‘escape from psychology’ (Giocoli, 2003). It may be claimed that it was because Marshall was even more impressed than Walras by the manifold interdependencies of economic phenomena that he demurred from Walras’s concept of general equilibrium as the principal basis of analysis. Walras tried to encapsulate all interdependencies in a single perspective, which was later to be extended by Arrow and Debreu (who denatured both time and uncertainty, and thereby dehumanized economic agents); but even with these extensions the general equilibrium model is incomplete, and yet is not capable of generating any satisfactory general account of how such an equilibrium is to be attained. Marshall preferred to work with a multi-layered concept of equilibrium, in which the layers related both to different time-periods of activity and different groups of actors (Raffaelli, 2003). This accorded with his general advice that complex problems must be first broken up for analysis, before these analytical components can be assembled into a coherent structure; it also reflected the need for agents to find a stable basis on which to reach their own decisions. A Marshallian equilibrium represents a balance of forces that allows people to rely on established routines and relationships; a disturbance of

70

Handbook of knowledge and economics

this balance stimulates a search for new actions, which if successful will generate new routines (as in his psychological model). But in undertaking this search people must continue to rely on the continued adequacy of many established practices and assumptions: ceteris paribus is a necessary basis not only for the analysis of change but for the orderly conduct of change. What is impossible is to respond to a variation in any of the established parameters by recalculating the optimal configuration of all activities; yet this is precisely what the general equilibrium model requires. All change depends on continuity – on the absence of change in most respects. It is almost a trivial conclusion from this argument that perfect competition is not a credible setting for these adjustment processes; continuing contacts between economic agents are needed to provide both knowledge and reassurance, as Richardson (1960) argued. The selectivity of connections and extensive reliance on the routines that they embody are essential to the working of the human mind; we should not therefore be surprised that they are essential to the working of the economy. As Marshall said, organization (which implies selectivity and routine) aids knowledge, and it has many forms.

3.7 AFTERWORD The current relevance of Marshall’s treatment of human knowledge is demonstrated by Vernon Smith’s 2002 Nobel Lecture – the more effectively because it contains no reference to Marshall. (There are appropriate references to Hayek and a partial recognition of Adam Smith.) Smith’s argument is based on the limitations and capabilities of human cognition, and especially the economizing principle that ‘human activity is diffused and dominated by unconscious, autonomic, neuropsychological systems that enable people to function effectively without always calling upon the brain’s scarcest resource – attentional and reasoning circuitry’ (Smith, 2003, p. 468). These systems organize behaviour by context, which may differ between individuals, and this organization operates on evolutionary principles. Smith acknowledges, like Hayek (1952), that evolution may occur both at the genetic level and within individuals and groups, but (also like Hayek) does not distinguish between levels, preferring to relate his theoretical exposition to practical examples and carefully constructed experiments. Nor does he introduce Marshall’s theme of economic development by context-influenced variation, although the connection is easy to make. Finally, Smith (2003, p. 466) notes that among the implications of human cognition are ‘the severe limitations it imposes on our development of economic theory’. Smith calls this a ‘missing chapter’; but as we

Knowledge in Marshall 71 have seen, a draft of this chapter has been written by Marshall, on the foundations that Smith has rediscovered.

REFERENCES Babbage, Charles (1864), ‘Passages from the life of a philosopher’, in M. Campbell-Kelly (ed.) (1989), The Works of Charles Babbage, VIII, London: Pickering and Chatto. Bain, Alexander (1864), The Senses and the Intellect, 2nd edn, London: Longman, Green, Longman, Roberts and Green. Bain, Alexander (1865), The Emotions and the Will, 2nd edn, London: Longmans, Green and Co. Becattini, Giacomo (1975), ‘Invito a una rilettura di Marshall’, Introduction to Alfred Marshall and Mary P. Marshall, Economia della Produzione, Milano: ISEDI. Butler, Robert W. (1991), ‘The historical context of the early Marshallian work’, Quaderni di Storia dell’ Economia Politica, IX (2–3), 269–88. Creedy, John and O’Brien, Denis P. (1990), ‘Marshall, monopoly and rectangular hyperbolas’, Australian Economic Papers, 29 (55), 141–53. Reprinted in Denis P. O’Brien (1994), Methodology, Money and the Firm: The Collected Essays of D.P. O’Brien, 1, Aldershot and Brookfield, VT, USA: Edward Elgar, pp. 300–312. Cyert, Richard M. and March, James G. (1963), A Behavioral Theory of the Firm, Englewood Cliffs, NJ: Prentice-Hall. Dardi, Marco (2002), ‘Alfred Marshall’s partial equilibrium: dynamics in disguise’, in Richard Arena and Michel Quéré (eds), The Economics of Alfred Marshall, Basingstoke: Palgrave, pp. 84–112. Giocoli, Nicola (2003), Modeling Rational Agents: From Interwar Economics to Early Modern Game Theory, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Groenewegen, Peter D. (1995), A Soaring Eagle: Alfred Marshall 1842–1924, Aldershot, UK and Brookfield, US: Edward Elgar. Hayek, Friedrich A. (1945), ‘The use of knowledge in society’, American Economic Review, 35, 519–30. Hayek, Friedrich A. (1952), The Sensory Order: An Enquiry into the Foundations of Theoretical Psychology, Chicago, IL: University of Chicago Press. Hicks, John R. (1976), ‘Time in economics’, in A.M. Tang et al. (eds), Evolution, Welfare and Time in Economics. Reprinted in John R. Hicks (1982), Collected Essays in Economic Theory, Volume II: Interest, Money and Wages, Oxford: Basil Blackwell, pp. 282–300. Knight, Frank H. (1921), Risk, Uncertainty and Profit, Boston, MA: Houghton Mifflin. Reprinted Chicago, IL: University of Chicago Press, 1971. Kuhn, Thomas S. (1962, 1970), The Structure of Scientific Revolution, 1st and 2nd edns, Chicago, IL: University of Chicago Press. Loasby, Brian J. (1990), ‘Firms, markets and the principle of continuity’, in John K. Whitaker (ed.), Centenary Essays on Alfred Marshall, Cambridge: University of Cambridge Press, pp. 108–26. Marshall, Alfred (1919), Industry and Trade, London: Macmillan. Marshall, Alfred (1920), Principles of Economics, 8th edn, London: Macmillan. Marshall, Alfred (1961), Principles of Economics, 9th (variorum) edn, 2, London: Macmillan. Marshall, Alfred (1994a), ‘The law of parcimony’, in Raffaelli (1994), pp. 95–103. Marshall, Alfred (1994b), ‘Ferrier’s proposition one’, in Raffaelli (1994), pp. 104–15. Marshall, Alfred (1994c), ‘Ye machine’, in Raffaelli (1994), pp. 116–32. Marshall, Alfred (1994d), ‘The duty of the logician or the system-maker to the metaphysician and to the practical man of science’, in Raffaelli (1994), pp. 133–48. Marshall, Alfred and Marshall, Mary P. (1879), The Economics of Industry, London: Macmillan.

72

Handbook of knowledge and economics

Penrose, Edith T. (1959), The Theory of the Growth of the Firm, Oxford: Basil Blackwell, 3rd edn (1995), Oxford: Oxford University Press. Quéré, Michel (2000), ‘Competition as a process: insights from the Marshallian perspective’, in Jackie Krafft (ed.), The Process of Competition, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 49–64. Raffaelli, Tiziano (1994), ‘Alfred Marshall’s early philosophical writings’, Research in the History of Economic Thought and Methodology, Archival Supplement, 4, 51–158. Raffaelli, Tiziano (2003), Marshall’s Evolutionary Economics, London and New York: Routledge. Richardson, George B. (1960), Information and Investment, Oxford: Oxford University Press. Ryle, Gilbert (1949), The Concept of Mind, London: Hutchinson. Schumpeter, Joseph A. (1934), The Theory of Economic Development, Cambridge, MA: Harvard University Press. Smith, Adam (1776), An Inquiry into the Nature and Causes of the Wealth of Nations, reprinted in Roy H. Campbell, Andrew S. Skinner and W.B. Todd (eds) (1976), Glasgow Edition of the Works and Correspondence of Adam Smith, 1, Oxford: Oxford University Press. Smith, Adam (1795), ‘The principles which lead and direct philosophical enquiries: illustrated by the history of astronomy’, in Essays on Philosophical Subjects, reprinted in William P.D. Wightman (ed.) (1980), Glasgow Edition of the Works and Correspondence of Adam Smith, 3, Oxford: Oxford University Press. Smith, Vernon L. (2003), ‘Constructivist and ecological rationality in economics’, American Economic Review, 93 (3), 465–508. Spencer, Herbert (1855), The Principles of Psychology, London: Longman, Brown, Green and Longmans. Whitaker, John K. (ed.) (1975), The Early Economic Writings of Alfred Marshall (1867– 1890), 2, London: Macmillan. Whitaker, John K. (ed.) (1996), The Correspondence of Alfred Marshall, Economist, 3 vols, Cambridge: Cambridge University Press. Whitehead, Alfred N. (1948), An Introduction to Mathematics, Oxford: Oxford University Press. First published 1911. Ziman, John M. (2000), Real Science: What it is and What it Means, Cambridge: Cambridge University Press.

4

Carl Menger and Friedrich von Wieser on the role of knowledge and beliefs in the emergence and evolution of institutions* Agnès Festré

4.1 INTRODUCTION The revival of interest in the issue of knowledge in recent years1 has rarely given way to systematic studies of the nature of knowledge within mainstream economics. Within this tradition, an entire generation of economists, following the seminal work of Arrow (1955, 1962), has confined scientific and technical knowledge to information, and argued that the knowledge generated by research activities possessed certain generic properties of public goods that implied that such activities could not be produced or distributed through the workings of competitive markets. By contrast, within the Austrian tradition,2 but also in Polanyi (1967) and in the case of evolutionary economics, we find explicit recognition of the influential aspect of knowledge in human action. The contribution of the Austrian tradition to this topic, culminating in the works of Hayek, is indeed indisputable.3 Several reasons may explain this special focus on knowledge. First and most importantly, it should be recalled that the Austrian economic tradition assumes that human action takes place in time and in a context of uncertainty.4 Such an assertion implies, first, that in contrast to neo-classical economics, data such as individual preferences or production techniques are not given, but gradually take shape through individual action, experience and learning as well as under the influence of institutions and collective norms or beliefs. Second, from a methodological perspective, the Austrian tradition is associated with the subjectivist viewpoint that it develops. This implies that, though often on a rather diverging basis according to the author considered, a fundamental heterogeneity in individuals is assumed. From these two general arguments it follows that knowledge and beliefs play a fundamental role in connecting agents’ decisions through time. Several questions then arise: how do individuals acquire knowledge? What is the nature of knowledge? How is knowledge created and diffused within the society? And finally how does coordination arise? 73

74

Handbook of knowledge and economics

The answers to those questions vary substantially among Austrian authors. To give only two polar cases, Mises, who endorses an aprioristic approach, considers that knowledge consists of a priori logical categories such as causality and that human action is always intentional, rational and conscious, although he views history and law as playing some role in the determination of individual action (Mises, 1949/1996, p. 11). By contrast, for Hayek, who favours a connectionist approach based on his 1952 book Sensory Order,5 knowledge is conceived as an abstract process undergoing change in relation to new individual experiences.6 This oversimplified comparison gives a hint of the variety of the conceptions of the nature of knowledge within the Austrian tradition. It follows that conceptions regarding the role played by knowledge in economic and social activity will also vary by author. For Hayek, since the process of classification into mental categories that underlies knowledge acquisition by individual agents is both abstract (in the sense of meta-conscious) and idiosyncratic, it is obvious that coordination of individual actions as well as communication between agents constitute crucial problems. As far as Mises’s analysis is concerned, communication issues are less central, although coordination of individual actions remains an important question since it involves time as well as uncertainty. The importance attached to the problem of coordination of individual actions has also led the Austrian tradition to wonder about the role of institutions in economic and social life. The first to address this issue is the founding father of the Austrian School, Carl Menger. In his Principles (1871), he devoted an entire chapter to the question of the emergence of money to illustrate the distinction between organic and pragmatic institutions as regards their mode of emergence. For Menger, indeed, the case of the emergence of money is a striking example of how organic institutions – that is, institutions that are not the result of the individual’s will, nor the fulfilment of a collective objective, in contrast with pragmatic institutions that are the result of an individual or group of individuals pursuing an intentional goal – emerge. This idea remained a key proposition of most Austrians economists – as the well-known statement by Hayek: ‘human action but not human design’ reminds us – even if both the methodological background and the conception of knowledge of the author considered again reflect on the way he views institutions and their role in economic and social activity. In this chapter, we shall focus on two authors: the founding father of the Austrian School, Carl Menger, and one of his direct successors, Friedrich von Wieser, who are emblematic of the way the Austrian School deals with the problems of knowledge and coordination. More particularly, we shall

Carl Menger and Friedrich von Wieser on the role of knowledge 75 contrast their respective conceptions of knowledge (and beliefs) in relation to the role of institutions. We shall stress that, starting from a rather similar methodological background and being interested in solving questions such as how institutions emerge in an environment characterized by individuals’ heterogeneity, time and spatial constraints, they however provide divergent perspectives of institutional dynamics. Before entering into the details of their respective contributions to this field of analysis, some general comments are in order. First, Menger and Wieser share the view that economic institutions are the product of individual action and not necessarily of human design. In other words, they are interested in the self-regulating or spontaneous order properties of economic institutions or collective entities while, at the same time, they cannot conceive of them as not resulting from the interactions between individual agents, in compliance with the principle of methodological individualism. The process of emergence of institutions must therefore be explained from the perspective of heterogenous individual agents. It follows that assumptions concerning the nature of individual (or inter-individual) knowledge, the way it is transmitted from one individual to another, and how it evolves through time, determine to a large extent the role and the properties of institutions. In the following, we endeavour to contrast Menger and Wieser’s respective conceptions of institutions from this angle. Second, they both refer, even if implicitly, to two kinds of knowledge (or belief), namely, individual knowledge (or belief), on the one hand, and collective knowledge (or beliefs) on the other. As we shall show, the articulation between these two kinds of knowledge appears to be essential in order to analyse such phenomena as the emergence or the maintenance of institutions as well as institutional change. However, Menger and Wieser do not perceive institutions the same way, the former being more focused on the problem of emergence of institutions, and the latter more interested in the problem of institutional change in relation to the forces of power.

4.2 MENGER: THE ROLE OF KNOWLEDGE IN THE EMERGENCE OF INSTITUTIONS For Menger, knowledge is central to economic phenomena in a very general and broad way. First, individual knowledge provides the foundations of a subjectivist conception of value. As recalled by Hayek in his introduction to the English edition of Menger’s Grundsätze (translated into English as Principles of Economics), Menger defines value as ‘the

76

Handbook of knowledge and economics

importance which concrete goods, or quantities of goods, receive for us from the fact that we are conscious of being dependent on our disposal over them for the satisfaction of our wants’ (Menger, quoted by Hayek, Introduction to Principles of Economics, 1976, p. 18). On the basis of this definition, it appears that value can hardly be conceived as a static and objective concept, reducible to the concept of marginal utility, as sometimes alleged. Unlike classical and neo-classical economists, Menger is indeed not concerned with static resource allocation, but rather with resource use as a result of a better knowledge of production processes. Moreover, perceptions by individuals of their economic needs, conceived as the knowledge of the relationships between means and objectives, is essential in the definition of economic value. In another passage of his Principles, Menger explicitly endorses this vision by writing that ‘the quantities of consumption goods at human disposal are limited only by the extent of human knowledge of the causal connections between things, and by the extent of human control over these things’ (Menger, 1871/1976, p. 74). In this quotation, knowledge is conceived as general knowledge that is likely to expand with economic development. Menger indeed considers that any satisfaction of human needs begins from acquiring knowledge. For him, the driving force of economic life lies in gaining knowledge about relevant situations on a twofold basis. On the one hand, agents must know what their economic objectives are, that is, their economic needs and how those ends can be achieved through time given the time-consuming nature of economic processes: ‘clarity about the objective of their endeavour is an essential factor in the success of every activity of men’ and, moreover, ‘it is also certain that knowledge of requirements for goods in future periods is the first prerequisite for the planning of all human activity directed to the satisfaction of need’. On the other hand, given any definite objective, people must know which are the means available to them in order to achieve their objectives: ‘the second factor that determines the success of human activity is the knowledge gained by men of the means available to them for the attainment of the desired ends’ (ibid.). Those two directions of knowledge growth permit us to define the usefulness of things, that is, individuals’ knowledge of the causal relationship between means and ends. This kind of knowledge can be illustrated by Menger’s reference, borrowed from Aristotle, to ‘imaginary goods’ as a counter-example. Menger indeed defines those imaginary goods as things ‘that are incapable of being placed in any kind of causal connection with the satisfaction of human needs [but] are nevertheless treated by men as goods’ (ibid., p. 53). For Menger, this situation arises ‘when attributes, and therefore capacities, are erroneously ascribed to things that do not

Carl Menger and Friedrich von Wieser on the role of knowledge 77 really possess them’ or ‘when non-existent human needs are mistakenly assumed to exist’ (ibid.). The first case may be exemplified by cosmetics, love potions or medicines that were administered to the sick peoples of early civilizations, while examples of the second case may be medicines for diseases that do not actually exist, or statues, buildings and so on used by pagan people for the worship of idols (ibid.). In sum, people may, owing to their ignorance or misperception of either their means or their ends, unduly consider some things as being useful, even though they are not. With the passing of time and thanks to economic progress, people are assumed to learn and have a better knowledge of their means and needs, as well as of the relationships between means and ends in terms of usefulness. As mentioned by Menger in a footnote, the distinction between imaginary goods and (useful) goods can be connected to individuals’ rationality, as Aristotle suggests when he distinguishes between true and imaginary goods according to whether the needs arise from rational deliberation (in the case of true goods) or are irrational (in the case of imaginary goods) (ibid., fn. p. 53). However, Menger does not elaborate on this line of argument even if he seems to endorse such a distinction. One may however consider that this distinction is unfortunate since it is possible to conceive rationality in a more general perspective, as for instance along the lines of Raymond Boudon (1981), and, therefore, to consider that individuals have ‘goods reasons’, even from the viewpoint of the criterion of usefulness, to connect some means with ends in cases that would be considered by Menger as falling into the category of imaginary goods. A second question arises about individuals’ rationality: to what extent can people connect means and ends given the time constraints and the uncertainty that characterizes economic activity? Clearly, for Menger, the correct foresight of the quantities available for the satisfaction of intended needs is unrealistic, so that ‘in practical life . . . men customarily do not even attempt to obtain results as fully exact as is possible in the existing state of the arts of measuring and taking inventory, but are satisfied with just the degree of exactness that is necessary for practical purposes’ (ibid., p. 90). More generally, knowledge of causalities between means and ends is a principle that guides economic activity as a whole and, therefore, must not be understood as restricted to exchange analysis. First, Menger’s theoretical framework is based on the idea of the temporal antecedence of the production activity over that of exchange. Second, as is well known, he provided a microanalysis of the production structure defined as a temporal process characterized by vertical relations between different goods. The vertical hierarchy of goods or, in Menger’s terms, the order of the respective goods involved in the productive process is defined according

78

Handbook of knowledge and economics

to the degree of closeness of the good to the final stage of consumption. In terms of causal relations, this means that the basic direct causality between means and ends of one consumer does not differ in nature from the causality between productive goods of different orders within the production structure of one producer, but only in degree: the more remote in the production structure the good, the more indirect the causality. Moreover, the value of remote goods is derived from the value of the final consumer goods they contribute to produce. As Menger notes, this perspective challenges Adam Smith’s conception of economic progress based on the principle of increasing division of labour (cf. Menger, 1871/1976, p. 73). For Menger, it is clear that the increase in the volume of consumption goods going with economic progress is not the exclusive effect of the division of labour, but should also be imputed to an increased variety of economic goods, and as a corollary, to an increased knowledge of the causal relationships between means and ends, which requires an increased knowledge of the already mentioned dual-purpose knowledge (knowledge of the ends/knowledge of the means) but now extended due to time constraints: the knowledge of the ‘quantity of goods they will need to satisfy their needs during the time period over which their plans extend’ and the knowledge of ‘the quantities of goods at their disposal for the purpose of meeting those requirements’ (ibid., p. 80). This is the only sense in which it is possible to understand Menger’s sentence ‘the quantities of consumption goods at human disposal are limited only by the extent of human knowledge of the causal connections between things . . .’ (ibid., p. 74). But, as already noticed, this knowledge cannot, by nature, be complete: ‘error is inseparable from knowledge’ (ibid., p. 148). With the extension of the production structure and the strengthening of time constraints, perfect foresight is an even less realistic assumption. Indeed, Menger saw the roots of uncertainty in the time-consuming nature of economic processes. Assuming that all production takes time, producers have no way of knowing for certain the market conditions prevailing when the product is ready for delivery. The result is that the price of the finished product bears no resemblance to the costs of production, since the two represent market conditions at very different points on time. To a certain extent, one could say that the principle of increasing knowledge goes hand in hand with a principle of increasing uncertainty. As economic development proceeds, some individuals specialized in the acquisition of knowledge emerge, for instance merchants and industrialists, who act as middlemen between ‘members of the society with whom they maintain trading connections’ (ibid.). With the passing of time, such

Carl Menger and Friedrich von Wieser on the role of knowledge 79 middlemen constitute a class in its own right: they are ‘a special class of economizing individuals who take care of the intellectual and mechanical parts of exchange operations for society and who are reimbursed for this with a part of the gains for trade’ (ibid., p. 239). These middlemen are referred to by Menger when he discusses the passage from the ‘isolated household’ to the ‘organized economy’ involving a transitory state of organization corresponding to the system of production on order (Menger, 1871/1976; see Arena and Gloria-Palermo, 2001, pp. 137–8). From this perspective, economic development can be seen as a process of division of knowledge – a principle that will be later systematized by Hayek, though from a different methodological perspective. This naturally leads Menger to elaborate further upon the role of organization and institutions with respect to the problem of knowledge. However, he is not very forthcoming concerning the conditions in terms of knowledge (explicit or tacit knowledge,7 heterogeneity in agents’ cognitive capabilities) for those organizations to emerge, to be maintained or to be efficient. There is in Menger’s writings evidence that those middlemen who arise with the development of the market are more aware of the deficiencies of the previous organization of markets or have a better knowledge of their personal interest, which leads to an improvement in the efficiency of exchanges. But this does not imply that those intermediaries possess a distinctive kind of knowledge, such as practical knowledge that would be associated with their intermediation activity. More convincingly, what Menger puts forward is that they display a different kind of rationality, in the sense that they act as innovators or as leaders (see Arena and GloriaPalermo, 2001, p. 138). This assumption of a fragmented population made up of leaders and followers is also implicit in Menger’s description of the emergence of money, as we shall show. Moreover, the increasing number of kinds of goods raises the problem of factor complementarity and substitutability. Let us now envisage how Menger deals with these two issues. Menger emphasizes the principle of complementarity between goods of different orders, which he states as follows: ‘We can bring quantities of goods of higher order to the production of given quantities of goods of lower order, and thus finally to the meetings of our requirements, only if we are in the position of having the complementary quantities of other goods of higher order simultaneously at our disposal’ (Menger, 1871/1976, p. 85, italics in original). Although this passage stresses the intertemporal complementarity constraint, it does not prevent factor substitutability from occurring, provided this constraint is met. This suggests that the specialization/adaptability dilemma, borrowed from Richardson (1990),8 does not necessarily apply to Menger’s analysis.

80

Handbook of knowledge and economics

As Lachmann puts it, the argument rests on the idea that factor complementarity and substitutability do not represent two mutually exclusive alternatives, but ‘are phenomena belonging to different provinces of the realm of action’ (Lachmann, 1977, p. 200). While the notion of factor complementarity applies to capital goods utilized in the prospect of a joint output, the idea of substitutability ‘is a phenomenon of change, the need for which arises whenever something has gone wrong with a prior plan’ (ibid.). Nevertheless, this argument does not remove the lack of precision of Menger regarding organizational implications of both time constraints and uncertainty.9 This weakness also reflects on the problem of knowledge. As we stressed earlier, Menger’s explanation of successive stages of productive organization in terms of time and space constraints stands at a very general and abstract level of analysis. In retrospect one can consider that Menger paved the way for a vast field of research on capital theory, industrial organization and the role of knowledge in economic activity. Nevertheless, within the Austrian tradition, the road that will be followed focuses more on the prominent role of entrepreneurship than on the firm seen as an economic and social device for managing productive constraints. Kirzner’s contribution (1973), judging from his insistence on the awareness of the entrepreneur, provides a striking example of this tendency. From a different perspective, Menger’s emphasis on time constraints and limits to human knowledge also brings up the issue of learning and its place in economic life. Menger’s analysis of the emergence of money provides a good illustration of the importance of learning in economic phenomena and, in particular, in the emergence of the specific institution of money. The case of the emergence of money is also typical of how Menger viewed the problem of emergence of organic – as opposed to pragmatic – institutions in his Untersuchungen:10 How can it be that institutions which serve the common welfare and are extremely significant for its development come into being without a common will directed toward establishing them? (Menger, 1883/1963, p. 146)

In his discussion of the origins of money (Menger, 1892; 1871/1976, chs 7 and 8; 1883/1963, Book 3, ch. 2), Menger points out that individuals do not always get what they want using the barter system. It is both costly and time-consuming to find the exact match, identified by Jevons as the ‘double-coincidence problem’ between individual needs. They tend to abandon direct exchange and exchange their goods with more marketable ones. The causes of marketability – also referred to as saleability11 – in commodities is related, according to Menger, to different circumstances: to the

Carl Menger and Friedrich von Wieser on the role of knowledge 81 organization of supply and demand (number of buyers, intensity of their needs, importance of their purchasing power, available quantity of the commodity), to the inner characteristics of goods (divisibility, for instance) and to the organization of the market (degree of development of exchanges, of speculation and of free trade). Furthermore, the saleability of commodities is also conditioned by spatial limits (degree of transportability, degree of development of the means of transport, commerce and communication between markets) as well as time limits (permanence of needs, durability and cost of preservation of commodities, periodicity of the market, the rate of interest, the development of speculation, the weight of political restrictions to intertemporal transfers of the commodity) (Menger, 1871/1976, pp. 246–7). Agents thus progressively learn to select increasingly marketable goods and consequently proceed to indirect exchange, although it does not permit immediate satisfaction of their needs: The economic interest of the economic individuals, therefore, with increased knowledge of their individual interests, without agreement, without legislation compulsion, even without any consideration of public interest, leads them to turn over their wares for more marketable ones, even if they do not need the latter for their immediate consumer needs. (Menger, 1883/1963, p. 154)

This positive feedback12 process eventually singles out one commodity, a commodity that becomes money. This selection process, however, is not the result of purposefully thinking about the advantages of commonly understood or used money. Market participants successfully experience smoother ways of trading for the sake of personal goals and, thus, are prone to carry on. In this case, the use of money by market participants is a spontaneous outcome of the market process. In other words, they do not invent money, nor are they able to know beforehand the superior properties of money in exchange, since it is an unintended result of their self-oriented activities. But it is also the use of prior or ex ante knowledge that helps people find better ways of carrying out transactions. As Menger explains in his 1892 paper on money, ‘the willing acceptance of the medium of exchange presupposes already a knowledge of these interests on the part of those economic subjects who are expected to accept in exchange for their wares a commodity which in and by itself is perhaps entirely useless to them’ (Menger, 1892, p. 261). As such, Menger’s explanation is not satisfactory and involves some kind of circular reasoning: the question arises as to where this prior knowledge comes from since it is at the same time the result of a selection process. Menger’s answer to this question is as follows: ‘this knowledge never arises in every part of a nation at the same time. It is only in the first instance a limited number of economic subjects who will recognize the advantage

82

Handbook of knowledge and economics

in such a procedure, an advantage which, in itself, is independent of the general recognition of a commodity as a medium of exchange . . .’ (ibid.). To sum up, the process of selection consists in four mechanisms. First, it is based on an asymmetry of knowledge between two classes of agents: ‘leaders’, who are more able to see the advantages of indirect exchanges because they possess better knowledge concerning the saleability of specific commodities: they are referred to by Menger as ‘the most effective’ and the ‘more intelligent bargainers’ (Menger, 1892, p. 254); and ‘followers’, who imitate them and who progressively become aware that through the use of these specific goods they can proceed ‘to [their] end much more quickly, more economically and with a greater probability of success’ (Menger, 1871/1976, p. 258). Second, it involves a process of learning by imitation: ‘followers’ indeed imitate leaders in their use of money. They want what their neighbours possess because they observe that their neighbours perform better by using ‘money’ than they do themselves without it. As Menger explains, ‘it is admitted, that there is no better method of enlightening anyone about his economic interests than that he perceive the economic success of those who use the right means to secure their own’ (Menger, 1892, p. 247). The kind of imitation involved here is essentially of the informational type,13 since followers imitate leaders because they are supposed to have a better knowledge of the properties of money in exchange and perform better. Third, the selection process is depicted as a self-organizing procedure. As already emphasized, Menger only reluctantly admits the intrusion of external or legal compulsion in the process of emergence of money. In his 1892 article, he makes clear that ‘money has not been generated by law. In its origin it is a social, and not a state institution.’ He only admits that ‘by state recognition and state regulation this social institution of money has been perfected and adjusted to the manifold and varying needs of evolving commerce .  .  .’ (Menger, 1892, p. 255). Clearly, in his analytical framework, legal or state compulsion is, at the most, of secondary importance since social and economic institutions such as money or the organization of markets are the unintended result of interacting agents. Fourth, the emergence of money may be depicted as a self-enforcing learning process. Menger indeed emphasizes the existence of what economist today would call network externalities or network effects, that is, the idea that the more the commodity is used as an intermediary of exchange, the more it becomes an efficient medium of exchange. In this way, a good that was initially used as an intermediate of exchange is converted, through ‘customs and practice’, into a ‘commodity that [comes] to be accepted, not merely, but by all economizing individuals in exchange for their own commodities’ (Menger, 1871/1976, p. 261).

Carl Menger and Friedrich von Wieser on the role of knowledge 83 With respect to the problem of knowledge, Menger appreciates the role of non-articulable, unconscious knowledge when the use of money becomes ever more widespread, though never using the term ‘tacit knowledge’. Rather, he implicitly refers to some social or collective knowledge that is embodied in social organic institutions.14 In other words, under civilization, the individual benefits from more knowledge than he is aware of. This is just one way to explain the fact that useful organic institutions cannot be conceived only as a result of deliberate actions.15 In other words, actors frequently do better than they know merely because they know better than they are aware of knowing. It is then likely that with the passing of time the use of money becomes so anchored within habits and customs of market participants that using it no longer requires the knowledge of its inner qualities in exchange. At this moment, using money becomes completely collective tacit knowledge. Although this argument anticipates Hayek’s analysis of the process of abstraction of rules, the following passage from Menger gives some support to this line of interpretation: With economic progress, therefore, we can everywhere observe the phenomenon of a certain number of goods, especially those that are more easily saleable at a given time and place, becoming, under the powerful influence of custom, acceptable to everyone in trade, and thus capable of being given in exchange for any other commodity. (Menger, 1871/1976, p. 260)

In sum, Menger’s conception of knowledge is very general and farreaching. It includes the two kinds of categories of knowledge that are usually distinguished within the literature (explicit versus tacit knowledge), but also involves the articulation between individual and collective knowledge. Although Menger does not elaborate fully on the mechanisms at work in the articulation of individual and collective knowledge, the existence of shared knowledge is essential for understanding the process of emergence of institutions. Moreover, this shared knowledge cannot be built through mere conscious imitation of leaders by followers. It is indeed the superposition of a collective/implicit knowledge that may explain the self-organizing and self-enforcing dimensions of the process of diffusion of money. Finally, it explains why organic institutions are the unintended result of interacting individual agents.

4.3 WIESER: THE ROLE OF BELIEFS IN THE EVOLUTION OF INSTITUTIONS Friedrich von Wieser is the least-known author within the triumvirate of the founders of the Austrian School. On the one hand, he is often viewed

84

Handbook of knowledge and economics

as a faithful follower of Menger, paling into insignificance beside BöhmBawerk. On the other hand, his inclination in favour of authoritarian political regimes did not contribute to making him very well liked. These circumstances certainly contributed to the underestimation of his work. However, as pointed out, Wieser’s contribution to capital and imputation theory had a great influence on the development of the intellectual achievements of the second generation of the Austrian School even if it is mostly still unappreciated in the literature16 (Streissler, 1987; Endres, 1991). Moreover, from a methodological viewpoint, he offers a very interesting mixture of economic analysis and economic sociology that shares some features with Schumpeter’s theoretical construction (see Streissler, 1988, p. 195 and Arena and Gloria-Palermo, 2001). In particular, Wieser developed a very interesting analysis of the role of knowledge and beliefs in economic and social phenomena in relation to the issues of the emergence and evolution of institutions. The remainder of this chapter will focus on this aspect of Wieser’s contribution. The accomplished version of this analysis is to be found in his late book Das Gesetz der Macht (translated as The Law of Power), which he published only a few months before his death, even if the overall scheme of thought of this last piece of writing is already sketched out in one of his previous works: Theorie der Gesellschaftlichen Wirtschaft (translated as Social Economics). More precisely, in Social Economics, Wieser is concerned with the deficiencies of a theory of marginal utility independent of the distribution of wealth or economic and social inequalities, for those factors, according to him, do affect subjective individual valuations.17 In particular, in contrast to Menger, Wieser maintains that political compulsion and power play a decisive role in the formation of individual preferences.18 In this same book, Wieser contrasts the Theory of the Simple Economy with Social Economics, stating that in the latter, social stratification between classes exerts substantial effects on the economic activity and, in particular, on individual preferences valuations (see Arena, 2003, p. 303). The existence of three classes implies that the group of ‘mass commodities’ has to be evaluated by the marginal utility of the poor, the set of ‘intermediate goods’ by the preferences of the middle classes and the group of ‘luxury goods’ by the valuations of rich people (Wieser, 1914/1927, 1967, pp. 157–8). But it is in his last contribution, the Law of Power, that Wieser provides an overall analysis of society that emphasizes power – power play, the psychology of power – and the role of beliefs in the emergence and evolution of institutions. We shall therefore concentrate on this work by Wieser. First, we want to focus on some important methodological features and terminology of Wieser’s overall theoretical approach that might be

Carl Menger and Friedrich von Wieser on the role of knowledge 85 helpful for understanding his perspective on the problems of emergence and evolution of institutions. From a methodological standpoint, it is important to note that Wieser departs from Menger’s strict methodological individualism and promotes an original view that mixes methodological individualism and holism.19 This perspective is, for Wieser, necessary if one wants to deal with social phenomena such as the emergence of institutions. He is certainly not satisfied with individualistic explanations that provide no other explanation ‘than the one which suggests itself in the personal sphere for the relations between individuals . . .’. In particular, such explanations afford no room for ‘the element of constraint or command without which the [S]tate could neither originate nor endure and which can be clearly established for money as well’ (Wieser, 1926/1983, p. 146). In this passage, Wieser implicitly refers to Menger’s explanation of the emergence of money, which he considers unsatisfactory for reasons that are not only related to the absence of legal or state compulsion, but also on other grounds that will be explained in the following. Wieser is no more satisfied with the polar case of collectivist explanations, stating that ‘in a roundabout way, [they] lead back to the individualistic explanation[s] by taking the people and the masses as a magnified individual’ (ibid.). For Wieser, there is no hope in those two kinds of monist explanations. He also criticizes ‘dualist explanations’, arguing that they also confront us with problems and do not, therefore, constitute a satisfactory alternative. Wieser takes as an example of the dualist explanations the classical distinction between the subjective (use) and the objective (exchange) value of goods, supposedly able to reconcile respectively the ‘personal or individual influences’ and ‘those influences which transcend the personal or individual’. But without a suitable way of connecting those two elements, this approach cannot be accepted either. For Wieser, the manner in which classical economics connects the two dimensions is misleading because the ‘so-called objective exchange value does not by any means apply objectively to everybody’. More precisely, on the demand size of the market of a particular good, the principle of the objective exchange value holds true ‘only for those who can pay the current price, i.e., for those for whom the acquisition of the good brings an increase in utility which at least offsets the decrease in utility brought about by the payment of the price’. Symmetrically, on the supply side, the principle holds good ‘only for those to whom the attainable price brings an increase in utility sufficient to compensate for the sacrifice which giving up possession of the goods involves’ (ibid., p. 147). Wieser therefore concludes that ‘the objectively determined price gives only the proximate base and not the ultimate standard for valuation, for one and the same quantity if money means a quite different

86

Handbook of knowledge and economics

utility experience for the poor and for the rich person . . .’ (ibid., italics in the original). In other words, this amounts to saying that the so-called objective exchange value is the subjective exchange value for market participants since those who participate in exchange are oriented towards the same objective base, that is, the market price. But for those excluded from the market, the objective exchange value has no meaning since ‘its outcome is as subjectively determined as is the personal use value in each individual case’ (ibid., p. 148). In fine, the contrast between objective exchange value and subjective use value is nothing more than a ‘contrast between a multitude of parallel subjective cases and the isolated case’ (ibid., italics in the original). From a terminological viewpoint, Wieser refers to ‘social institutions’ as distinguished from ‘historical formations’ with respect to their mode of emergence. Social institutions are characterized by the fact that they ‘are created by governments or by other orders of power for a deliberate purpose and following a deliberate plan’ or ‘in the awareness of a task to be done’, while ‘historical formations’ ‘grow up without the possibility of one’s becoming aware of a specific creator’ but rather as a ‘searching force’ (Wieser, 1926/1983, pp. 141–5). At first sight, Wieser’s distinction is reminiscent of Menger’s distinction between organic and pragmatic institutions. By analogy, historical formations could be conceived as organic institutions since they are the result of unintended action while social institutions should be considered as pragmatic since they are the result of ‘intentions, opinions, and available instrumentalities of human social unions or their ruler’ (Menger, 1883/1963, p. 145). Nevertheless, this analogy no longer holds if one recalls that, for Menger, the distinction between pragmatic and organic institutions follows directly from another distinction of a methodological nature: the distinction between two orientations of theoretical research, namely the ‘realistic– empirical’ approach and the ‘exact science’ one.20 Even though Menger admits that these two orientations can supplement each other, they nevertheless constitute two logically distinct perspectives. For Wieser, by contrast, explanation of social institutions and historical formations cannot be the subject of independent analyses. On the one hand, Wieser explains that social institutions are always embedded in ‘historical formations’ (Wieser, 1926/1983, p. 146) in the sense that an emerging social institution must necessarily fit or be consistent with the contemporary existing historical formation. Wieser takes as examples the market institution and monetary institutions, indicating that ‘the market system presupposes the market as created by the coincidence of supply and demand’, or monetary institutions, and that ‘the special monetary arrangements of a country are

Carl Menger and Friedrich von Wieser on the role of knowledge 87 based on the general characteristics of money, which has come about as a result of the tortuous paths of commerce . . .’ (ibid., p. 143). Generalizing this argument, Wieser maintains: [with] all institutional arrangements it can be clearly seen how in their effect they always depend on being properly adjusted to the nature of historical formations which serve as their foundations. (ibid.)

For Wieser, indeed, a market system that is not consistent with the law of supply and demand cannot succeed. The same is true for ‘a monetary system attempting to maintain a value of money which has become untenable by an excess of monetary media issued by the state’ (ibid.). On the other hand, historical formations are defined in relation to power, which plays an important role in Wieser’s analysis of institutions. This characteristic sets him apart from Menger, who takes into account the role of power only in the sense of the command of economic resources. In other words, Menger deals with power as a necessary condition for being (or not) in a position to use economic goods but does not elaborate further on how power is distributed among individuals and how this distribution evolves through time. By contrast, when referring to historical formations, Wieser notes that they constitute particular states of the evolution of the society, characterized by a certain social stratification of powers, and which result from ‘the accord of many personal units of consciousness which to a certain degree give up their independence, but without a higher encompassing unit of consciousness taking their place’ (Wieser, 1926/1983, p. 146, italics in the original). More precisely, Wieser’s analysis of power is based on two distinctions: on the one hand, the distinction between masses and leaders; on the other hand, the distinction between internal and external power. First, Wieser differentiates between leaders and masses more to suggest that their respective behaviours or rationality obey different laws or display a distinct psychology of power than to suggest an idea of intellectual superiority of leaders with respect to masses or to give a pejorative meaning to the term ‘mass’. In Wieser’s own words, being a leader ‘means nothing but to be the first in matters of common concern [and] [t]he social function of a leader is to walk in front .  .  .’ (ibid., p. 38). However, the phenomenon of leadership is based on the existence of inequalities within a given population: it is governed by the ‘law of small numbers’ based ‘on the social success of small groups’ (ibid., p. 1). Moreover, it is possible to distinguish between two types of leadership. On the one hand, there is the authoritarian personal or cooperative kind of leadership, including despotic leadership or small social groupings such as guilds. On the other hand, there is the impersonal or anonymous kind of leadership, which

88

Handbook of knowledge and economics

predominates in the sphere of private life in a free society and complements the authoritarian kind of leadership in many instances, such as in the case of the development of language, art, science, law or ethics as well as in the creation of money (ibid., p. 40).21 Both forms of leadership are however governed by the principle of success, which implies imitation by the masses. Wieser indeed describes the masses as ‘following the leader’ (ibid., p. 38), although he distinguishes between two forms of following: dead masses or blind following and true following. Blind following refers to a passive form of following that is reducible to sheep-like behaviour, that is, ‘following [the] close surroundings rather than the leader’ (ibid., p. 44). True following, on the contrary, refers to a reflective, searching type of following or active following, ‘which demands of the masses a certain independence of conduct and the capacity to adapt to the given circumstances’ (ibid., p. 45). Moreover, in the case of true following, the psychology of the masses also displays a process of internalization (or even true identification) of power: Internal power arouses in the masses the urge for ready emulation. In this connection the individual obeys not only his own instinct, but his behavior is also determined by the contact he has with the attitude of his environment and that of the masses in their entirety. The experience of power is intensified by the fact that the individual submitting to power thereby enhances the effective weight of internal power in society: he joins the ranks of the social rulers, albeit with a minimal share of power. (ibid., p. 57)

This passage suggests a dynamics between masses and leaders that is more complex than the term ‘following’ suggests. In fact, as we shall explain, this dynamics cannot be described in terms of imitation only. In particular, it involves the interplay of ‘internal power’, which constitutes a critical factor of the dynamics between masses and leaders. Second, Wieser defines internal power as distinguished from external power. He conceives ‘internal power’ as ‘impersonal and anonymous’ (ibid., p. 3), while ‘external power’ corresponds to the power that some persons or some groups exert on people with the help of ‘external’ means such as ‘numerical superiority, arms or wealth’ (ibid.). Internal power may be channelled by several means: science and knowledge, for instance, through ‘historical education’, contribute to the creation of the social interactions that support social internal power (ibid., p. 107), while arts rather fall into the category of external power and leaders, even if it is rooted in the populace (ibid., p. 113). But the power of knowledge, contrary to ‘faith power’, is not a direct power but needs many intermediaries within the ruling classes in order to reach the masses. It is therefore

Carl Menger and Friedrich von Wieser on the role of knowledge 89 associated with Wieser’s first law of social growth: the tendency toward increasing stratification (ibid., p. 26). To a certain extent, the distinction between internal and external power again echoes the one between historical formations and social institutions. Internal power indeed refers to historical formations conceived as the result of unintended actions, while external power can be associated with social institutions viewed as human devices designed with a specific task or purpose in mind. Similarly, as in the case of the distinction between historical formations and social institutions, internal and external power can hardly be dealt with independently of each other.22 As Wieser emphasizes, there must be some complementarity and consistency between internal and external power. Related to this issue, he criticizes Nietzsche and Spencer’s too emphatic conception of the leader or the ‘great man’, which is, according to him, out of touch with the reality of the masses (ibid., p. 46): [i]ndispensable as is the performance of the leader in front for the achievements of society, no less so is the following by the masses. If the leader is viewed as the sower casting out the seed, the masses may be viewed as the ground which absorbs it. (ibid., p. 47)

This quotation brings us back to the issue of methodological individualism and holism. As we have emphasized, Wieser is satisfied neither with monist approaches nor with dualist ones, as he understands them. So what is the suitable method to be applied? Wieser’s answer is not always clear or his argument easy to follow. As has also been stressed, when he describes the dynamics between masses and leaders, Wieser clearly has in mind something more than passive imitation or ‘blind borrowing’. We could argue that the kind of following that he labels ‘true following’ has some common features with self-reference imitation, defined by Orléan (1988) as a kind of imitation occurring when agents imitate but, by so doing, create a social value or a convention that gathers some momentum and gains some autonomy vis-à-vis the individuals who initiated the dynamics. By the same token, the idea of ‘anonymous leadership’ could also be interpreted as a sophisticated mechanism that involves more than mutual interaction between masses and leaders. For, as Wieser indicates, this kind of leadership is characterized by the fact that ‘the social success of small groups can be magnified to full-fledged social success if the new strength, which first was formed by the small group in its own interest, is removed from its control and placed at the disposal of the society as a whole’ (ibid., p. 33).23 The following arguments may indeed support this view. First, Wieser refers to the notion of ‘objective spirit’, which he borrowed from Dilthey and Freyer, whose teaching was widespread in Germany. He

90

Handbook of knowledge and economics

argues that the existence of an ‘objective spirit’ plays an essential role in the articulation between individuals or masses and leaders, and strengthens the coherence of social systems. More precisely, this ‘objective spirit’ relates to the commonly observed psychology of social human beings to infer from others and share some significations or values: Because we ourselves are moved by emotions, follow impulses, act purposeoriented, connect mental images, forge concepts, and because this structural coherence of minds, characteristic of our very nature, falls within the realm of our experience, we can imagine ourselves as partaking in the consequences of the acts of foreign human beings and can re-create what spiritual values they contain . . . What is foreign becomes a signpost which we are able to follow even when it does not guide us simply in a certain direction but leads us to a plenitude of heterogeneous realities: languages, literatures, states, architectural styles, churches, customs, arts, and systems of sciences. (Freyer, Theory of the Objective Spirit, quoted by Wieser, 1926/1983, p. 147)

This idea of ‘objective spirit’ can be interpreted as a means to articulate individual beliefs or values with collective ones in a manner that is more sophisticated that the one implied by mere imitation. More precisely, for Wieser, the objective spirit of a community is more than a signpost: it is ‘like a current to which we are glad to yield because we feel its supporting power and whose superior strength we possibly may not be able to escape at all even when we are terrified to discover that it will carry us toward the abyss’ (ibid., p. 148). In other words, the ‘objective spirit’ becomes an entity that has its own developmental mechanisms, such as inertia, selfpreservation (ibid., p. 124) and destructive power. However, its autonomy vis-à-vis individuals ‘must not undermine our recognition that it is borne out of the spirit of the united individuals’ (ibid., p. 149). The idea of ‘objective spirit’ can also be related to Wieser’s notion of social egoism, which he developed in Social Economics. For Wieser, social egoism is conceived as an intrinsic component of the psychology of human beings implying that ‘by reason of the social egoism a man is ready to fit into the social order which includes both submission and domination’ (Wieser, 1914/1927, 1967, p. 161). Second, Wieser refers to the notion of success, a concept that, incidentally, he shares with Menger. But, in contrast to Menger, for Wieser, the notion of success encompasses more than the idea of the replication by followers of supposedly efficient behaviours displayed by leaders. As pointed out by Samuels (introduction to Wieser, 1926/1983), Wieser’s concept of ‘success’ is not defined in abstracto as the achievement of the fittest economic state. In particular, depending on the fact that it is actual or perceived, success can also lead to negative outcomes, such as dictatorships:

Carl Menger and Friedrich von Wieser on the role of knowledge 91 Success constitutes a mechanism, as it were, of historical selection. The course of history is marked by a path of success vis-à-vis other paths which might have been. Success in this context signifies survival . . . Success, in Wieser’s analysis, has no independent positive or normative, ex ante, test. It is circumstantial, episodic, and without external or internal value basis independent of the fact of survival. It is the consequence of successes, however, which marks the course of history. (Samuels, introduction to Wieser, 1926/1983, p. xxxi)

Third, Wieser refers to ‘the law of upward mobility of classes’,24 which implies the existence of a tendency towards the congruence of beliefs between masses and leaders. Indeed, at first, masses have no share in public power, but through social interaction – through labour and art essentially, they may have an opportunity to further their personal achievements and, therefore, to resist the pressure from the leaders and not completely succumb to it. As far as the ruling leaders are concerned, they are themselves aware of their own interest in augmenting the vigour of the people to utilize it better, so that the more enlightened rulers have a strong affinity with the populace and begin sharing public power with it. As Wieser summarizes: In the present epoch, the face of the earth is being technically transformed by the alertness of .  .  . both those in command and those in subordination positions. All these quietly evolving and ascending collective forces have in due time been transformed into social power or they will do so, acting as a resistance first but eventually also sharing leadership roles. (Wieser, 1926/1983, p. 26)

All these arguments give strong support to an interpretation of Wieser’s approach to the emergence of institutions in terms of a more sophisticated dynamics of social interaction than the one put forward by Menger. This interpretation is reinforced by thorough investigation of Wieser’s remarks regarding Menger’s explanation of the emergence of money. In Social Economics, Wieser dedicates several pages to Menger’s approach to the emergence of money. If he clearly regards money, as does Menger, as one of the founding institutions of social economy, he has reservations about the way Menger deals with the issue of its emergence. On the one hand, he appreciates Menger’s novel ‘penetrating investigation’ into the problem of money which – by taking ‘the phenomenon of money as a paradigm’ allowing him to show that ‘all social institutions of the economy are nothing more than unintended social results of individual teleological factor’ (Menger, 1883/1963, pp. 171–87) – put an end to the ‘long series of writers who sought to explain money as an individualistic institution’ (Wieser, 1914/1927, 1967, p. 163). On the other hand, a few pages later, he points out:

92

Handbook of knowledge and economics Menger’s explanation would be entirely satisfactory if he had appreciated as fully the part that the masses play in the development of money as he did that taken by leaders. The function of the masses consists in that their imitation establishes the universal practice which gives to a rule its binding force and social power . . . It is in keeping with . . . [his] individualistic point of view that he . . . [did] not fully appreciate the part played by the masses in the creation of money . . . Therefore it is not possible for him . . . to concede that money represents something more and stronger than the will of participating individuals. A money for which a mass habit of acceptance has once been established is no longer the mere result of the individual aims of leaders whom the masses follow. Neither in the beginning nor later did the leaders have in mind a social institution . . . The final form of money is not a mere resultant; because of the universal social resonance that it awoke it represents a tremendous strengthening of their endeavors. (Wieser, 1914/1927, 1967, p. 165, italics added)

From this passage, it follows that the intervention of the masses cannot be interpreted as an act of pure recognition of the social utility of leaders’ decisions. For reasons that still need to be elaborated upon, the masses tend to create a final rule ‘far beyond [leaders’] expectations’ (ibid.). This suggests again an interpretation of Wieser’s conception of the emergence of institutions in terms of conventions (or social norms) or in terms of the supervenience approach (see Dupuy, 1992, p. 247). Such an interpretation is also consistent with Wieser’s description of some of the characteristics of the social dynamics between leaders and masses. As we have shown, for Wieser, this dynamics is not only governed by purely economic interests but also includes some extra-economic sociological laws such as the law of power, the law of increasing social stratification, the law of upward mobility of classes or the role of success. The resulting effect of these mechanisms is all but determinist; nor can it be considered as welfare improving. From this viewpoint, Wieser differentiates himself from Menger, who implicitly assumes that organic institutions serve the common welfare. Additional arguments taken from Wieser’s writings may be put forward in support of this view. First, Wieser makes reference to ‘inner rules’ such as inertia effects or self-destruction mechanisms that underlie the law of power, as the following quotation exemplifies: A social group, once it has been formed into a unit by the sacrificium voluntatis of its members, cannot easily be jolted by the sacrifices which it demands of them. Once success has induced leaders and masses to go together, failure will not automatically induce separation in spite of the losses caused by it. (Wieser, 1926/1983, p. 26)

Those mechanisms belong to the psychology of power and constitute what Wieser refers as the ‘supra-social’ or ‘anti-individual’ or even

Carl Menger and Friedrich von Wieser on the role of knowledge 93 ‘anti-social’ character of power, which stands ‘in complete reversal to the law of success’ (ibid., p. 71). They also contribute to explain the emergence of collective wholes or social entities that have acquired some autonomy vis-à-vis individuals. This ‘holistic’ feature does not, however, lead to neglect of the role of individuals within social dynamics. Wieser indeed explains that personal strength is at the origin of the growing of power but, ‘by aligning itself with the strength of a like-minded individuals, [it] is being enhanced way beyond its inherent potential[,] [a]longside it, there is a strengthening of the feeling of power, though at the same time strength in no small degree is being deprived of its personal roots’ (ibid., p. 70). Moreover, by emphasizing the relation of individuals to their neighbourhood but also to society considered as a whole entity, Wieser is able, in contrast to Menger, to deal with the issue of the evolution (and not only of the emergence) of institutions. For instance, Wieser points out the possibility of conflicts between a new historical task and existing historical powers (ibid., p. 203), or the existence of tensions regarding the sharing of power between the leadership strata and the strata representing the masses (ibid., p. 52). In mathematical terms, leaving aside the well-known reservations of Austrian economics about the use of mathematical relationships, these conflicts or tensions could be assimilated to path-dependence or hysteresis effects. To conclude, Wieser is far from being a faithful follower of Menger. His overall scheme of thought displays original thinking on several issues. First, on methodological grounds, Wieser’s combination of individualism and holism and of economic analysis and economic sociology is very insightful and demonstrates quite convincingly, by using the method of successive approximations, the limits of pure economics, in particular regarding power considerations. From this perspective, his contribution is comparable to that of Pareto (see Ragni, Chapter 2 in this volume) or Schumpeter (concerning Schumpeter, see the introduction to Arena and Dangel-Hagnauer, 2002). Second, by facing the problem of the influence of individual and social knowledge as well as systems of beliefs about economic and social phenomena, he offers a broader conception of rationality, where beliefs and action cannot be dealt with separately but determine each other. This concern, which is today very lively among social philosophers and sociologists, makes Wieser’s contribution highly topical. Third, Wieser’s contribution is also very enlightening regarding institutional matters. In particular, it provides some foundations for an analysis of the dynamics of institutions based on interlocking groups of agents and conflicting interests rooted in power and social strata.

94

Handbook of knowledge and economics

4.4 CONCLUDING REMARKS Menger’s and Wieser’s approaches to the problem of the emergence of institutions have some common features. On the one hand, both see it as the ‘unintended social results of individual tendencies’ (Menger, 1927/1967, p. 165). On the other hand, both authors introduce a distinction between leaders and followers or masses, which underlies a further distinction between innovative and imitative behaviours. However, for Wieser, those two kinds of behaviour may overlap, because they are also rooted and subject to the law of power. This original feature of Wieser also permits him to provide an analysis not only of the emergence but also on the evolution of institutions. Furthermore, they both appreciate the role of knowledge in economic and social phenomena in general, and in the more particular case of the emergence of institutions. As I have shown, both explanations of the emergence of institutions involve the interplay of individual and collective knowledge (or beliefs). If they both attempt to analyse the phenomenon of economic or social institutions from the perspective of interactions of individuals, they differ from one another as regards the dynamic process underlying those interactions. On one hand, Menger takes for granted the involuntary formation of shared knowledge about the validity of social institutions such as money. On the other hand, Wieser favours an explanation whereby collective beliefs are more than shared knowledge since they have some autonomy vis-à-vis individuals. As I have stressed, Wieser’s emphasis on the psychology of masses and leaders leads him to consider the influence of compulsion forces, besides the forces of freedom or ‘natural controls’, on the historical formations underpinning institutions, whereas they are discarded by Menger. This also explains why Wieser views historical development and social institutions as radically non-deterministic, and possibly welfare damaging, while Menger implicitly assumes, judging from his analysis of the emergence of money, that they are always welfare enhancing. To this extent, Menger’s analysis is limited to the emergence of institutions, viewed as a ‘discovery’ process, while Wieser’s is more focused on the dynamics of institutions, seen as a creative–destructive process.

NOTES * 1.

We thank the International Center for Economic Research (ICER, Turin) for its financial support. I acknowledge comments and suggestions by Richard Arena, Massimo Egidi and André Orléan on an earlier version of this chapter. This renewed interest has already swept many fields in economics such as decision theory, game theory, finance and organizational theory as well as neighbouring

Carl Menger and Friedrich von Wieser on the role of knowledge 95

2.

3. 4. 5. 6. 7.

8. 9. 10.

11. 12. 13. 14. 15.

16.

17.

disciplines such as sociology, psychology or social philosophy. This increasing interest in analysing the role of knowledge and beliefs in economic phenomena is also characterized by increasing confusion regarding the use of the concepts of knowledge and beliefs in the various fields of social sciences. It is clear that a more systematic and general reflection on the role of knowledge and beliefs still needs to be carried out, but this is beyond the scope of this chapter (see Arena and Festré, 2006a). Apart from the most famous Austrian authors, one should mention, in particular, Fritz Machlup, who is less known as an Austrian economist but also contributed to this field of research. His contribution is summarized in three of the ten projected volume series: Knowledge: Its Creation, Distribution and Economic Significance, published respectively in 1980, 1982 and 1984. For a study of the place of knowledge and economic beliefs in the second generation of the Austrian School see Arena and Festré (2006b). Cf. O’Driscoll and Rizzo (1984) on the opposition between real time and Newtonian time. In support of this interpretation, see, for instance, Birner (1999) and Garrouste (1999). Hayek defines knowledge as a ‘system of rules of action supported and modified by rules indicating similarities and differences between combinations of stimuli’ (Hayek, 1978, p. 41). Menger’s conception of knowledge is indeed difficult to specify precisely because it represents a very far-reaching form of knowledge. It includes general and explicit knowledge such as scientific knowledge but also some more local forms of knowledge that might be assimilated to the ‘knowledge of the circumstances of time and space’, to use Hayek’s terminology. Concerning these more local forms of knowledge, one can distinguish between explicit kinds of local knowledge that are articulable on one hand, and tacit, unconscious and non-articulable local knowledge on the other (see Fleetwood, 1997, pp. 164–6). This dilemma can be expressed as follows: if a firm always seeks specialization, this can be done only at the expense of adaptability in face of unexpected changes on the demand side. On this point, cf. Dulbecco and Garrouste (1999). In passing, this is an issue that Hicks will address seriously and that gives his contribution a strong Austrian complexion. From a methodological viewpoint, the case of the emergence of money or more generally of organic institutions comes under the ‘exact orientation of economic research’, while pragmatic institutions refer to the ‘empirical realist approach’ (Menger, 1883/1963, pp. 55–61). Menger coined the German word Absatzfähigkeit to refer to the property of marketability by merging two words: Absatz, meaning something like ‘possibility of sale’ or ‘to find a market for’, and Fähigkeit, meaning ‘capability’ or ‘ability’. In the sense that a more marketable good is more exchangeable and then becomes even more marketable. Referring to Orléan’s typology of imitation (informational, self-reference and normative imitation). One may refer here to some kind of knowledge creation following Nonaka and Takeuchi (1995). The process of the emergence of money can also be depicted as a ‘stochastically stable strategy’ (Young, 1998; see Garrouste, 2003, pp. 110–11). In other words, the institution of money emerges partly as the result of chance, and partly as the consequence of agents’ better knowledge of the intrinsic properties of the commodity that is likely to be commonly accepted. As Stigler (1941) puts it, Wieser’s contribution on capital theory ‘occupies a position of indisputable importance in the history of economics’ and he ‘presented one of the best theories of capital which had emerged’ in his time (Stigler, 1941, pp. 158, 177; quoted by Endres, 1991, p. 68). For Wieser, there are not one but two theories of distribution: the first measures the efficiency of productive services; the second determines the allocation of wealth. This

96

18.

19. 20. 21.

22. 23.

24.

Handbook of knowledge and economics distinction is made by a French sociologist/economist, Roche Agussol, in ‘Friedrich von Wieser’, Revue d’Economie Politique, 1930, nos 4 and 5, p. 53, quoted by F. Perroux in his introduction to the French translation of Schumpeter’s Theory of Economic Development, Paris: Dalloz, p. 31. As indicated by Warren J. Samuels in his introduction to the 1983 English translation of Das Gesetz der Macht, as early as in Social Economics, power is taken seriously into account since it constitutes, together with the principle of utility, the ‘twin organizing principles’ of Wieser’s theoretical economic framework. This feature had not gone unnoticed by Oskar Morgenstern, who wrote in his obituary to Wieser that Social Economics had been the ‘greatest systematic treatise that has been written by an Austrian in which the principle of marginal utility is analyzed in all its ramifications’ (Morgenstern, 1927, p. 671, quoted by Samuels in the introduction of the Law of Power, Wieser, 1926/1983, p. xv). This is a feature that he shares with Schumpeter. See Festré and Garrouste (2008). In passing, it is interesting to note that Menger’s distinction between pragmatic and organic institutions is purely formal since he neither elaborates on, nor give examples of, pragmatic institutions. Here again, one may be tempted to apply Menger’s typology of pragmatic versus organic to the notion of leadership. The analogy is, however, quite superficial. What is arguable, however, is that Wieser is more focused on historical formations and anonymous leadership and that he privileges causal–genetic explanations, which places him squarely in Menger’s tradition in this respect. In contrast to Menger’s treatment of organic versus pragmatic institutions, as already pointed out in this chapter. In passing, this quotation could be put in perspective with the notion of heteronomy, borrowed from political philosophy, and defined as the subordination or subjection of individuals to the law of another or to something else that individuals fail to see. This notion is in sharp contrast with the widespread idea among Austrian authors of the autonomy of the individual vis-à-vis the state. As rightly noted by J.-P. Dupuy, this boils down to the theoretical problem of the articulation between two kinds of autonomy: (1) the autonomy of the individual freed from any relation of subordination towards the sacred, the state or the social whole; (2) the social autonomy, which does not mean that men control society, but quite the opposite: the society escape them; it seems to be endowed with a life of its own, foreign to the individuals that form it (see Dupuy, 1992, p. 247). This law defines the second law of social growth, the first one being, as already mentioned, the ‘law of increasing social stratification’.

REFERENCES Arena, R. (2003), ‘Economic agents and social beliefs in the Austrian tradition: the case of Friedrich von Wieser’, Rivista Internazionale di Scienze Economiche e Commerciali, 50 (3), 291–309. Arena, R. and Dangel-Hagnauer, C. (2002), The Contribution of Joseph Schumpeter to Economics: Economic Development and Institutional Change, London and New York: Routledge. Arena, R. and Festré, A. (2006a), Knowledge and Beliefs in Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Arena, R. and Festré, A. (2006b), ‘Knowledge and beliefs in economics: the case of the Austrian tradition’, in R. Arena and A. Festré (eds), Knowledge and Beliefs in Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 35–58. Arena, R. and Gloria-Palermo, S. (2001), ‘Evolutionary themes in the Austrian tradition: Menger, Wieser and Schumpeter on institutions and rationality’, in P. Garrouste and

Carl Menger and Friedrich von Wieser on the role of knowledge 97 S. Ioannides (eds), Evolution and Path Dependence in Economic Ideas, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 133–47. Arrow, K.J. (1955), ‘Economic aspects of military research and development’, RAND Corporation Memorandum D-3142, 30 August (unpublished). Arrow, K.J. (1962), ‘Economic welfare and allocation of resources for invention’, in R. Nelson (ed.), The Rate and the Direction of Inventive Activity: Economic and Social Factors, New York: NBER, pp. 609–26. Birner, J. (1999), ‘The surprising place of cognitive psychology in the work of F.A. Hayek’, History of Economic Ideas, 7 (1–2), 43–84. Boudon, R. (1981), The Logic of Social Action, London and Boston, MA: Routledge and Kegan Paul. Dulbecco, P. and Garrouste, P. (1999), ‘Toward an Austrian theory of the firm’, Review of Austrian Economics, 12, 43–64. Dupuy, J.-P. (1992), Le sacrifice et l’envie (Sacrifice and Envy), Paris: Calmann-Lévy. Endres, A.M. (1991), ‘Austrian capital and interest theory: Wieser’s contribution and the Menger tradition’, The Review of Austrian Economics, 5 (1), 67–90. Festré, A. and Garrouste, P. (2008), ‘Rationality, behaviour, institutional and economic change in Schumpeter’, Journal of Economic Methodology, 15 (15), 1–26. Fleetwood, S. (1997), ‘Hayek III: the necessity of social rules of conduct’, in Stephen Frowen (ed.), Hayek: The Economist and Social Philosopher – A Critical Retrospect, London: Macmillan, pp. 155–78. Garrouste, P. (1999), ‘Is the Hayekian evolutionism coherent?’, History of Economic Ideas, 7 (1–2), 85–103. Garrouste, P. (2003), ‘Learning in economics: Austrian insights’, in S. Rizzello (ed.), Cognitive Developments in Economics, London: Routledge, pp. 302–15. Hayek, F.A. (1952), The Sensory Order: An Inquiry into the Foundations of Theoretical Psychology, Chicago, IL: University of Chicago Press. Hayek, F.A. (1978), New Studies in Philosophy, Politics, Economics, and the History of Ideas, London: Routledge and Kegan Paul. Kirzner, I.M. (1973), Competition and Entrepreneurship, Chicago, IL: University of Chicago Press. Lachmann, L. (1977), Capital, Expectations, and the Market Process: Essays on the Theory of the Market Economy, edited with an introduction by Walter E. Gringer, Kansas City: Sheed Andrews and McMeel. Machlup, F. (1980), Knowledge: Its Creation, Distribution, and Economic Significance (vol. 1, Knowledge and Knowledge Production), Princeton, NJ: Princeton University Press. Machlup, F. (1982), Knowledge: Its Creation, Distribution, and Economic Significance (vol. 2, The Branches of Learning), Princeton, NJ: Princeton University Press. Machlup, F. (1984), Knowledge: Its Creation, Distribution, and Economic Significance (vol. 3, The Economics of Information and Human Capital), Princeton, NJ: Princeton University Press. Menger, C. (1871), Grundsätze der Volkswirtschaftslehre, English translation: Principles of Economics, New York and London: New York University Press (1976). Menger, C. (1883), Untersuchungen über die Methode der Sozialwissenschaft und der Politischen Oekonomie insbesondere, English translation: Problems of Economics and Sociology, Urbana, IL: University of Illinois Press (1963). Menger, C. (1892), ‘On the origin of money’, Economic Journal, 2, 239–55. Mises, L. von (1949), Human Action: A Treatise of Economics, 4th rev. version, San Francisco, CA: Fox and Wilker (1996). Morgenstern, O. (1927), ‘Friedrich von Wieser, 1851–1925’, American Economic Review, 17 (4), 669–74. Nonaka, I. and Takeuchi, H. (1995), The Knowledge-creating Company: How Japanese Companies Create the Dynamics of Innovation, New York: Oxford University Press. O’Driscoll, G.P. and Rizzo, M. (1984), Time and Ignorance in Economics, New York: Basil Blackwell.

98

Handbook of knowledge and economics

Orléan, A. (1988), ‘The evolution of imitation’, in P. Cohendet, P. Lleran, H. Stahn and G. Umbhauer (eds), The Economics of Networks, Berlin and New York: Springer, pp. 325–39. Polanyi, M. (1967), The Tacit Dimension, New York: Anchor Books. Richardson, G. (1990), Information and Investment, 2nd edn, Oxford: Clarendon Press. Stigler, G.J. (1941), Production and Distribution Theories, New York: Macmillan. Streissler, E. (1987), ‘Wieser, F. von’, in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of Economics (vol. 4), London: Macmillan. Streissler, E. (1988), ‘The intellectual and political impact of the Austrian School of Economics’, History of European Ideas, 9 (2), 191–204. Wieser, F. von (1914), Theorie der Gesellschaftlichen Wirtschaft, English translation: Social Economics, New York: Adelphi Co. (1927). Reprinted in ‘Reprints of Economic Classics’, New York: Augustus M. Kelley (1967). Wieser, F. von (1926), Das Gesetz der Macht, Vienna: Julius Springer. English translation: The Law of Power, Bureau of Business Research, University of Nebraska, Lincoln (1983). Young, P. (1998), Individual Strategy and Social Structure. An Evolutionary Theory of Institutions. Princeton, NJ: Princeton University Press.

5

The pragmatist view of knowledge and beliefs in institutional economics: the significance of habits of thought, transactions and institutions in the conception of economic behavior Véronique Dutraive

5.1 INTRODUCTION A major trend of modern economics is to emphasize the role of institutions in economic phenomena, long after the work of old institutionalist economists. The revival of the theory of institutions owes much to North’s contribution to economic history and growth macroeconomics based on property rights and transaction costs. Subsequently, stemming from the influential works of Coase and Williamson, the interest in institutions has spread to the theory of the firm and of economic organization, and more widely to microeconomics. One of the important questions now discussed is the relation between institutions and cognition, and North himself has drawn attention to this point (Denzau and North, 1994, p. 5). The competences and capabilities theories of the firm and more widely evolutionary economics are now connecting evolutionary processes and cognition (Egidi and Rizzello, 2004). In this context, old American institutionalism, in particular the contributions of Veblen and Commons, have been rediscovered and recognized as having anticipated some of the trends of contemporary economic analysis (Hodgson, 2004; Rutherford, 2001; Samuels, 1995). Indeed, Veblen and Commons, dissatisfied with the conception of human agency in the exclusive terms of the theory of rational choice, pioneered the focus on the importance of institutions for our understanding of the dynamics of economic phenomena in modern societies and made early claims that economics must have some interaction with other sciences and particularly with the psychological sciences. There is a common agreement that Veblen’s and Commons’s conceptions of knowledge rely on American pragmatist philosophy, a philosophy that offers an original perspective on mental processes. The underlying ontology and general view of human behavior entail a distinctive methodology. In order to resume this thesis, the first part of the chapter deals with how 99

100

Handbook of knowledge and economics

pragmatism treats knowledge and beliefs while the rest of the chapter explores its transcription to institutional economics according to Veblen and Commons.

5.2 THE PRAGMATIST THEORY OF KNOWLEDGE Pragmatism is an original trend of American philosophy that emerged at the end of the nineteenth century. The main distinction from European philosophy of mind, focused on consciousness, and from empirical English philosophy, focused on sensual perception, is that pragmatism focuses on experience and activity (Deledalle, 1998, p. 14). Pragmatism contrasts both with empiricism and rationalism, which share the idea that decision is primarily based on reason. The first trend considers that mental activity is the reflection of the external world while the second assumes that it only depends on our internal world (the characteristics of mind). A third idea is related to idealism, according to which there is no external world independent from our mind. Pragmatism breaks with all these philosophical traditions, emphasizing the connection of cognitive functions to experience and action. Human beings are not in search of pure reality and truth. Their intellectual functions are related to more global, vital functions and are not separated from activities that are (more or less directly) necessary for their survival. ‘A pragmatist . . . turns away . . . from bad a priori reasons, from fixed principles, closed systems, and pretended absolutes and origins. He turns towards concreteness and adequacy, towards facts, towards action, and towards power. That means the empiricist temper regnant and the rationalist temper [are] sincerely given up’ (James, 1907, p. 20). ‘Theories thus become instruments, not answers to enigmas, in which we can rest’ (ibid., p. 21). Different variants of pragmatism may be distinguished: the logical and semiotic approach of Peirce’s interpretation, the psychological perspective of James’s version and the emphasis on education and democracy in Dewey’s works. These differences notwithstanding, the following sections discuss the unified ideas of pragmatism to be found in all three perspectives. 5.2.1

The Nature of Mind and of Reality: Mental Process and Beliefs

Pragmatism rejects the dualism between subject and object, mind and body, thought and action, theory and action, and so on. All pragmatists were impressed by the Darwinian revolution in science and philosophy (for instance Dewey, 1910), an implication of which being what is now

The pragmatist view of knowledge 101 called ‘the naturalization of mind’. This process of naturalization implies the necessity of entering into the system of mind and its cognitive functions. Thought is considered to be one of the biological means of achieving existence. On the one hand, matter-of-fact life and activity stimulate the mind and, on the other hand, a fundamental function of thought is to ensure the accomplishment of actions. Human beings are not related to the external word by passive perceptions but by intentions and assumptions about the successive effects of their actions.1 Accordingly, human beings are connected to reality by mental preconceptions. Since such preconceptions may be clear or confused, Peirce laid down the pragmatist foundation with the maxim ‘How to make our ideas clear’: ‘Consider what effects, that might conceivably have practical bearings, we conceive the object of our conception to have. Then, our conception of these effects is the whole of our conception of the object’ (Peirce, 1878, CP5. 402). This means that there is no immediate relation with reality but a perception of things mediated by their anticipated effects and their practical consequences.2 Peirce’s maxim implies that the prospective effects and consequences of action are beliefs that are the object of a cognitive elaboration. A main function of thought is to ‘fix a belief ’ and to ‘settle an opinion’ about the consequences of our actions because we cannot act in a state of doubt about this anticipated result. Thus the ‘quest for certainty’ (the title of one of Dewey’s books) is a cardinal topic of the pragmatist conception of knowledge. ‘Doubt is an uneasy and dissatisfied state from which we struggle to free ourselves and pass into the state of belief; while the latter is a calm and satisfactory state which we do not wish to avoid, or to change to a belief in anything else’ (Peirce, 1877, CP5. 372). A belief is a guiding principle for our desires and ‘a general disposition to act’ (Bain quoted by Deledalle, 1998, p. 56). Peirce further considers that a ‘belief is of the nature of a habit’ (Peirce, 1877, CP5. 377). ‘The feeling of believing is a more or less sure indication of there being established in our nature some habit which will determine our actions’ (ibid., CP5. 372). These habits are dispositions to act, instinctive and non-conscious in part, but that can also be subjected to control and reflexivity (Tiercelin, 1993, p. 33).3 A habit is not a repetitive mechanical conduct but a potentiality to observe a general rule applied in similar contexts (a ‘would be’, according to Peirce). Habits are thus intelligent dispositions to reproduce a behavior, having already produced satisfactory results in past experience. How do such beliefs arise? Peirce distinguishes four modes of ‘the fixation of belief ’ (Peirce, 1877). The first is the method of ‘tenacity’. It is a method of self-enforcing credence in which one avoids confrontation with experience (because the state of belief implies a feeling of security).4 However, this type of method can be easily destabilized by contrary beliefs

102

Handbook of knowledge and economics

because individuals are always confronted with other persons with different beliefs. A second method is that of ‘authority’, corresponding to situations in which some institutions (political, religious and so forth) impose their will upon individuals through various doctrines. Such collective beliefs have the capacity to coerce individuals’ minds through education and enforcement (punishment). But the existence of other doctrines related to other institutions or countries, and the sometimes necessary violence perpetuating beliefs, cannot definitely shield one from doubt. The third approach is the a priori method conveniently summarized by the idea that we believe in what is ‘agreeable to reason’. In the same vein of inspiration, James defined a true idea as an idea whose consequences bring satisfaction. Nevertheless, according to Peirce, there is a last and superior method: the ‘scientific method’.5 The scientific method is the best one to ‘fix our beliefs’ because it is the only one to be grounded in experience. According to Peirce and other pragmatists, this method is not limited to the professional scientist, since ‘everybody uses the scientific method about a great many things, and only ceases to use it when he does not know how to apply it’ (ibid., CP5. 385). Accordingly, in the pragmatist perspective, knowledge is made up of beliefs that are necessary for learning and for behaving because one cannot do anything in a state of doubt and insecurity. At the same time, the way our beliefs are formed resembles the scientific method because it relies deeply on experience. 5.2.2

The Pragmatist Ontology and Method: Relation and Inquiry

The pragmatist point of view does not admit a discontinuity between subjects and objects but considers them to be in relation. The relations are, therefore, the ontological entities.6 Relations create things and not the other way around. A consequence of this ontology is that, in contrast with both rationalist and empiricist conceptions, for which (T)ruth is static in time and space, pragmatism develops an evolutionary perspective of truth and reality. Thought does not describe a definitive external world but an ever-changing reality. Moreover, an idea or a concept is active not only in the conceptual realm but also causes actions that consequently change reality (Lapoujade, 1997). Dewey in particular developed this evolutionary perspective (Dewey, 1910).7 Dewey, like other pragmatists, rejects traditional dualisms. Specifically, fundamental entities are participants in what he calls ‘transactions’. There is no (ontological) discontinuity between living organisms and their biological and social environment. Behaviors and knowledge are not considered as the inherent expressions of individuals but as the result of a process that relies on the organism and

The pragmatist view of knowledge 103 its environment in a specific situation, the ‘transaction’. James well illustrated this idea when saying that ‘I’ is a name of a position just as ‘this’ and ‘here’ are (quoted by Deledalle, 1998, p. 104), which means that ‘I’ is at the crossroads of different types of transaction. In pragmatist terms, the knowledge of this reality in process arises from the method of ‘inquiry’ that rests on creative hypotheses and experiences. As emphasized above, this inquiry is oriented by the search for stable beliefs out of a state of doubt. However, this reference to doubt must be differentiated from the ‘Cartesian radical doubt’ (the state of mind that Descartes mentions in order to demonstrate the exercise of rationality by isolation from all environmental influences) because it is in contradiction with the nature of the cognitive operations of the active mind. Knowledge is always dependent on previous beliefs and preconceptions acquired in experiments and not on axioms or a priori assumptions. Accordingly, Peirce does not support the hypothetico-deductive method of investigation for the ‘fixation of opinion’. Moreover, if induction is indispensable, observation (and perceptions) cannot be isolated from previous representations. Peirce defines a specific procedure of inference in order to account for human creativity in the process of inquiry and to surpass the epistemological opposition between induction and deduction. He calls ‘abduction’ the inference that begins with a hypothesis (new idea) different from what has already been observed (i.e. different from induction which is the inference of the existence of something from observation of some similar cases). Abduction does not begin with observation but with a supposition of a general principle, which, if it were true, would explain facts as they are. Abduction is a creative insight, a hypothesis produced by the mind in order to solve a problem arising from experience and breaking a previous established belief. Scientific methods do not monopolize the use of abduction since all types of mental processes begin with abduction. Such is the specificity of the pragmatist method of science that it can be generalized to all types of knowledge. In fact, the process of inquiry combines three types of inference: we form hypotheses (abduction), of which we infer the effects (deduction); the consequences of the hypotheses we elaborate are, then, confronted with experience (induction). However, if the combination of the three types of inference is a source of progress in knowledge, abduction alone produces the new insights necessary for new knowledge. This process is continuous and our knowledge is always temporary, not only because reality is in perpetual evolution, but also because our ideas and our concepts contribute to the transformation of reality in the continuum of experience. Pragmatism seeks to define practical concepts that we need for our reasoning and for our actions instead of searching for an absolute, metaphysical truth.

104 5.2.3

Handbook of knowledge and economics The Collective Dimension of Knowledge

A last characteristic of the pragmatist theory of knowledge is that it does not subscribe to the idea of a private (subjective) criterion of knowledge. Peirce does not accept the Cartesian idea that doubt is voluntary and finally that cogito is a private affair. On the contrary, self-consciousness is not the starting point but the final term, the result of public or social interactions in which language, objects, other individuals and public beliefs are the real origins of self-consciousness. Among all the pragmatists, Mead most explicitly insisted on the social dimension of individual thought. While following Dewey in the evolutionary idea of a continuum between the human mind and its environment (transaction), Mead considers that this environment is not only natural, but also social.8 In the continuum of transactions, a person also influences society, and his thoughts and actions emerge from the problems generated by the environment because thoughts and actions are motivated by ‘problem-solving’ (and by the associated doubt). In the realm of science (in the Peircian sense of the best way of producing a stable belief), the ‘scientific method’ leads to knowledge preserved from the discretionary power of individuals because it stems from the community of opinion, the ‘common sense’ that is the result of the collective process of inquiry. Consequently, knowledge is not subjective but intersubjective. Kantian categories9 that mediate internal and external realities are not definitive but modified by experience. As for all other dualistic conceptions, pragmatism transcends the opposition between individualism and collectivism (or holism) or between the analytical priority of either parts or wholes. Dewey explicitly develops this stance regarding his discussions of democracy. Democracy is the political organization in which principles emerge from plural points of view (Peirce emphasizes that truth arises ‘probabilistically’ from the scientific or social community levels), because opinions come from experiences that are relative to the ‘multiplicity of worlds’ and consequently are diverse. For Dewey, thoughts and actions are motivated by ‘problem-solving’ (and by the associated doubt). Dewey thus stresses the importance of social cooperation in terms of problem-solving, underlines the importance of democracy and of broadly based political participation as a means of achieving social cooperation and finally argues for the primacy of education as the vehicle through which citizenship (democratic participation in social life) is accomplished. Above all, education (schooling) is seen as a community experience. But Dewey did not see the community as a set of social compromises constraining individuals and private claims. Indeed, individuals do not pre-exist the community and their identities and aspirations emerge as parts of the social links of the community in which they live (Berstein, 1991, pp. 128–9).

The pragmatist view of knowledge 105 If one reduces pragmatism to its main propositions, these would be: (1) knowledge is based on beliefs that are formed in the processes of our active minds and our experiences; (2) the nature of reality cannot be explored from the sole perspective of isolated individuals but only from that of relational entities; therefore, (3) the social (contextual) dimension and the individual dimension of knowledge are closely intertwined. These three related pragmatist ideas constitute the foundations of institutional economics’ conception of knowledge.

5.3 PRECONCEPTIONS, HABIT OF THOUGHT AND INSTITUTIONS IN VEBLEN’S EVOLUTIONARY CONCEPTION OF KNOWLEDGE AND BELIEFS The birth of institutional and evolutionary economics is frequently associated with Veblen. Despite some notable contributions to economic theory (in terms of business cycles, of consumption theory, of the theory of the firm etc.), Veblen is, above all, well known for his epistemological essays in the field of the theory of knowledge and especially of economic knowledge (Veblen, 1919a/1990). Even in the absence of explicit references, there is a strong similarity between Veblen’s theory of knowledge and pragmatist philosophy10 when it comes to questions of mind and of ontology, of scientific method and of the collective character of knowledge.11 Moreover, Veblen had extensive knowledge of philosophy.12 He explicitly built his conception of knowledge on Kant’s notion of judgment relative to mental processes and on Darwin’s theory of evolution and its philosophical implications for epistemology. These two authors were also major sources of inspiration for pragmatism. From the conjunction of these sources, Veblen builds an evolutionary theory of knowledge related to social and cultural considerations. 5.3.1

The Philosophical ‘Preconception’ in Veblen’s Theory of Knowledge

Veblen’s concern with the philosophical foundations of human action and social organization can be related to basic considerations about the nature of reality and of human beings. First, Veblen holds that each historical epoch (or type of society) is dominated by what was later called an episteme (Foucault, 1969) or a paradigm (Kuhn, 1962), that is, common attributes in the different areas of human knowledge and activities. Veblen’s own term is ‘preconceptions’. Human beings do not have a ‘brute’ perception of reality but a perception mediated by something similar to Kantian ‘a priori categories’. But

106

Handbook of knowledge and economics

contrary to Kant, these frameworks are not fixed; they are changing with time. Veblen argues that human mental processes are shaped by instinctual propensities and above all by the conditions of life, by the activity at hand as well as by social institutions. In accordance with pragmatism, Veblen defends ‘realist social theorizing’ (Lawson, 2002), allowing him to assert the impact of a changing reality, which is experienced, on thought and behavior. He writes: ‘the scheme of thought or of knowledge is in good part a reverberation of the scheme of life’ (Veblen, 1899/1990, p. 105). With the advent of modern society, the scheme of life is characterized by the impact of machinery and by the growing importance of the ‘place of science’ (particularly of empirical sciences), a process that progressively revolutionizes previous representations and preconceptions. According to Veblen, the episteme of modern times (the end of the nineteenth century and the beginning of the twentieth century) can be qualified as ‘post-Darwinian’. Darwin’s theory of evolution is, in fact, exemplary from the viewpoint of the method of inquiry and more generally of the conception of knowledge it implies. Veblen admitted the biological grounds of cognition and behavior, and was also interested in finding an equivalent mechanism of evolution for social organizations and institutions. Above all, Veblen retains from Darwinian theory its conception of causality (Hodgson, 2001) oriented by the search for the mechanisms of evolution untainted by any idea of predetermined finality. Thus Veblen distinguishes between two types of scientific theories:13 pre-Darwinian and post-Darwinian theories. Pre-Darwinian theories are considered ‘taxonomic’ because they essentially perform a classification of phenomena according to some preexistent norms of inquiry (for instance, progress, consistency, efficiency or equilibrium). Given the a priori finality of those theories, Veblen designates them as ‘teleological’. It is noteworthy that many of these theories establish ‘natural laws’. In this context, natural laws are ‘the rules of the game of causation. They [formulate] the immutable relations in which things “naturally” [stand] to one another before [a] causal disturbance [takes] place between them’ (Veblen, 1908/1990, p. 37). By contrast, Veblen argues in favor of the non-teleological character of the Darwinian method of analysis. Indeed, reasoning in terms of ‘consecutive change process’ is a characteristic of modern science. In post-Darwinian sciences, ‘the process of causation, the interval of instability and transition between initial cause and definitive effect, has come to take the first place in the inquiry’ (ibid.). Thus ‘modern science . . . is becoming . . . a theory of the process of consecutive change, which is taken as a sequence of cumulative change, realized to be self-continuing or self-propagating and to have no final term’ (ibid.). With its focus on the process of ‘opaque cause and effect’ (meaning without any presupposed finality), modern science has

The pragmatist view of knowledge 107 embraced the episteme of evolutionary knowledge because the process itself is investigated given that the chain of causes and effects never comes to rest and the result is undetermined. 5.3.2

Where do ‘Preconceptions’ come from and how do they Evolve?

Like the pragmatists, Veblen argues that mental processes are linked to actions and experiences. On the one hand, some habits of thought (beliefs) are formed by practical and social life and, on the other hand, beliefs determine decisions and actions. ‘It is by the use of their habitual canons of knowledge and belief, that men construct those canons of conduct which serve as guides and standards in practical life’ (Veblen, 1919b, ch. 1, p. 3). Veblen names these beliefs ‘preconceptions’ and distinguishes their two main interrelated sources: some stem from social institutions (organizational, political and legal rules), while others are related to practices and activities. Life in society imposes on people various rules of conduct built on norms, values and so on, which are internalized in the form of beliefs. Some beliefs, and the associated knowledge, are more related to practical, material situations that people face while going about their everyday life (they come from repeated empirical observation and manipulation). Both sources (social relations and material practices) are intertwined in the schemes of knowledge made up of habits of thought. Veblen speaks of ‘institutions’ when these habits of thought acquire a collective dimension, defining institutions as ‘settled habits of thought common to the generality of men’. Thus, according to Veblen, institutions are not only codified rules or organizations; they are collective beliefs and internalized rules of conduct. ‘The fabric of institutions intervenes between the material exigencies of life and the speculative scheme of things’ (Veblen, 1908/1990, p. 44). Following the realist perspective – in the pragmatist sense – Veblen argues that reality affects beliefs and that these evolve with time. Throughout history, collective beliefs (as well as scientific theories) were characterized by a ‘teleological’ interpretation of facts, from religious beliefs to the belief in natural laws. However, in Veblen’s view, the time of an ‘impersonal’ conception of phenomena has dawned. In other words, the epistemological implication of Darwinian evolutionary theory is precisely an account of ‘cumulative causation’ not relying on ‘animistic’ or ‘anthropomorphic’ preconceptions. For Veblen, such must be the maxim of knowledge in the social sciences because of its correspondence with the very nature of an ever-changing and undetermined reality. Moreover, this also constitutes Veblen’s conception of ontology: ‘an ontology of (the genetic process of) cumulative causation’ (Lawson, 2002, p. 289). The evolutionary canon of knowledge applied to social phenomena allows one to understand both

108

Handbook of knowledge and economics

the nature of social order (the stability of institutions) and the process of change. More precisely, institutions, which are crystallized collective habits of thought, have the double property of inertia and transformation. In this context, Veblen proposes his own conception of institutional change based on the evolution of habits of thought and of beliefs. Depending on the varying influence of the material conditions of existence, Veblen distinguishes the ‘matter-of-fact institutions’ from the ‘ceremonial institutions’ which are more ‘regardless of facts’. The rhythms of change of these institutions differ, the ‘less matter-of-fact’ being less plastic than the ‘more matterof-fact’ (Veblen, 1919b). Moreover, prevailing institutions select behaviors that thus enforce their predominance. This process of ‘cumulative causation’ can give rise to the type of phenomenon that modern evolutionary economists call ‘path-dependence’. Given this general framework of institutions ‘likened to strata that shift slowly at different rates’ (Hodgson, 1994, p. 26), some institutions adapted to a past ‘state of material arts’ can persist (in the self-reinforcing process) while the material conditions of life and the associated habits of thought have long since evolved. Thus changes in the institutional framework often emerge from contradictions. 5.3.3

Human Behavior, Typology of Knowledge and Social Process

Veblen considers that social sciences such as economics are based on preconceptions going back to eighteenth-century economic conditions and episteme, and have not yet embraced the Darwinian revolution. In particular, he argues against ‘associationist psychology’ on which much of economic theory is based on the grounds that it depicts an overly passive and individualistic view of mental processes. ‘The later psychology [James’s pragmatism in other words], re-enforced by modern anthropological research, gives a different conception of human nature. According to this conception, it is the characteristic of man to do something, not simply to suffer pleasures and pains through the impact of suitable forces’ (Veblen, 1898/1990, p. 74). Veblen considers that economic rationality must be embedded in an evolutionary context as well as in a cultural system of beliefs (institutions) that are ‘dispositions for action’. Various instinctual propensities (Veblen, 1914/1922) of human nature lead to the rejection of the behavioral preconceptions of economic theory: the instinct of ‘workmanship’ (invoked against usual economic assumptions of effort aversion), the ‘parental bent’ (as opposed to the idea of exclusively self-seeking behavior), ‘idle curiosity’ (as opposed to the idea that actions are guided solely by pecuniary interests) and the ‘predatory instinct’ (contrary to the belief in the natural harmony of interests).

The pragmatist view of knowledge 109 Two of these instincts can be related to the typology of knowledge based on the distinction between ‘scientific knowledge’ and ‘pragmatic knowledge’ related to different types of activities. Idle curiosity plays an essential role because of its connection to scientific knowledge that, as with the ‘aptitude of play’, is a very useful activity not motivated by ‘expediency’ but by a disinterested and imaginative inclination to inquiry. Such idle curiosity must be distinguished from ‘pragmatic intelligence’, which is the basis of the ‘instinct of workmanship’, defined as ‘the response into the form of a reasoned line of conduct looking to an outcome that shall be expedient for the agent’ (Veblen, 1906/1990, pp. 6–7). Veblen underlines the fact that this type of intelligence implies ‘teleological’ interpretations in the sense that facts are always assumed to be associated to some purposes. Thus human cognition is related to human actions because human beings want to exercise causality on things. The knowledge based on experience only is not enough to satisfy the need of anticipating the actions’ effects. Indeed, experience ‘can, at the best, give what is or what has been, but cannot say what is to be’ (Veblen, 1884/1998, p. 176). Human judgment is not based only on the data of experience (anti-empiricism) but relies on cognitive inference operations. In accordance with Kant, Veblen argues that the product of knowledge (theories, for instance) must be connected to a priori principles of the intellect and states, in a pragmatist perspective, that ‘the nature of [these] principle[s] is to be found from a consideration of the work it is to do’ (ibid., p. 180). Veblen qualifies this principle as ‘teleological’ because most of the time (except when it is the product of ‘idle curiosity’) human beings follow a purpose. Given this tendency towards action-oriented mental processes, human beings often commit the fallacious step of considering that the external world is itself oriented and goaldirected. In other words, ‘the finality which is attributed to external reality .  .  . is simply and only an imputed finality’ by an extension of our own human motives and intellectual frameworks (ibid., p. 186). In a Kantian spirit, Veblen considers causality as one of the a prioris of understanding but distinguishes different types of causality. ‘Sufficient reason’ and ‘efficient causality’ correspond respectively to the type of inference used in classical and neoclassical economics. Furthermore, the contrast between the two expressions crystallizes the distinction between the so-called finalist and mechanist philosophies of biology. In the finalist version, associated with Aristotle, living beings tend to an end assigned by nature itself in which a directing principle or finality is inscribed. The mechanist conception, associated with Descartes, holds that natural beings can be compared to machines, driven by some mechanical and physical causality. In Veblen’s view, ‘the preconceptions of economic science’ hinge somewhere between these two positions, both

110

Handbook of knowledge and economics

of which yield an incoherent view of individual behavior. On the one hand, actions are determined by reasons, by the conscious reference to predefined ends. In this perspective, an agent’s cognition is reduced to a pure rational consciousness (Veblen, 1898 speaks of the ‘lightning calculator’) and his behavior is nothing more than the result of an instrumental evaluation of the odds. On the other hand, economic actions are mechanical reactions to some causes (‘efficient causality’) that constitute a process in which the agent is like a particle that reacts mechanically and without resistance to a combination of external forces (like ‘a globule of desire’). Veblen, highly critical of these conceptions of causality, introduces a different type of inference discovered in ‘Darwinian science’ (Veblen, 1909/1990, pp. 237–8), namely a process of ‘cumulative causation’ between the individuals’ actions and their environment. In this perspective, a causal explanation is no longer a ‘teleological’ explanation (in terms of natural law or of intentionality).14 More precisely, human intentionality is embedded in a larger context of accumulation of step-by-step causal mechanisms (Hodgson, 2003). Thus Veblen’s main message is that while most behaviors are teleological, the phenomenon resulting from the interaction of these behaviors cannot be explained in terms of teleology but in terms of cumulative causation. Put differently, ‘while knowledge is constructed in teleological terms, in terms of personal interest and attention, this teleological aptitude is itself reducible to a product of unteleological natural selection. The teleological bent of intelligence is an hereditary trait settled upon the race by the selective action of forces that look to no end’ (Veblen, 1906/1990, p. 5). Causality in social sciences such as economics cannot be understood from the point of view of the individual intentions but from the evolutionary method of thought based on ‘cumulative causation where both the agent and his environment [are] at any point the outcome of the last process’ (Veblen, 1898/1990, p. 74). The last section of this chapter discusses some other implications of pragmatist philosophy for the institutionalist conception of knowledge that can be found in the works of Commons.

5.4 ‘FUTURITY’, COLLECTIVE ACTION AND THE COGNITIVE DIMENSION OF INSTITUTIONS IN COMMONS’S INSTITUTIONALISM Commons shares some of Veblen’s fundamental views of economic behavior and of social order as well as some methodological principles characterizing institutional economics, but introduces several original aspects.

The pragmatist view of knowledge 111 The similarity relates to the evolutionary appreciation of social order, to the defense of an institutional analysis enabling an understanding of social phenomena in which economic actions are embedded, and to the recognition of the main habits and rules of modern capitalism. However, Commons emphasizes the importance of both formal rules of legal institutions and organizations in contemporary society. In Commons’s view, individuals take part in different types of ‘collective actions’ that have become the main sources of economic phenomena. Moreover, contemporary society exhibits a certain capacity for ‘artificial selection’ of those activities emerging from the evolutionary process. Depending on the type of political governance, this selection can be oriented towards the satisfaction of either ‘public welfare’ or some ‘vested interests’. In the following subsections, this particular conception of social order and organization is related to the theory of knowledge and of economic behavior. 5.4.1

Hume, Peirce and Weber: Contributions and Shortcomings in Terms of a Theory of Economic Knowledge

Commons’s conception of knowledge builds on the double influence of Hume’s radical empiricism and skepticism and of American pragmatism (Commons, 1934/1990, pp. 140–57). Hume’s ‘skepticism’ implies that knowledge cannot be rooted in pure rationalism. This view is also shared by pragmatism and institutionalism. Hume is also qualified as a ‘radical empiricist’ due to his belief that ideas originate from experiences and sense data. However, contrary to Locke’s empiricism, Hume did not profess the belief that sense data could allow the discovery of the reality of the external world: percepts appear to the mind as impressions and ideas and all thought derive from some particular and contingent impressions. Therefore any abstract (complex) idea is a composition of simple ideas that are articulated by some operation of association, such as resemblance, difference, contiguity or succession. Sophisticated operations of the mind – of which the process of reasoning – are associated with a repetition of impressions, that is, habits that contribute to form beliefs. Thereafter, knowledge relies on beliefs that imply some inferences of previous similarities and no longer need to be observed. This is the case of ‘causality’ based on habitual relations between impressions that give way to expectations of the future from observed past regularities. Hume’s presentation, in which experience, habits and beliefs play important roles, may thus be said to have anticipated pragmatism in some respects. Nevertheless, for Commons, Hume’s is a much too passive vision of the mind, essentially a simple ‘receptacle of impression .  .  . whereas Peirce’s idea is that of an active, continuing organizer and reorganizer

112

Handbook of knowledge and economics

of impressions [and of] active beliefs reaching forward for future action’ (ibid., p. 153). Subjective ideas come from sense data but the same fact could give rise to different interpretations. According to Commons, in a pragmatist perspective, mental tools confer ‘meanings’ and ‘valuation’ (variable depending on the circumstances and the purposes of an action and influenced by ‘habitual assumptions’ or preconceptions) rather than ‘pure knowledge’. This point is particularly crucial for the social sciences (Ramstad, 1986). Indeed, social phenomena are different from physical phenomena due to the intentionality of human action. Building on this point made by Weber (and the German Historical School), Commons argues that the specificity of scientific explanation of social facts implies that one refers to the meanings and values of those facts for the actors from whom they originate.15 In this sense, the construction of knowledge in the social sciences is a particular compilation of the ordinary cognitive processes (of the common man). This being said, Commons offers a typology of ‘ideas’ arranged according to their increasing complexity. (1) Percepts come from the senses but previous beliefs (that Commons calls ‘habitual assumptions’ or ‘insights’) confer ‘meanings’ and classifications of importance (valuation). (2) Concepts are elaborated through the cognitive process of ‘analysis’ consisting of a classification of facts according to the similarities or differences of their attributes. (3) Principles are ideas capturing the similarity of actions in their temporal dimension consisting of a classification of facts according to the similarities or differences of their causes, effects and purposes (an operation dubbed ‘genesis’). (4) Formulas are mental constructions resulting from operations of ‘synthesis’ understood as mental pictures of external reality. The external world is apprehended, according to the pragmatist view, in terms of the changing relations between parts and wholes. (5) Finally, social philosophies are systems of explanation with a performative dimension, the classification of which depends on the similarities or differences of their goals (ibid., pp. 93–108). Commons considers that his predecessors had insufficiently investigated two main factors associated with the specificity of human beings and social facts. The first is what Commons calls ‘futurity’ (Commons, 1950, pp. 104–9), which captures the fact that from the perspective of human behavior and economic action, economic values are strongly associated to the representation of the future state of society. The second point Commons insists on is that individual actions come within numerous collective actions such that these should be the ultimate unit of analysis of economic phenomena (Commons, 1934/1990, pp. 144–50). Building on these philosophical considerations, Commons offers his original institutionalist contribution based on ‘futurity’ and ‘collective action’.

The pragmatist view of knowledge 113 5.4.2

Futurity and the Security of Expectations: Psychological Principles, Causal Mechanisms and the Foundations of Capitalism

Human beings are pragmatic beings, meaning that their thoughts are connected to their activities and oriented by goals. ‘Futurity’ encompasses the mental field of beliefs that guides actions in relation to ends to be attained by the means of actual activities. Commons argues that the problem of causality should be viewed not from the past to the present but from the future to the present, because projects and beliefs orient actual actions. However, Commons uses the term ‘futurity’ instead of ‘future’ because the future is uncertain and unknowable, even through the cognitive process of backward induction, whereas ‘futurity’ is knowable as it is an actual projection shaped by objectives and possibilities of action and allowed by actual rules of conduct. A psychological consequence of this principle of futurity (already outlined by pragmatist psychology) is that ‘the security of expectations’ is a necessary condition that prevails over (and is a condition of) selfish interest. In the realm of economic activity, this security is particularly necessary for the functioning of a market economy. Without this security of expectations, there would be little or no present value, present enterprise, present transactions, or present employment. Value is present worth of future net income. No school of economists was clear-cut on these time dimensions . . . One reason why early economists did not separate out future time for special investigation in their theories of value was the assumption, taken from the physical sciences, that cause precedes effect. Labor precedes its product; sensations precede action; scarcity and want precede effort and satisfaction. But here is an effect that precedes its cause. (Commons, 1950, p. 104, emphasis in original)

Economic values are associated with the representation of the future state of society. Thus, economic theory must rely on this idea of ‘futurity’ because in modern capitalism ‘the present transfer of legal control .  .  . take[s] effect in the future production and consumption, or labor process. Production and consumption cannot be carried on without first obtaining legal control (which must precede physical control). Possibly, this changes the idea of causation. It places causation definitely in the future instead of in the past’ (Commons, 1934/1990, p. 7). Numerous economic institutions support the importance of this time lag between legal control and the physical realization of economic transactions and their consequences for decision. Property is an actual right of legal control over the future use of an asset; money is a debt (the issuance of which permits the evaluation of economics quantities); the modern conception of capital, including ‘intangible assets’ (such as patents, goodwill,

114

Handbook of knowledge and economics

trademarks, corporate franchises and ‘various rights to do business’, whose present value depends on their expected exchange value or expected income), implies that part of the value of capital is based on the expectation of future earnings. Consequently, a market economy relies on the confidence in the future, on the expectation of the acquittal of all debts in good faith, on the continuity of activities and so on. In other words, that business will go on as planned. Such confidence, in Commons’s view, could not rely solely on a mechanism of natural, spontaneous harmonization of individual interests and on a social order based on contracts only. Institutionalism offers a different representation of social order that underlines the nexus of rules organizing collective action. Seen from the point of view of the theory of knowledge, one must adopt a vision that surpasses individuals’ ideas and beliefs. 5.4.3

The Formulas of Collective Action and the Cognitive Dimension of Social Rules: Transactions, Going Concerns and Institutions

Commons follows Hume’s idea that scarcity implies conflict instead of a harmony of interests. Not surprisingly, Commons raises a classical question in social philosophy: if we start with individual interests in conflict, why is it that most of the time society does not collapse in generalized violence and chaos? Commons believes that there is something more than just individual interactions, a ‘visible hand’ of human institutions instead of an ‘invisible hand’ of natural laws. Therefore individual action should not be taken as the ultimate unit of analysis of social facts. The unit of analysis must resume three dimensions of all social relations: conflict (related to scarcity); dependence (efficiency through cooperation); and order (resulting from an equilibrium between the two preceding dimensions). Instead of individuals, Commons states that the ‘transaction’ provides such ‘a unity of activity’ and should therefore be taken as the unit of analysis. The important idea, here, is that this view of transactions cannot be reduced to an intersubjective and voluntary contractual relationship between two self-interested individuals (as in the ‘transaction cost theory’). The transaction should be viewed in its pragmatist sense of a relation, which, as an ideal-type (that Commons calls ‘formula’), is primary over participants. Individual preferences and goals, and the means of attaining these ends, are shaped by a nexus of working rules of different types (moral, habitual, legal) and at different levels (local or general, organizational or institutional) that constitute incentives and/or obligations and/or possibilities of transaction. The issue is that if ones wishes to understand the logic behind a particular transaction, one must first establish the set of rules in which it is embedded. Regarding knowledge, rules are essential

The pragmatist view of knowledge 115 because no one would take part in a transaction without the supporting security of expectations. This does not imply that all transactional outcomes are perfectly knowable, without surprises and challenge. In this perspective, Commons proposes a taxonomy of transactions according to the type of knowledge involved. On the one hand, ‘routine transactions’ are those related to habitual activities involving stabilized knowledge (embodied in rules). On the other hand, ‘strategic transactions’ are those related to situations of novelty implying new practices and new opportunities and for which there is no stabilized knowledge. This captures a conception of rationality quite different from the traditional view, which reduces it to optimization and continuous deliberation. Most lasting transactions are supported by routinized practices. Not every one of them is subjected to mental examination (something that the human brain could not face) if previous experiences produced satisfactory results. The mind is consciously active only in case of innovation or change, when past habits are inappropriate. In these circumstances, the mind reveals ‘a creative agency looking towards the future and manipulating the external world and other people in view of expected consequences’ (Commons, 1934/1990, p. 7). According to this (pragmatic) conception of human behavior, any notion of ‘rationality’ – in fact, a strategic ability to exert power on the environment (natural or social) – depends in fine on habits and rules. Habits amount to the condition that reduces cognitive limits in the face of uncertainty and complexity; they are the basis of the process of learning, a condition for active and creative human behavior facing new situations (Bazzoli and Dutraive, 1999). The process of habituation is to be grounded in the social context of individual activities. Many of the elements of behavior and choice, namely preferences, information and knowledge acquisition, conceptual frameworks, representations and even perceptions and mental categorizations, are formed through the social context of experience. This context is that of various organizations and communities (that Commons calls ‘going concerns’ in order to express the idea that they are not fixed external structures but evolving sets of transactions and working rules) in which behaviors are learned and actions are set. Rules of various going concerns produce ends and means for action by offering various incentives and possibilities while setting the limits. These diverse going concerns are linked together by common rules in the more global social order. According to Commons, social order is not natural but institutional, an institution being understood as ‘collective action in restraint, liberation, and expansion of individual action’ (Commons, 1934/1990, p. 73). From this viewpoint, social order is a nexus of evolving rules, from beliefs and habits to collective customs of various going concerns and to legal and

116

Handbook of knowledge and economics

public rules. Institutions thus play a cognitive role. Indeed, individuals are not only externally coordinated by rules but are ‘institutionalized minds’ (ibid.) and ‘institutionalized personalities’ (ibid., p. 874). The embeddedness of their actions in institutions confers meaning on events and allows expectations related to repeated transactions and projects, some of which are of longer duration than the participants’ horizons. Beyond individual ideas, some collective ideas arise from ‘corporate actors’ and organizations out of individual interests and conflict. Commons argues that from the process of collective action, ‘reasonable’ beliefs orienting actions in the direction of ‘welfare’ instead of in that of ‘vested interests’ can emerge. The cement of society is not a sum of isolated peoples but a community of experience and values that emerge from a voluntary process of cooperation through collective action. In particular, it is through conflicts of interests giving rise to negotiated resolution processes that human beings voluntary acquire their ‘reasonable’ character. In accordance with Dewey, Commons regards democracy as a social process (grounded on the ‘social philosophy of reasonable values’) able to produce evolutionary social values serving the common welfare.

5.5 CONCLUDING REMARKS Institutionalism states that economics cannot be autonomous from philosophical preconceptions. It sees in the pragmatist philosophy of knowledge an alternative to the standard psychological and methodological preconceptions of economics. The specificity of social sciences vis-à-vis natural sciences lies in the fact that meanings are given to actions and their intentionality. Thus, if mental processes matter, the pragmatist’s perspective implies that ideas are related to activities and experiences, and that meaning and intentionality are contextual. Human cognition is not reducible to unrestricted utilitarian calculus. Instead of a simple individual mechanism that processes external data, mental processes are understood to be more active and less infallible. Rationality is extended to and made reliant on habits of thought. Routinized procedures are the condition for more sophisticated cognitive processes such as expectations and strategic calculations, which, consequently, lose their axiomatic status. Pragmatism carries strong ontological and methodological consequences for economic analysis. When one bases cognitive functions on action and experience, one introduces heterogeneity not only of individuals but also of social and historical contexts. This entails a greater uncertainty and complexity that discredits the idea of ‘substantive rationality’ and the predictability of the results of those actions grounded on it. Therefore individuals can no

The pragmatist view of knowledge 117 longer be the unit of economic analysis. This unit must be able to organize and stabilize expectations. According to institutionalists, due to bounded rationality, the reduction of uncertainty and conflict relies on habits and regularities that are the expressions of institutions. The institution is not immutable but is a relational category that connects (mediates) individuals and social contexts, or cognitive processes and social structures. Economic analysis should combine individual and institutional causalities in a process of mutually reinforcing influence based on a major consequence of pragmatism as an evolutionary method of thought. This type of inquiry seeks to blend the freedom of knowing, creative individuals and the self-organizational dimension of economic dynamics. Old institutional economics has, from this point of view, a great bearing on contemporary perspectives.

NOTES 1.

2.

3.

4.

Pragmatists do not share the empiricist idea that the external world is perceived through sense data or by pure reasoning. However, James qualified himself as a ‘radical empiricist’ because, in his view, perception of the world stems from the influence it produces on us and the influence we produce on it, resumed by experience. It is distinct from sensualist empiricism grounded on the duality of the internal and external world. Peirce also rejects idealist and nominalist positions according to which external objects are nothing more than the concepts or words that qualify them. Peirce’s position, on the contrary, makes claims for ‘realism’. He writes: ‘That which any true proposition asserts is real [emphasis in original], in the sense of being as it is regardless of what you or I may think about it’ (Peirce, 1905, p. 432). Put differently: ‘That those characters are independent of how you or I think is an external reality. There are, however, phenomena within our own minds, dependent upon our thought, which are at the same time real in the sense that we really think them. But though their characters depend on how we think, they do not depend on what we think those characters to be . . . Thus we may define the real as that whose characters are independent of what anybody may think them to be’ (Peirce, 1878, CP5. 405). According to James, ‘[T]ruth is an idea with consequences that we consider as satisfactory and that brings us some gratification. But this perspective must be distinguished from utilitarianism because “satisfaction” is not arbitrary or subjective, but submitted to the constraints of context and of previous and accumulated habits of thought’ (Lapoujade, 1997, p. 51). Tiercelin explains that Peircian pragmatism concerns rational and finalized action, close to the Kantian idea of ‘conduct’ that cannot be reduced to ‘behavior’ or to ‘action’. On this matter, Peirce differs from James’s ‘radical empiricism’ overly oriented by experience and psychology. Peirce suggests calling his own conception ‘pragmaticism’ in order to distinguish it from that of James (Peirce, 1905). ‘This simple and direct method is really pursued by many men. I remember once being entreated not to read a certain newspaper lest it might change my opinion upon free trade. “Lest I might be entrapped by its fallacies and misstatements”, was the form of expression. “You are not”, my friend said, “a special student of political economy. You might therefore easily be deceived by fallacious arguments upon the subject. You might, then, if you read the paper, be led to believe in protection. But you admit that free trade is a true doctrine; and you do not wish to believe what is not true”’ (Peirce, 1877, CP5. 377).

118 5.

6.

7. 8. 9. 10.

11. 12.

13.

Handbook of knowledge and economics ‘To satisfy our doubts, therefore, it is necessary that a method should be found by which our beliefs may be determined by nothing human, but by some external permanency – by something upon which our thinking has no effect . . . Our external permanency would not be external, in our sense, if it was restricted in its influence to one individual. It must be something that affects, or might affect, every man. And, though these affections are necessarily as various as are individual conditions, yet the method must be such that the ultimate conclusion of every man shall be the same. Such is the method of science. Its fundamental hypothesis, restated in more familiar language, is this: There are Real things, whose characters are entirely independent of our opinions about them; those Reals affect our senses according to regular laws, and, though our sensations are as different as are our relations to the objects, yet, by taking advantage of the laws of perception, we can ascertain by reasoning how things really and truly are; and any man, if he have sufficient experience and he reason enough about it, will be led to the one True conclusion. The new conception here involved is that of Reality’ (Peirce, 1877, CP5. 384). One of the major contributions of pragmatism to philosophy is Peirce’s ‘triadic’ phenomenology and semiotics. Stated in simple terms, Peirce’s phenomenology articulates three states of phenomena: ‘firstness’ is the possibility of existence of an entity; ‘secondness’ is the effect of the entity that qualifies its existence; ‘thirdness’ is the representation of the entity by an interpretation that makes it intelligible. Peirce’s realism, rejecting both rationalism and empiricism, considers that we attain reality through ‘signs’ in accordance with the triadic phenomenology (very different from Saussure’s semiotic distinction between the signifier and the signified): a sign is an ‘icon’ related to the abstract quality of the entity in itself, an ‘index’ representing the entity by virtue of real connections to it, and a ‘symbol’ of habitual representation of the entity. According to Peirce, thought can be seen as a process of interpretation of reality through an articulation of signs. Peirce’s contribution was of such importance that, after him, the theory of interpretation of signs was called ‘pragmatic’ (see Deledalle, 1998 and Tiercelin, 1993). Dewey’s conception is qualified as ‘instrumentalist’ because, in his view, all manifestations of mind, even the conscious ones, are considered as expressions of natural (biological) functions. This pragmatist orientation is called ‘instrumentalism’ because the human mind is not only considered as a product of natural selection and evolution but also as a tool of the transformation of the human environment. The a priori patterns of judgment, such as ‘time and space’, are independent of any experience. While, according to most commentators, there are some pragmatist influences on Veblenian ideas, this influence has been the object of debates. Veblen attended Peirce’s lectures on logic in 1881 at Johns Hopkins University (Griffin, 1998), but the former never explicitly quoted the works of the latter. Dewey was his faculty fellow at Johns Hopkins University, at the University of Chicago (which Mead attended as well) and at the New School of Social Research (Dorfman, 1934). But the only pragmatist work Veblen really referred to is James’s (1890) Principles of Psychology. Another difficulty is that Veblen frequently referred to pragmatism in the general sense of an ‘expedient conduct’ (Veblen, 1906/1990) instead of its philosophical sense (see below). The aim of the following account is not to compare Veblen’s views with pragmatist conceptions, which is a theme often examined (see for instance Dyer, 1986; Griffin, 1998; Mirowski, 1987; Samuels, 1990), but to introduce Veblen’s conception of knowledge. This interest is manifest in some of his first essays (which probably originated from his PhD) building on the works of Kant and Spencer. His dissertation, entitled ‘Ethical grounds of a doctrine of retribution’, has been lost since before 1935, according to Dorfman (1934). This typology is much more analytic than diachronic because Veblen considered that the economic science of his time, even when interested in economic change, had not yet operated its Darwinian epistemological revolution (Veblen, 1898/1990).

The pragmatist view of knowledge 119 14.

15.

‘It is, of course, true that human conduct is distinguished from other natural phenomena by the human faculty for taking thought, and any science that has to do with human conduct must face the patent fact that the details of such conduct consequently fall into the teleological form; but it is the peculiarity of the hedonistic economics that by force of its postulates its attention is confined to this teleological bearing of conduct alone. It deals with this conduct only in so far as it may be construed in rationalistic, teleological terms of calculation and choice. But it is at the same time no less true that human conduct, economic or otherwise, is subject to the sequence of cause and effect, by force of such elements as habituation and conventional requirements’ (Veblen, 1909/1990, pp. 238–9). Indeed: ‘the subject-matter with which the economist deals is not a mechanism or organism whose motions the investigator cannot understand, it is human beings whose activities he can fairly well understand by putting himself “in their place” and thus construct the “reasons” in the sense of motives and purposes, or values, of their activity’ (Commons, 1934/1990, p. 723) in order to ‘understand the reason why people act as they do under the particular circumstances selected’ (ibid., p. 725).

REFERENCES Bazzoli, L. and Dutraive, V. (1999), ‘The legacy of J.R. Commons’s conception of economics as a science of behavior’, in J. Groenewegen and J. Vromen (eds), Institutions and the Evolution of Capitalism: Implications of Evolutionary Economics, EAEPE, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 52–77. Berstein, R.J. (1991), ‘Dewey et la démocratie: la tâche qui nous attend’, in J. Rajchman and C. West (eds), La pensée américaine contemporaine, Paris: Presses universitaires de France (French translation of Post-Analytic Philosophy, New York: Columbia University Press, 1985), pp. 119–32. Commons, J.R. (1934), Institutional Economics: Its Place in Political Economy, New Brunswick, NJ: Transaction Publishers, 1990. Commons, J.R. (1950), The Economic of Collective Action, New York: Macmillan. Deledalle, G. (1998), La philosophie américaine, 3rd edn, Bruxelles: De Boeck Université. Denzau, A. and North, D.C. (1994), ‘Shared mental models: ideologies and institutions’, Kyklos, 47(1), 3–31. Dewey, J. (1910), The Influence of Darwin on Philosophy and Other Essays, New York: Henry Holt & Company. Dorfman, J. (1934), Thorstein Veblen and His America, New York: A.M. Kelley. Dyer, A. (1986), ‘Veblen on scientific creativity: the influence of Charles S. Peirce’, Journal of Economic Issues, 20(1), 21–41. Egidi, M. and Rizzello, S. (2004), ‘Cognitive economics: foundations and historical roots’, in M. Egidi and S. Rizzello (eds), Cognitive Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 1–22. Foucault, M. (1969), L’Archéologie du savoir, Paris: Gallimard, 1998. Griffin, R. (1998), ‘What Veblen owed to Peirce – the social theory of logic’, Journal of Economic Issues, 32(3), 733–57. Hodgson, G. (1994), ‘Precursors of modern evolutionary economics’, in R.W. England (ed.), Evolutionary Concepts in Contemporary Economics, Ann Arbor, MI: University of Michigan Press, pp. 9–35. Hodgson, G. (2001), ‘Darwin, Veblen and the problem of causality in economics’, History of Philosophy and Life Science, 23, 385–423. Hodgson, G. (2003), ‘Instinct and habits before reason: comparing the views of J. Dewey, F. Hayek and T. Veblen’, Behavioral Research Council conference on ‘Dewey, Hayek and Embodied Cognition: Experience, Beliefs and Rules’, Great Barrington, A, USA, 18–20 July.

120

Handbook of knowledge and economics

Hodgson, G. (2004), The Evolution of Institutional Economics: Agency, Structure and Darwinism in American Institutionalism, London: Routledge. James, W. (1907), Pragmatism: A New Name for Some Old Ways of Thinking, New York: Longman Green & Co. Kuhn, T. (1962), The Structure of Scientific Revolutions, Chicago, IL: University of Chicago Press. Lapoujade, D. (1997), William James: Empirisme et pragmatisme, Paris: Presses universitaires de France. Lawson, T. (2002), ‘Should economics be an evolutionary science? Veblen’s concern and philosophical legacy’, Journal of Economic Issues, 36(2), 279–92. Mirowski, P. (1987), ‘The philosophical basis of institutional economics’, Journal of Economic Issues, 21(3), 1001–38. Peirce, C.S. (1877), ‘The fixation of belief ’, popular science monthly, 12, November, 1–15, on line: http://www.cspeirce.com/menu/library/bycsp/fixation/fx-frame.htm. Peirce, C.S. (1878), ‘How to make our ideas clear’, Popular Science Monthly, 12, January, 286–302, on line: http://www.cspeirce.com/menu/library/bycsp/ideas/id-frame.htm. Peirce C.S. (1905), ‘What pragmatism is’, The Monist, 15(2), 161–81, on line: http://www. cspeirce.com/menu/library/bycsp/whatis/whatpragis.htm. Ramstad, Y. (1986), ‘A pragmatist’s quest for holistic knowledge: the scientific methodology of J. R. Commons’, Journal of Economic Issues, 20(4), 1067–105. Rutherford, M. (2001), ‘Institutional economics: then and now’, Journal of Economic Perspectives, 15(3), 173–94. Samuels, W. (1990), ‘The self-referentiability of Thorstein Veblen’s theory of the preconceptions of economic science’, Journal of Economic Issues, 24, 695–718. Samuels, W. (1995), ‘Present state of institutional economics’, Cambridge Journal of Economics, 19 (August), 569–90. Tirercelin, C. (1993), C.S. Peirce et le pragmatisme, Paris: Presses universitaires de France. Veblen, T. (1884), ‘Kant’s Critique of Judgment’, in Essays in Our Changing Order, New Brunswick, NJ: Transaction Publishers, 1998, pp. 175–93. Veblen, T. (1898), ‘Why is economics not an evolutionary science?’, in The Place of Science in Modern Civilization and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990, pp. 56–81. Veblen, T. (1899), ‘The preconceptions of economic science I’, in The Place of Science in Modern Civilization and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990, pp. 82–113. Veblen, T. (1906), ‘The place of science in modern civilization’, in The Place of Science in Modern Civilization and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990, pp. 1–31. Veblen, T. (1908), ‘The evolution of the scientific point of view’, in The Place of Science in Modern Civilization and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990, pp. 32–55. Veblen, T. (1909), ‘The limitations of marginal utility’, in The Place of Science in Modern Civilization and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990, pp. 231–51. Veblen, T. (1914), The Instinct of Workmanship and the State of the Industrial Arts, New York: Huebsch, 1922. Veblen, T. (1919a), The Place of Science in Modern Civilisation and Other Essays, New Brunswick, NJ: Transaction Publishers, 1990. Veblen, T. (1919b), The Vested Interests and the Common Man, New York: Huebsch. Electronic publication: http://etext.lib.virginia.edu/toc/modeng/public/VebVest.html.

6

Imagination and perception as gateways to knowledge: the unexplored affinity between Boulding and Hayek Roberta Patalano

6.1 INTRODUCTION In 1956, Kenneth E. Boulding claimed that human behaviour depends on mental representations. Every action, including economic action, is guided by the images of the world that the subject has built up in his/her mind. In those times, Boulding’s revolutionary message was not understood. His ideas remained widely unheard, especially by the economic public to which they were addressed. In the same period, cognitive scientists – and some economists among them – were reacting against behaviourism, thus bringing back mental phenomena to the centre of theoretical inquiry in social sciences. Herbert Simon supported the revolution from the inside: by highlighting the bounds of rationality he pointed out the necessity of exploring mental activity more deeply (Simon, 1955). Years later, when introducing the concept of ‘problem space’, he argued that solutions to problems are not found in the external world but in the mental representation of the problem that the agent elaborates (Newell and Simon, 1972). Although Boulding and Hayek were not directly involved in the cognitive revolution, two of their most eclectic contributions on knowledge pertain to these years. The Sensory Order: An Inquiry in the Foundations of Theoretical Psychology (Hayek, 1952) and The Image: Knowledge in Life and Society (Boulding, 1956) were both looking at the interaction between the individual cognitive structure and the external environment as a determinant of human behaviour. Perception, for Hayek, and imagination, for Boulding, were considered the main cognitive activities involved in the construction of knowledge. Recently, a new field of research has emerged and consolidated within economic theory from the contamination of economics with cognitive science (Egidi and Rizzello, 2004). Cognitive economists analyse economic phenomena with an interdisciplinary approach that is focused on the functioning of the human mind. From such a perspective, and as already suggested by Hayek and Boulding decades ago, the intermingling of 121

122

Handbook of knowledge and economics

perception, imagination and behaviour helps clarify the process through which knowledge is constructed. The main aim of this chapter is to develop a comparative analysis of The Sensory Order (1952) and The Image (1956). As we shall see, these works have much in common and, most of all, present an approach to knowledge that still appears to be innovative under different profiles. While analysing two distinct cognitive processes, perception and imagination, both the authors conceive the mind as an interpretative filter that classifies and elaborates information. The construction of knowledge is thus seen as a sort of metabolic process during which the objective dimension of information is gradually lost and transformed into a subjective content.1 The implications of such a perspective will be explored in detail. While published in a historical period that was particularly prolific for cross-fertilization among disciplines, these books have shared an unfortunate fate of oblivion in the history of economic ideas. If The Sensory Order was ‘rediscovered’ by economists in the mid-1970s, and, in the last decades, has been increasingly investigated in the literature (Butos and Koppl, 1993; Smith, 1997; Birner, 1999; Garrouste, 1999; Loasby, 1999; Rizzello, 1999a; Butos and McQuade, 2002; Arena, 2004; Caldwell, 2004; Rizzello, 2004), The Image is still little known. In order to increase its diffusion among economists and social scientists, the main contents of the book will be summarized and discussed in the next section of this chapter.

6.2

THE IMAGE: KNOWLEDGE IN LIFE AND SOCIETY

In 1956, after spending 11 months at the Center for Advanced Study in the Behavioral Sciences at Stanford University, Boulding2 wrote The Image. As he explained in a note of 1988, the draft of the book caused him an intense experience of contact with inner thoughts and feelings.3 At the end of August 1955 the entire text was actually dictated to the secretarial staff of the Center at Stanford in only nine days. In this book mental images are interpreted as a filter for perceptual data and mostly as an interface between human perception and knowledge. The author analyses how individual images are created and how they evolve or may be shared. A fundamental hypothesis concerns the existence of a direct and persistent influence of images on behaviour. As individual worldviews affect behaviour, the study of mental processes through which such worldviews are developed is a crucial step in the comprehension of human action. Boulding’s ideas are applied to biology, economics,

The unexplored affinity between Boulding and Hayek

123

political philosophy, organization theory, history and sociology. The discussion covers 11 chapters written in a thought-provoking style and are quite accessible to a wide range of readers. Everyone has a self-image, which includes a picture of his/her location in space, his/her acknowledgement of being part of a time flow, perception of the universe as a world of regularities and the sensation of being part of a human relational network. He/she then has an image to evaluate reality (‘value image’), which intervenes in his/her relationship with the external environment, infusing information with meanings, an ‘affectional or emotional image’ that gives rise to feelings, attitudes and motivations, and a ‘public image’ that helps him/her to compare personal with collective views. Knowledge is built up through a process of subjective selection and interpretation of external data that is guided by the image. In Boulding’s words, ‘we do not perceive our sense data raw; they are mediated through a highly learned process of interpretation and acceptance’ (1956, p. 14). The meaning of an information flow depends on its consequences for the image.4 If the image remains unaffected, the information has not been absorbed5 by the subject and cannot be integrated into his/her pre-existing knowledge. By contrast, when new information becomes part of the image it may stimulate a process of learning, for example a reorganization of the image in accordance with the new data. Through such a mechanism of learning and reorganization of knowledge, the subject may modify their worldviews over time. According to Boulding, apart from the conceptual context in which it is analysed, the image is characterized by plasticity, dependence on the personal living of the subject and resistance to change. These regularities are likely to make mental images temporarily stable, despite their dynamic potentiality. As clarified by the author, the image contains elements that enhance its evolutionary tendency as much as internal obstacles to its own development. Change in the image originates from the adjustment to new messages received from the environment. Such process of adjustment may involve different phases: our image is itself resistant to change. When it receives messages which conflict with it, its first impulse is to reject them as in some sense untrue .  .  . As we continue to receive messages which contradict our image, however, we begin to have doubts, and then one day we receive a message which overthrows our previous image and we revise it completely. (Boulding, 1956, p. 9)

As this passage indicates, the author emphasizes the role played by exogenous factors, such as the pressure and recurrence of messages, to stimulate the revision of images.

124

Handbook of knowledge and economics

Moreover, Boulding introduced ‘resistance to change’ among the factors that influence economic action. Having acknowledged that the obstacles to change may come from the cognitive attitude of agents, he argues that conservative attitudes at the representational level also influence market behaviour. The stickiness of prices, for example, may be caused by the lack of a clear and widely understandable image of what goes on in the market: The buyer or seller in an imperfect market drives on a mountain highway where he cannot see more than a few feet around each curve; he drives it, moreover, in dense fog. There is little wonder, therefore, that he tends not to drive it all but to stay where he is. The well-known stability or stickiness of prices in imperfect markets may have much more to do with the uncertain nature of the image involved than with any ideal of maximizing behavior. (ibid., pp. 85–6)

As this passage appears to suggest, resistance to change may represent a strategy to face uncertainty. When information is not fully known, avoiding changes may be a choice that helps agents stabilize their behaviour and make it more predictable.6 To explain the evolutionary path of the stock of images that exists in society, an analysis of the factors that shape individual imagery (e.g. learning processes) is required. This approach turns out to be particularly relevant in economic contexts because the process of reorganization of economic images through messages is the key to understand economic dynamics. The great over-all processes of economic life – inflation, deflation, depression, recovery, and economic development – are governed largely by the process of reorganization of economic images through the transition of messages. (ibid., p. 90)

It is not so much the quantity of information that matters, nor the speed of its circulation, but the existence of images that are able to absorb incoming messages without strongly resisting change.7 Boulding comes to this conclusion by considering the following points. The impact of a message on the image depends on resistance to change, which is stronger if the message contradicts the image. On the contrary, ‘messages which are favourable to the existing image of the world are received easily’ (ibid., p. 13). Messages that are often repeated or that come from a source that the subject regards as authoritative have good chances of weakening resistance; however, resistance also depends on internal qualities of the image, such as its logical consistency and the existence of aesthetic relationships among its parts. The more consistent and elegant the image is, the more it resists revision: ‘The resistance may take the form of simply ignoring

The unexplored affinity between Boulding and Hayek

125

the message, or it may take the form of emotive response: anger, hostility, indignation’ (ibid., p. 12). While representing ambivalent grounds for change, individual imagery has a relevant social function because it enables sharing of values and meanings. As Boulding says, ‘the basic bond of any society, culture, subculture, or organization is a “public image”, that is, an image, the essential characteristics of which are shared by the individuals who participate in the group . . . Indeed, every public image begins in the mind of some single individual and only becomes public as it is transmitted and shared’ (ibid., p. 64). It is interesting to investigate the mechanisms that allow the sharing of the image at the social level. Interpersonal communication is considered by the author as the main channel through which private images come into contact with those of others and start to overlap. The ‘universe of discourse’ is thus a conceptual space in which collective images emerge as a result of prolonged interaction based on conversation and personal exchange. Boulding also considers the role played by symbols and cultural background by acknowledging that ‘a public image almost invariably produces a “transcript”; that is, a record in more or less permanent form which can be handed down from generation to generation’ (ibid.). Through intergenerational transmission of legends, rituals, proverbs and ceremonials, people come to share a set of traditions and informal rules of behaviour. They also conform to a similar interpretation of symbols that is then reinforced by the establishment of institutions that make use of the same symbolic language. On the one hand, such language reflects a set of meanings and values shared by subjects at the imaginative level; on the other, it is externalized through the emergence of institutions. By way of institutional sanctioning, the interpretation of symbols becomes durable, and even more stable in the case of institutions themselves. At the collective level, the image also has cohesive power, which may exert a strategic function both in an organizational context and in cooperative interaction.8 At the same time it may enhance the building up of shared expectations about future market performance. Such an effect is the more effective the more society is supplied with a widespread information structure that reaches a large number of market niches. In such contexts, cohesion at the imaginative level can consolidate existing images and make them more stable through self-enforcement processes, even when they are not adequate for interpreting external reality. The presence of constitutive regularities in the image underlines its unifying power as a concept through which a better understanding of human

126

Handbook of knowledge and economics

behaviour and social dynamics may be achieved. In order to develop an interdisciplinary perspective on the topic, Boulding suggests founding a new science, ‘Eiconics’, which is intended to be a conceptual space for cross-disciplinary connections. In the final chapter of the book the author explores some philosophical implications of his approach.

6.3

THE IMAGE: SOME REMARKS

In 1890, Marshall placed imagination among the three basic conceptual instruments that any economist should use for his work: The economist needs the three great intellectual faculties, perception, imagination and reason: and most of all he needs imagination, to put him on the track of those causes of visible events which are remote or lie below the surface, and of those effects of visible causes which are remote or lie below the surface. (Marshall, 1890/1961, p. 43)

In Marshall’s words, the main role of imagination is that of guiding the theorist ‘below the surface’ to grasp what cannot be understood at first glance. For Boulding, the role of imagination is even more subtle. Imagination allows the subject to build up a mental picture of reality. Any level of understanding, even the most superficial, is not possible without the image that filters, interprets or ignores the information received from the mind. With great originality, Boulding centres his theory of knowledge on the imaginative capacity of individuals, a feature that is still almost neglected by contemporary economic literature despite its significant influence on market behaviour (Patalano and Rizzello, 2003; Patalano, 2010).9 Moreover, Boulding’s intuitions on imagination, and specifically the existent link between images and behaviour, are still valid in the light of a new approach to decision-making that emerged in psychology (Beach and Mitchell, 1987; see note 9) and has been recently extended to economics (Patalano, 2005a, 2005b). According to Boulding, imagination has a cognitive role. By filtering and interpreting raw information, it favours the metabolic production of knowledge. Beyond this function, imagination is also seen as the main source of creativity. For example, in the case of technological change the innovator takes advantage of the ability to modify his current image of the productive system by introducing new instruments and tools. Not accidentally, the presence of unconventional social groups, ethnic minorities or diversified educational systems increases the innovative potential of society by contributing to the emergence of less stereotyped images of the cultural and economic context.10

The unexplored affinity between Boulding and Hayek

127

Boulding does not explicitly explain the mechanisms through which differentiation among individuals comes about. However, the idea that imaginative aptitudes are not the same for everybody is often mentioned. On the one hand, he attributes the subjective dimension of imaging to its dependence on personal living and experiences. On the other, when innovative potentialities are considered, the author appears to suggest that individuals possess a different propensity to creativity, which is fundamentally innate, and when present, may be enhanced by a favourable and non-conformist environment (see also Section 6.2). An interesting affinity of thought exists between Boulding (1956) and Penrose (1959) on the role of imagination in entrepreneurship. Penrose makes an important distinction between the ‘objective’ opportunities of the firm and the ‘subjective’ ones, suggesting that the latter depend on the mental attitude of the entrepreneur.11 Even more explicitly, she talks of an ‘image’ that orients the entrepreneur’s judgement about the environment: Within the unknowable limits placed by the environment on successful action there is a wide scope for judgements. We shall be interested in the environment as an ‘image’ in the entrepreneur’s mind, for we want, among other things, to discover what economic considerations, as contrasted with ‘temperamental’ considerations, determine entrepreneurial judgements about the environment. (Penrose, 1959, p. 42)

For Boulding and for Penrose, imagination seems to be the main source of subjective evaluation of economic reality.12 What can be seen as an opportunity of profit, for example, depends highly on the individual view of the market. Imagination guides the entrepreneur by shaping his/her process of selection between projects that have innovative potential.13 When interpreted as a guide to behaviour, imagination also represents an interface between the agent and the society. By evaluating imagined perspectives, the agent makes his/her choices. From this point of view, imagination is not merely an individual resource. Images may be shared at the social level, thus acquiring cohesive power or enhancing coordination among agents. The process through which images are socialized leads to developing common visions of reality and, consequently, common modalities of behaviour (see also note 7). The main effect on society is the emergence of multiple forms of coalitions, all characterized by the sharing of images and worldviews among their members. The same process of cohesion that is effective in bringing organizations together, for example having a common organizational vision, may act also for political parties, social groups and ethnic minorities (Boulding, 1956).14 From the point of view of a strictly rigorous scientific analysis, The Image may sometimes appear vague or elusive. Boulding’s approach is

128

Handbook of knowledge and economics

certainly based on intuitions more than on fully developed models. In particular, the concept of ‘image’ is never explicitly defined by the author, who, however, seems to be aware of this lack as, for example, he asks: What, however, determines the image? This is the central question of this work. It is not a question which can be answered by it . . . One thing is clear. The image is built up as a result of all past experience of the possessor of the image. Part of the image is the history of the image itself. (Boulding, 1956, p. 6)

From the perspective of cognitive science, the main question raised by Boulding’s concept of ‘image’ addresses the distinction (if any) of mental images from other forms of mental representations. Does the term ‘image’ in his work refer to a generic kind of representation that could be alternatively defined as a ‘mental model’ or a ‘set of beliefs’? Or, on the contrary, does the concept refer to a specific form of symbolic representation with unique figurative characteristics?15 Theoretically, at least two answers are possible. First, mental images are equivalent to other representational formats, such as beliefs or schemas. If so, the most original insight that Boulding develops in his work concerns the dependence of human behaviour on mental representations, which he defines generically as images. Second, the ‘image’ has a figurative content that could not be entirely transmitted through language. Moreover, the visual nature of the image defines its symbolic characteristics that make it different – and thus not reducible to – verbal statements (see also note 11). From this perspective, the analysis of imagining suggests distinguishing among multiple modalities of thought that influence behaviour at different levels of awareness and involve the activation of distinct brain areas. Patalano (2005a, 2005b) has explored such an interpretative line16 that is, however, still little analysed by economic literature. Indeed, recent advances in cognitive economics have been mostly concerned with the idea that agents express their knowledge through sets of propositions, mental models and systems of interrelated beliefs. The specificity of symbolic and iconic forms of knowledge representation – and their impact on behaviour – have not yet been addressed systematically by economists.

6.4

THE SENSORY ORDER

Among the many works of Friederich A. von Hayek on the relationship between knowledge and economics, The Sensory Order occupies the most

The unexplored affinity between Boulding and Hayek

129

eclectic position. Hayek started this book when he was very young and did not know then if he wanted to be an economist or a psychologist, and worked on it for more than 20 years (Hayek, 1952). The Sensory Order examines the psychology of perception and has some interesting points in common with the theory of cellular assembly that was elaborated by Hebb (1949).17 Hayek compares the mind to a classifying structure that does not receive sensory stimuli passively but directs and interprets them: ‘The transmission of impulses from neuron to neuron within the central nervous system . . . is thus conceived as the apparatus of classification (Hayek, 1952, p. 52). First of all, the mind acts on perceptual data as a framework that orders perception through interpretation. The process through which sensorial data are interpreted is guided by the neuronal structure of each individual, which associates classes of stimuli with classes of responses in accordance with past neural activity. As a result, ‘the qualities which we attribute to the experienced objects are strictly speaking not properties of that object at all, but a set of relations by which our nervous system classifies them’ (ibid., p. 143). Hayek uses the concepts of ‘map’ and ‘model’ to describe the perceptual mechanisms in more detail. The ‘map’ is a plastic structure made up of the neural linkages that have been created by the brain on the grounds of past experience. The ‘model’ is, instead, ‘the pattern of impulses which is traced at any moment within the given network of semipermanent channels’ (ibid., p. 114). The model is thus generated by the map in accordance with the stimuli that the subject is currently experiencing. Through the interaction of the map and the model, the past perceptual experience and the present one interact. Interpretation at the neural level is thus dependent on two main factors: the genetic phenotype and the previous sensorial experience. Genetic characteristics are relevant as they provide each subject with a constellation of neurons and neural connections (synapses) that then evolve over the life of the individual under the influence of experience. The development of neural circuitries is mainly shaped by learning: when a new stimulus reaches the brain and is integrated in the pre-existing class of stimuli and responses, it is likely to alter the structure of the class, thus promoting a process of reorganization of synapses. As a result, previous sensorial experience plays a crucial role in the interpretation of perceptual data because it moulds the neural structure that has been genetically inherited, into accordance with personal history. Furthermore, as underlined by Hayek, starting from its native structures the brain continuously evolves over time and modifies its way of processing external data without awareness. Recent neurobiological research confirms Hayek’s major ideas and underlines that they form an interesting basis for the understanding of the

130

Handbook of knowledge and economics

interaction between innate aspects of cerebral structures and their possible development towards unforecastable directions (Damasio, 1994; Paller, 2001). The development of the mind and specifically its ability to create meanings on the basis of perceived information depends on the presence of neural connections that have existed since the individual’s birth and change according to new experiences. Changes include functional evolution of neural groups, learning how to perform new ‘tasks’ when the individual faces unexplored situations and the recombination of synaptic connections into a configuration that is more suitable to a present situation. An interesting consequence of Hayek’s approach concerns the historical character of the brain’s functioning. If any stimulus is interpreted on the grounds of previous perceptive experience, then sensory events that have no link to anything in the perceptual past of the subject cannot be interpreted or classified. Information is then comprehensible only if it recalls something already familiar (Dempsey, 1996). On the theoretical grounds provided by his inquiry into the mechanisms of perception, Hayek differentiates between information and knowledge. By processing the information that has been acquired from the environment at neural level, the economic agents build up their personal knowledge. Information has, thus, an objective dimension and consists of the external data that can be discovered in the market. Knowledge is, instead, the fruit of the subjective elaboration of external information, through a series of neurocognitive processes that start from perception (see also note 1). In the mind of the subject a cognitive dynamic takes place that leads from undifferentiated data perceived in the environment to mental representations of the world. Perception is the first source of subjectivity in such a dynamic. Similarly, Boulding argues that knowledge is the final result of a ‘metabolic process’ that takes place in the mind of the subject in order to elaborate the informational flows that are filtered by the image.

6.5 UNEXPLORED AFFINITIES BETWEEN THE IMAGE AND THE SENSORY ORDER 6.5.1

Some Historical Remarks

As far as the history of ideas is concerned, The Sensory Order and The Image were edited in a cultural climate that was exceptionally prolific for the birth of The Mind’s New Science (Gardner, 1985). At the beginning of the 1950s, many different strands of thinking were pointing out

The unexplored affinity between Boulding and Hayek

131

the fallacies of behaviourism and the cognitive revolution made its first appearance. Launched by J.B. Watson in 1913, behaviourism had aimed at cutting off mental phenomena – such as desires, goals, intentions and, more generally, mental states – from the study of behaviour. The latter was explained entirely in terms of stimulus–response associations that were considered as observable and, thus, objectively testable. Having gained wide consensus throughout American psychology, by the mid-1950s the behaviourist movement was in its decline. During 1956 two important symposia pointed towards a significant turning point for the cognitive revolution.18 At the same time, Wiener’s cybernetics was gaining consensus and the Harvard Center for Cognitive Studies, founded by Bruner, became the reference point for psychologists who were searching for new foundations for their discipline (Miller, 2003). Important communication channels with psychologists abroad were also opened up. Bartlett’s approach to memory, Piaget’s insights into child psychology and Luriia’s study of the mind also started to exercise their influence outside European borders. As Miller notes, ‘By 1960 it was clear that something interdisciplinary was happening. At Harvard we called it cognitive studies, at Carnegie-Mellon they called it informationprocessing psychology, at La Jolla they called it cognitive science’ (Miller, 2003, p. 142). As far as economics is concerned, in the 1950s experimental results started to question the validity of the standard model of rational choice. Allais’s paradox of 1952 and the empirical study of decision processes in firms conducted by Cyert, Simon and Trow in 1956 were pointing to the roles played by mental attitudes to risk and uncertainty in decisionmaking and organizational contexts. The limits of rationality and the bounded capacity of human beings to process information were highlighted by Herbert Simon, who actively supported the cognitive revolution.19 In such a cultural climate, the need for a more realistic analysis of economic behaviour was emerging. Boulding and Hayek were not involved in the cognitive revolution directly and did not intentionally embrace its project. Nevertheless, in retrospect, The Image and The Sensory Order appear to be perfectly congruent with the debate on the functioning of the mind that animated social sciences in the 1950s. Moreover, by stressing the relevance of mental activity to explain economic phenomena, Boulding and Hayek’s contributions may be considered as fertile ground for the ongoing dialogue between economics and cognitive sciences, as we shall argue in the following sections. Important differences and complementarities among the authors will also be considered.

132

Handbook of knowledge and economics

6.5.2

The Subjective Nature of Knowledge

For both Boulding and Hayek, knowledge has a subjective nature because the cognitive process through which it is obtained requires the involvement of individual elements, mainly genetic inheritance and personal history. The mind is conceived as the intermediate variable through which the objective dimension of information is processed and transformed into the personal dimension of acquired knowledge. From this perspective, it is not a passive receptacle but an active and adaptive instrument. Such approach is in open contrast with behaviourism, as explicitly remarked by Boulding.20 One of its philosophical antecedents may be found in the pragmatist tradition, especially in the works of Peirce (1905) and Dewey (1929). The conception of the mind as an interface between the external environment, which is common to all agents, and the ‘internal environment’ (Simon, 2000), which differs for every individual being, has remarkable implications for economic theory. Boulding’s and Hayek’s approach to knowledge does not seem to be compatible with the traditional concept of the market as populated by anonymous agents. The process through which information is selected and then interpreted crucially depends on the images of the subject and on his/her neural structure. Being based on subjective characteristics, the construction of knowledge cannot lead to the same output for every agent. The cognitive dynamic through which knowledge is built, in fact, impresses a personal blueprint on its content. As a main consequence, even in contexts where – for reasons of simplicity – we might assume that the same information is equally accessible to anybody, the knowledge acquired by agents would differentiate them from each other (Rizzello, 1999b; see also note 1). Furthermore, the knowledge of one agent might not ‘represent’ that of any other, depending on its construction of unique and irreplaceable elements. An important difference between Boulding and Hayek concerns the factors that may explain the subjective nature of knowledge. Hayek stresses the neuropsychological underpinnings of the processes through which information is elaborated. Thanks to the order that exists in the mind and that is sustained at the neural level by the joint action of the model and the map, the subject interprets the sensations that have been extracted from the environment and builds up his/her representations of reality. The subjective nature of such representations is due to the differences that emerge among individual neural circuits. At birth, agents inherit the genetic traits that are embedded in the structure of their brain. From then on, the linkages among neurons evolve and combine according to personal

The unexplored affinity between Boulding and Hayek

133

history and environmental influences, within the semi-permanent bounds defined by the map. In building up knowledge, everybody subjectively elaborates information that is received from the environment. Perception and the subsequent process of interpretation give meaning to sensorial data by triggering the activation of neural structures that are partly genetically inherited and are partly continuously redesigned throughout life. In contrast to Hayek, Boulding does not have any specific knowledge of brain functioning.21 His rudimentary ideas on the state of the art in neurophysiology also suggest that in 1956 he was ignoring the content of The Sensory Order and even Hebb’s theory on cell assemblies (see note 14): Indeed, what we know about the brain suggests that is an extraordinarily unspecialised and, in a sense, unstructured object; and that if there is a physical and chemical structure corresponding to the knowledge structure, it must be of a kind which at present we do not understand. (Boulding, 1956, p. 17)

Instead, Boulding points to the principle that organizes knowledge, the image, as being the source of individual specificity. The image works as a psychic substrate whose main function is that of mediating between the body – and, thus, also the brain – and the external environment. The messages that reach the subject are filtered and interpreted through the interaction of two images. The first, ‘an image of fact’, mirrors the world as it is represented in the mind of the subject; the second, an ‘image of values’,22 expresses the value system according to which personal judgements about the world are elaborated. Information that does not pass the scrutiny of personal evaluation, for example is perceived as negative, has a low probability of being elaborated further. Most probably it will be ignored. The subjective nature of knowledge is thus related to the dependence of the images on personal history and experiences. Moreover, subjectivism affects two fundamental elements: the way the information is represented by the subject and the way it is evaluated. Although Boulding and Hayek give prominence to different factors in their comprehension of mental activity, their approaches do not exclude each other. On the contrary, we believe that the acknowledgment of their differences, as much as an analysis of their complementarities, may fruitfully contribute to contemporary research. As mentioned in Section 5.1, the ongoing cross-fertilization between economics and cognitive science has been fostering the development of an approach to knowledge that takes the functioning of the mind and of the brain into account. The research in this field is, however, still quite diversified. Most of all, the neurological and cognitive dimensions are often treated as separate objects of analysis, as the almost independent

134

Handbook of knowledge and economics

development of behavioural economics and neuroeconomics witnesses. Which guidelines would Boulding and Hayek have suggested we follow? In The Sensory Order Hayek explores the neuropsychological mechanisms that are involved in perception and anticipates the idea that neuroscience may help understand how knowledge structures differentiate themselves among economic agents. As also remarked by V. Smith (2003) in his Nobel Prize lecture, ‘Hayek (1952) was a pioneer in developing a theory of perception, which anticipated recent contribution to the neuroscience of perception’ (p. 532). However, Hayek does not claim that the isolated comprehension – or measurement – of brain activity is sufficient when explaining behaviour. First, in his theory, it is the intermingling of neuropsychological and cognitive processes of classification and interpretation that gives form to individual actions. Such processes are context-dependent, as they are moulded by the interaction of the subject and the social environment in which he/ she acts. Second, it is not the study of the brain as such but of the factors that affect its development that is worth consideration. Such factors crucially involve learning, the contact with other agents and the relationship with the socio-institutional context of reference. As underlined by Butos and Koppl (2006), ‘Hayek’s theory does not cast individuals as cognitive islands who can step outside themselves epistemologically or divorce themselves cognitively from the wider social reality they not only inhabit but to which they must also adapt’ (p. 30). As far as Boulding is concerned, the fundamental idea of his book is that behaviour depends on images, which is to say, on mental representations. The attribution of meanings and values to ‘raw data’ takes place at the representational level and may be partially shared among individuals. It is, thus, the comprehension of cognitive mechanisms and of their social functions, much more than an analysis of their neural correlates, that helps in understanding how knowledge is developed. In summing up, we believe that careful reading of The Image and of The Sensory Order would not support the idea that the neurobiological and the psychical dimension of behaviour can be studied in isolation from the social context that influences and structures them. Moreover, an integrated perspective on Hayek’s and Boulding’s contributions indicates that the critical questions for contemporary research address the interaction of cognitive, affective and neurological structures much more than their single roles. 6.5.3

Bounded Learning

In general terms, learning may be defined as a relatively permanent change in cognition, resulting from experience and directly influencing behaviour.

The unexplored affinity between Boulding and Hayek

135

For both Boulding and Hayek the interaction between learning and knowledge appears to be a relevant phenomenon. In The Sensory Order learning plays an active role in modifying the neural structure of the subject and, thus, his/her perceptive capacity. More specifically, the brain evolves under the combined influence of genetic factors and learning during which new connections among previously unlinked neurons are created (synapses). Then, when learning occurs, the perceptual channels through which information is selected and interpreted overcome a phase of reorganization and enrichment. Different perceptual experiences become possible and, consequently, the cognitive abilities of the subject amplify. Analogously, learning is the key factor that helps the images to evolve over time. Under the influence of new experiences and new interactions with the social context, individual worldviews can change. As suggested elsewhere (Patalano, 2007a), the process of change described by Boulding (1956) may be interpreted as a reorganization of the ‘picture’ that the image offers the mind. New elements are included or cut off from the picture and the relationships among the various representational elements undergo a process of revision. The main observable effect concerns behaviour that modifies according to the changes that have occurred at the imaginative level. For both the authors, learning achieves a compromise between the legacy of past experiences and the dimension of novelty. By moulding the cognitive structures that filter and interpret information, past cognitive experiences influence the construction of knowledge. At the same time, learning occurs when a deviation from past modalities in perceiving and processing information emerges and consolidates. As already mentioned, in The Sensory Order a double link exists between neural structure and perception. The first selects and interprets the perceptual stimuli that reach the subject; however, under the influence of prolonged interaction, the most frequent stimuli may alter the neural structure by leading to the emergence of new synapses. As an implication, the perceptual past influences the present by means of the existing neural connections that, in turn, may be modified by recurrent new stimuli. For Boulding, too, the past is the starting point for learning: new visions of reality, and thus new modalities of behaviour, may emerge only through evolution of the already existing ones. In other words, the process of change in the images does not proceed ex nihilo but unfolds from previous imaginary textures. Interestingly, the relationship with the past is not only managed at the individual level. Realization of the collective past is embedded in social traditions and institutions that are transmitted from one generation to the next. Being part of the social context, they are also

136

Handbook of knowledge and economics

included in the mental representations of individuals who share the same environment. At both the individual and the collective level, the evolution of images has to face intrinsic resistance to grasp new opportunities of development. As images are plastic but at the same time resistant to change, learning takes place only when plasticity prevails. From such a perspective, the past is considered, first of all, as a tie that becomes effective when the conservative support of consolidated worldviews blocks evolution. Stickiness to ‘old ideas’ is, for example, quite a common weak point of organizational structures (Boulding, 1956). The factors that may induce cognitive rigidities, thus locking in the evolution of mental representations, are an important topic of research for cognitive economics. As highlighted by the literature, in problem-solving mental representations tend to be stable even if they do not lead to optimal solutions23 (Egidi, 2002, 2003). Boulding seems to suggest that resistance to change does not emerge only in problematic situations but represents a constitutive feature of mental activity. Images have, in fact, an intrinsic tendency towards stability that has to be overcome in order to promote evolution. The relationship between present and past is also a central topic in the literature on path-dependence that has developed during the last two decades24 (Arthur, 1989; David, 1985, 1994; North, 1994; Patalano, 2007b, 2011; Rizzello, 2004). Within the framework of cognitive economics, The Sensory Order has been considered as an important source of intuition on the existence of non-ergodic processes in the evolution of the human brain and mind (Patalano, 2007b; Rizzello, 2004). Boulding’s contribution may help by adding new insights to this line of research.

6.6 FINAL REMARKS The Sensory Order and The Image were published in a period that was exceptionally prolific for the history of ideas. Unfortunately, for many years the works did not receive much attention. Many reasons may contribute to explain such a reaction of the economic community. The main one was probably highlighted by Boulding (1988) and concerns a sort of wrong timing between the evolution of economics as a science and the innovative potential of the two books. The relationship between mental phenomena and economic behaviour, which both the authors support in their works, was not completely new to economics in the 1950s but maybe it was still too far ahead of its time to be understood in its full implications.

The unexplored affinity between Boulding and Hayek

137

Nowadays, the situation is mature for a wider receptive capacity of interdisciplinary ideas. In the last 15 years the cognitive approach to economics has consolidated and diversified into multiple fields of applications. Decision making, problem solving, institutional analysis and innovation theory have taken the functioning of the human mind/brain into account. Important phenomena such as knowledge, complexity, learning, change and coordination have been investigated from an interdisciplinary perspective based on the contamination of economics with cognitive science. It is important not to interpret these contemporary developments as being detached from past economic ideas. The Sensory Order and The Image had already stressed the relevance of the neuropsychological and psychic underpinnings of economic behaviour half a century ago. While some ideas proposed by Boulding and, mainly, by Hayek have been developed with modern instruments, others are still neglected and are worth rediscovering. From this point of view, the most evident lack concerns imagination. The mental attitude of framing situations and developing images of what may happen in the future contributes to guide economic choices and should be analysed in more depth so as to clarify its influence on individual and collective behaviour. As argued by March, ‘imagination of the future, like imagination of the past, are devices for living in the present’ (1995, p. 427).

NOTES 1.

2.

3.

As far as Hayek is considered, the heterogeneity of agents is explained in terms of differences in agents’ preferences and, most important to our purposes, in terms of differences in their perceptions (Arena, 2004). As we shall argue, it is, thus, in the subjectivity of the perceptive processes that the subjective dimension of knowledge is acquired. In addition to that, for Hayek not all knowledge is conscious. Part of the individual knowledge being tacit and difficult to articulate explicitly (Hayek, 1967), institutions such as the market and its price structures act as signals of cognitive contents that could not be communicated otherwise. See Horwitz (2000) and Patalano (2007a) for a fuller analysis of this point. Kenneth Boulding (1910–93) was born in Liverpool, England. He started at New College, Oxford on a chemistry scholarship, and after one year he transferred into the school of politics, philosophy and economics. At first, he worked within the framework of orthodox economics, but by 1948 he had started his dissociation from mainstream theories by integrating economic analysis with biology and with the study of complex systems. During his interdisciplinary research activity, ‘Boulding was awarded honorary doctorates by over thirty universities; he had prizes not only for economics but also for political science, peace research, and scholarships in the humanities’ (Keyfitz, 1994). Four times nominated for the Nobel Prize for Peace, he was also an appreciated Quaker poet. For a complete bibliography see Boulding’s home page at CSF, Colorado: http:// csf.colorado.edu/boulding. ‘I dictated virtually the entire text of The Image, a very intense experience, reflected in the book’s format. There are virtually no footnotes and no bibliography. I simply

138

4.

5.

6. 7.

8.

9. 10. 11. 12. 13.

Handbook of knowledge and economics transformed my own image – part of the content of my own mind – into the English language’ (Boulding, 1988, p. 20). ‘Between the incoming and the outgoing messages lies the great intervening variable of the image. The outgoing messages are the result of the image, not the result of the incoming messages. The incoming messages only modify the outgoing messages as they succed in modifying the image’ (Boulding, 1956, p. 28). For Boulding, the image has an innate tendency to resist changes. Indeed, the idea of rigidities and frictions that do not allow cognitive structures to change continuously is still valid and has relevant implications both at the individual and at the social level (Patalano, 2011). However, Boulding’s vision of communication processes appears to be quite naïve if compared with recent achievements in the field of social psychology and communication studies. See, for example, Sperber (1995), Sperber and Wilson (2002). See Patalano (2011) for a discussion of resistance to change in contemporary literature. ‘If a totally new image is to come into being, however, there must be sensitivity to internal messages, the image itself must be sensitive to change, must be unstable, and it must include a value image which places high value on trials, experiments, and the trying of new things’ (Boulding, 1956, p. 94). Boulding does not explain clearly how an image can be formed ‘at the collective level’. His suggestions on this point are twofold. First, individuals build up images of society and of the world in which they live. Second, through socialization and market exchanges individual imagery is put in contact with that of others. At this level, it may be partially shared. For contemporary reinterpretations of the process through which imagination may be ‘socialized’, see Castoriadis (1975) and Patalano (2007a). March (1995), Penrose (1959) and Shackle (1972, 1979) have also acknowledged such influence; for critical discussions see Augier and Kreiner (2000) and Patalano (2005b). From a political perspective, Hayek also considers the protection of minorities and small groups as being important for the relationship between individuals and their institutional environment. See Hayek (1967). ‘Although the “objective” productive opportunity of a firm is limited by what the firm is able to accomplish, the “subjective” productive opportunity is a question of what it thinks it can accomplish’ (Penrose, 1959, p. 41). More recently, Witt (2000) has underlined the existence of a link between imagination and the ‘business conception’ of the entrepreneur. In the mid-1980s, psychologists Beach and Mitchell (1987) developed a descriptive theory of economic behaviour that attributes a very similar evaluative role to imagination. In their approach, when faced with a decision-making problem, the decision maker refers to four images: 1. 2. 3. 4.

The self-image, which consists of personal values, beliefs and ethics that guide one’s adoption (or rejection) of goals and contribute to the generation of potential new ones. The trajectory image, which is a mind picture of one’s desirable future, given the self-image. The action image, which consists of plans (intended as sequences of actions) currently used by the subject to achieve his goals. The projected image, which mirrors the expected future (in terms of the anticipated results of plans) if one persists with plans that are currently adopted.

A decision consists in adopting or rejecting new candidates as constitutive elements of one’s images (‘adoption decision’), or of evaluating whether progress towards goals has been made or not (‘progress decision’). Decisions are made according to two evaluative criteria, one based on compatibility and the other on profitability. Both these criteria are implemented through computational tests (Beach, 1998). As a result of the compatibility test, a candidate is rejected for adoption if the number of its violations of the

The unexplored affinity between Boulding and Hayek

14.

15.

16.

17.

18.

139

relevant image constituents exceeds a subjective rejection threshold. In other words, it is rejected if its compatibility turns out to be too low in the personal evaluation of the decision maker. Among the compatible candidates, the one who maximizes the probability of reaching the desirable goals is selected: it is in fact the potentially most profitable candidate. For further readings on this theory, see Beach and Mitchell (1987), and Beach (1998). For an extension of this theory to economics see Patalano (2005a, 2005b). Boulding’s approach to the interaction between imagination and society shows an interesting affinity with the process of ‘coordination among expectations’ postulated by Schelling (1960) and may offer important contributions to the research on coordination failures (Patalano, 2005b). Such questions remind us of an inflamed debate in the history of the philosophy of the mind, which regards the nature of mental images. The main controversy is between ‘pictorialism’ and ‘propositionalism’. According to the first line of thought, mental images have a visual–spatial nature that differentiates them from language. As a consequence, they carry a figural content that cannot be verbalized, nor transmitted through more abstracts forms of representation (Kosslyn, 1980, 1994; Pinker, 1984). On the contrary, propositionalists claim that images do not have any symbolic specificity and can be substituted by propositions or verbal descriptions (Pylyshyn, 1973, 1981). For further readings on such debate see Ferretti (1998) and Farah (1988). Patalano (2005a, 2005b) has proposed interpreting the ‘image’ as the final product of vision. Such an interpretation is based on the theory developed by Kosslyn (1980, 1994), according to which the production of mental images is similar to drawing a figure on a computer screen: the sensory information on the physical world, perceived the first time and/or stored in previous memories, is recalled and reassembled, like pixels. Mental images are thus generated on a real interior screen with ‘grain’ and more or less bright areas, called visual buffers. At the same time, vision appears to be a complex system with a vertical and a horizontal dimension (Ferretti, 1998). At the level of low vision the perception of objects reaches the retina. Subsequently, the data imprinted on the retina undergo further cognitive elaboration through which the image is generated. In this approach imagination and vision share similar information-processing procedures and the visual properties of mental images (creating their figural character) are developed by processing sensory data. For further readings on the topic see Farah (1988), and Kosslyn (1994). In The Organization of Behavior: A Neuropsychological Theory (1949) Donald O. Hebb (1904–85) claimed that ‘the problem of understanding behavior is the problem of understanding the total action of the nervous system, and vice versa’ (ibid., p. xiv). According to his approach, neurons that fire together tend to form a cell-assembly whose activity can also persist after the triggering event. In particular, such a cell-assembly will represent the triggering event within the neural structure. Thinking consists, thus, of the sequential activation of sets of cell-assemblies, e.g. neural representations of past events. In his preface of 1952, Hayek mentions the work by Hebb, regretting not to have known about it before: ‘it seems as if the problems discussed here were coming back into favour and some recent contributions have come to my knowledge too late to make full use of them. This applies particularly to Professor Donald Hebb’s Organization of Behavior, which appeared when the final version of the present book was practically finished. That work contains a theory of sensation which in many respects is similar to the one expounded here.’ During a symposium held at the MIT, Chomsky gave a preliminary exposition of his theory on transformational generative grammar, Miller (1956) presented his paper on the magic number seven in short-term memory, Newell and Simon gave a talk on their pioneer computational model (Newell et al., 1958), and IBM researchers used the largest computer available at those times to test Hebb’s psychological theory. In 1956, too, another important conference on artificial intelligence was organized at Dartmouth and attended by Chomsky, McCarthy, Minsky, Newell, and Simon and Miller, among others (Miller, 2003).

140 19. 20.

21.

22.

23.

24.

Handbook of knowledge and economics Marshall, Veblen and Menger can be considered as important precursors of the cognitive approach to economics that emerged after the cognitive revolution. For a historical overview see Egidi and Rizzello (2004). ‘Behavior then consists of acting in a way that is expected to realize the image of the most preferred future. This is a very different view of human behavior from that of the behaviorist, who thinks that behavior follows a stimulus. The idea behind The Image is that human behavior is the result of all previous stimuli, both internal and external, resulting in conscious images of the future. It seems the height of absurdity to an economist to suppose that behavior only comes out of immediate stimuli, even though these are certainly part of the total picture’ (Boulding, 1988, p. 20). ‘In summation, then, my theory might well be called an organic theory of knowledge. Its most fundamental proposition is that knowledge is what somebody or something knows, and that without a knower, knowledge is an absurdity. Moreover, I argue that the growth of knowledge is the growth of an “organic” structure. I am not suggesting here that knowledge is simply an arrangement of neural circuits or brain cells, or something of that kind. On the question of the relation between the physical and chemical structure of an organism and its knowledge structure, I am quite prepared to be agnostic’ (Boulding, 1956, p. 17). ‘The subjective knowledge structure or image of any individual or organization consists not only of images of “fact” but also images of “value” . . . The image of value is concerned with the rating of the various parts of our image of the world, according to some scale of betterness or worseness. We, all of us, possess one or more of these scales’ (Boulding, 1956, p. 11). Their resistance to change takes two main forms. The first is resistance within a given representation. Individuals may prefer not to re-think their mental representation of the current problem to avoid the mental effort involved in re-framing (Egidi, 2002). The second consists of the tendency to transfer strategies from one problem to another, even though they have proved not to be efficient. At an individual level, path-dependence emerges when history influences the choice set and the behavioural algorithms of agents irreversibly (Bassanini and Dosi, 1999). At a macro level, conventions and collectively shared norms, such as those that influence and shape institutions, are also an important ‘carrier of history’ (David, 1994). They give rise to a cumulative and self-enforcing process of development because, by structuring the social context, they also shape the cognitive and behavioural patterns that support their existence.

REFERENCES Arena, R. (2004), ‘Beliefs, knowledge and equilibrium: a different perspective on Hayek’, in S. Rizzello (ed.), Cognitive Developments in Economics, London: Routledge, pp. 316–38. Arthur, W.B. (1989), ‘Competing technologies, increasing returns and lock-in by historical events’, Economic Journal, 99, 116–31. Augier, M. and Kreiner, K. (2000), ‘Rationality, imagination and intelligence: some boundaries in human decision-making’, Industrial and Corporate Change, 9 (4), 659–81. Bassanini, A.P. and Dosi, G. (1999), ‘When and how chance and human will can twist the arms of Clio’, Laboratory of Economics and Management Sant’Anna School of Advanced Studies, Working Paper Series 1999/05. Beach, L.R. (1998), Image Theory: Theoretical and Empirical Foundations, Mahwah, NJ: Lawrence Erlbaum Associates. Beach, L.R. and Mitchell, T.R. (1987), ‘Image theory: principles, goals and plans in decision making’, Acta Psychologica, 66, 201–20. Birner, J. (1999), ‘The surprising place of cognitive psychology in the work of F.A. von Hayek’, History of Economic Ideas, 7 (1–2), 43–84.

The unexplored affinity between Boulding and Hayek

141

Boulding, K.E. (1956), The Image: Knowledge in Life and Society, Ann Arbor, MI: University of Michigan Press. Boulding, K.E. (1988), This Week’s Citation Classic, 48, 28 November. Butos, W.N. and Koppl, R. (1993), ‘Hayekian expectations: theory and empirical applications’, Constitutional Political Economy, 4 (3), 303–29. Butos, W.N. and Koppl, R. (2006), ‘Does The Sensory Order have a useful economic future?’, in E. Krecke and K. Krecke (eds), ‘Cognition and Economics’, special issue of Advances in Austrian Economics, 9, 19–50. Butos, W.N. and McQuade, J.T. (2002), ‘Mind, market, and institutions: the knowledge problem in Hayek’s thought’, in J. Birner, P. Garrouste and T. Aimar (eds), F.A. Hayek as a Political Economist, London and New York: Routledge, pp. 113–33. Caldwell, B. (2004), ‘Some reflections on F.A. Hayek’s The Sensory Order’, Journal of Bioeconomics, 6, 239–54. Castoriadis, C. (1975), L’Institution imaginaire de la société, Paris: Éditions du Seuil; English version The Imaginary Institution of Society (1987), Cambridge, MA: MIT Press and Cambridge Polity Press. Damasio, A.R. (1994), Descartes’ Error: Emotion, Reason, and the Human Brain, New York: G.P. Putnam’s Sons. David, P. (1985), ‘Clio and the economics of QWERTY’, American Economic Review, 75 (2), 332–7. David, P. (1994), ‘Why are institutions the “carriers of history”? Path-dependence and the evolution of conventions, organizations and institutions’, Structural Change and Economic Dynamics, 5 (2), 205–20. Dempsey, G.T. (1996), ‘Hayek’s terra incognita of the mind’, Southern Journal of Philosophy, XXXIV (1), 13–41. Dewey, J. (1929), Experience and Nature, New York: Norton and Co. Egidi, M. (2002), ‘Biases in organizational behaviour’, in M. Augier and J.J. March (eds), The Economics of Choice, Change and Organization: Essays in Memory of Richard M. Cyert, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 190–242. Egidi, M. (2003), ‘Razionalità limitata’, Sistemi Intelligenti, XV, 67–72. Egidi, M. and Rizzello, S. (2004), ‘Cognitive economics: foundations and historical evolution’, in M. Egidi and S. Rizzello (eds), Cognitive Economics, 2 vols in the series The International Library of Critical Writings in Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, vol. 1, pp. 1–22. Farah, M.J. (1988), ‘Is visual imagery really visual? Overlooked evidence from neuropsychology’, Psychological Review, 95 (3), 307–17. Ferretti, F. (1998), Pensare vedendo. Le immagini mentali nella scienza cognitiva, Roma: Carocci Editore. Gardner, H. (1985), The Mind’s New Science: A History of the Cognitive Revolution, New York: Basic Books. Garrouste, P. (1999), ‘Is the Hayekian evolutionism coherent?’, History of Economic Ideas, 7 (1–2), 85–103. Hayek, F.A. (1952), The Sensory Order: An Inquiry into the Foundations of Theoretical Psychology, London: Routledge and Kegan Paul. Hayek, F.A. (1967), Rules, Perception, and Intelligibility: Studies in Politics, Philosophy, and Economics, Chicago, IL: University of Chicago Press. Hebb, D.O. (1949), The Organization of Behavior: A Neuropsychological Theory, New York: John Wiley. Horwitz, S. (2000), ‘From The Sensory Order to the liberal order: Hayek’s non-rationalist liberalism’, Review of Austrian Economics, 13, 23–40. Keyfitz, N. (1994), ‘Kenneth Ewart Boulding, 1910–1993’, Biographical Memoir, http:// newton.nap.edu/html/biomems/kboulding.html. Kosslyn, S.M. (1980), Image and Mind, Cambridge, MA: Harvard University Press. Kosslyn, S.M. (1994), Image and Brain. The Resolution of the Imagery Debate, Cambridge, MA: MIT Press.

142

Handbook of knowledge and economics

Loasby, B.J. (1999), Knowledge, Institutions and Evolution in Economics, London: Routledge. March, J.G. (1995), ‘The future, disposable organization and the rigidities of imagination’, Organization, 2 (3–4), 427–40. Marshall, A. (1890/1961), Principles of Economics, London: Macmillan. Miller, G.A. (2003), ‘The cognitive revolution: a historical perspective’, Trends in Cognitive Sciences, 7 (3), 141–4. Newell, A. and Simon, H.A. (1972), Human Problem Solving, Englewood Cliffs, NJ: Prentice-Hall. Newell, A., Shaw, J.C. and Simon, H.A. (1958), ‘Chess-playing programs and the problem of complexity’, IBM Journal of Research and Development, 2, 320–35. North, D.C. (1994), ‘Economic performance through time’, American Economic Review, 84 (3), 359–68. Paller, K.A. (2001), ‘Neurocognitive foundations of human memory’, in D.L. Medin (ed.), The Psychology of Learning and Motivation, vol. 40, New York: Academic Press, pp. 121–45. Patalano, R. (2005a), Images and Economic Behaviour, Advances in Cognitive Economics, Sofia: New Bulgarian University Press. Patalano, R. (2005b), La mente economica. Immagini e comportamenti di mercato, Roma– Bari: Laterza. Patalano, R. (2007a), ‘Imagination and society. The affective side of institutions’, Constitutional Political Economy, 18 (4), 223–41. Patalano, R. (2007b), ‘Mind-dependence. The past in the grip of the present’, Journal of Bioeconomics, 9 (2), 85–107. Patalano, R. (2010), ‘Imagination and economics at the crossroad. Materials for a dialogue’, History of Economic Ideas, XVIII (1), 167–89. Patalano, R. (2011), ‘Resistance to change. Historical excursus and contemporary interpretations’, Review of Political Economy, 23 (2), 249–66. Patalano, R. and Rizzello, S. (2003), ‘Il concetto di image nel pensiero di Kenneth Boulding e le implicazioni per la teoria economica contemporanea’, Storia del pensiero economico, I (45), 3–32. Peirce, C.S. (1905), ‘What pragmatism is’, in C.S. Peirce (ed.), The Essential Peirce, vol. 2, Bloomington, IN: Indiana University Press, pp. 331–45. Penrose, E.T. (1959), The Theory of the Growth of the Firm, Oxford: Blackwell. Pinker, S. (1984), ‘Visual cognition: an introduction’, Cognition, 18, 1–63. Pylyshyn, Z.W. (1973), ‘What the mind’s eye tells the mind’s brain: a critique of mental imagery’, Psychological Bulletin, 80, 1–24. Pylyshyn, Z.W. (1981), ‘The imagery debate. Analog media versus tacit knowledge’, Psychological Review, 88, 1–25. Rizzello, S. (1999a), The Economics of the Mind, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Rizzello, S. (1999b), ‘The endogenous asymmetrical information: the market for “lemons” according to Hayek’s legacy’, History of Economic Ideas, VII, 13–41. Rizzello, S. (2004), ‘Knowledge as path-dependence process’, Journal of Bioeconomics, 6 (3), 255–74. Schelling, T. (1960), The Strategy of Conflict, Cambridge, MA: Harvard University Press. Shackle, G.L.S. (1972), Epistemics and Economics, Cambridge: Cambridge University Press. Shackle, G.L.S. (1979), ‘Imagination, formalism and choice’, in M. Rizzo (ed.), Time, Uncertainty and Disequilibrium, Lexington, KY: Lexington Books, pp. 19–31. Simon, H.A. (1955), ‘A behavioral model of rational choice’, Quarterly Journal of Economics, 69 (1), 99–118. Simon, H.A. (2000), ‘Bounded rationality in social science: today and tomorrow’, Mind and Society, 1 (1), 25–40. Smith, B. (1997), ‘The connectionist mind: a study of Hayekian psychology’, in S.F. Frowen (ed.), Hayek: Economist and Social Philosopher, London: Macmillan, pp. 9–29.

The unexplored affinity between Boulding and Hayek

143

Smith, L.V. (2003), ‘Constructivist and ecological rationality in economics’, Nobel Prize Lecture, at http://www.wcfia.harvard.edu/sites/default/files/Smith2003.pdf. Sperber, D. (1995), ‘How do we communicate?’, in John Brockman and Katinka Matson (eds), How Things Are: A Science Toolkit for the Mind, New York: Morrow, pp. 191–9. Sperber, D. and Wilson, D. (2002), ‘Truthfulness and relevance’, Mind, 111, 583–632. Witt, U. (2000), ‘Changing cognitive frames. Changing organizational forms. An entrepreneurial theory of organizational development’, Industrial and Corporate Change, 9 (4), 733–55.

7

The knowledge–rationality connection in Herbert Simon Salvatore Rizzello and Anna Spada

7.1 INTRODUCTION The central role of knowledge for economics was already manifest during the 1930s, but it was during the 1950s that it became especially evident. The 1950s are mainly characterized by a confrontation between the economists defending rational choice theory and those who intend to develop new tools, which will constitute the grounds for a psychologybased theory of decision. The confrontation revolves around the issue of uncertainty in decision making, which also emerged during the 1950s as an unintended consequence of expected utility theory (von Neumann and Morgenstern, 1944). Those economists who were trying to keep rational choice theory alive pursued their goal at two different levels: first, at the analytical level, with the attempt to make the neoclassical assumptions of rationality explicit, in order to reduce the psychological distortions which generated deviations from the expected utility model (Savage, 1954). The attempt was also pursued at the epistemological level, with the evolutionary foundation of rationality (Friedman, 1953). The major goal of this defence of rational choice, at both levels, is to free economic models from uncertainty. During the same period, however, there were several contributions that took another direction and tried to elaborate a descriptive theory of decisionmaking, including uncertainty as one of its unavoidable characteristics. During the 1950s, therefore, uncertainty was the central problem for economists, and the relevance of knowledge was mainly related to the role that this concept could play as an instrument to face uncertainty in decision making. We can find this concept of knowledge in all the contributions that laid the foundations for a psychology-based decision theory. Nevertheless, in the debate on uncertainty, knowledge remains essentially tacit and underdeveloped, even though Hayek had already highlighted the neurobiological foundations of knowledge and its acquisition and use with his The Sensory Order (1952). The implicit use of the concept of knowledge also characterizes Herbert Simon’s contribution, which appears as fundamental for the 1950s debate. 144

The knowledge–rationality connection in Herbert Simon 145 The relevance of Simon’s contribution is broadly acknowledged, both for the emergence of a psychology-based decision theory and for the changes in economics’ conception of rationality. It is precisely during the 1950s that he explored the nature of human rationality and the related problem of its role in the acquisition and use of knowledge, which he considered as an instrument to face decision-making uncertainty. He also played a central role in the renewal of the conception of economics, both as father of cognitive sciences and as leading figure of the CarnegieMellon School. Therefore, by means of a profound reflection on economics, psychology and epistemology, he brought a relevant contribution to the making of the theoretical arena that characterized the 1950s and that represented the basis for the emergence of the cognitive approach. In particular, his considerations concerned the concept of rationality. It is well known that Simon’s research exposed the divergences between the concept of rationality used in economic models and the one that characterizes actual decision making. This divergence, confirmed during the same period by experimental works (Allais, 1953), leads Simon to separate economic rationality, which he defines as absolute rationality, and human rationality, which he defines as bounded and procedural rationality. It is also well known that Simon, with the expression ‘absolute rationality’, or ‘economic rationality’, refers to the concept of rationality used in economic models, where decision-making is assumed to be characterized by risk and not by uncertainty, following Knight’s (1921) distinction, and the decision maker is assumed to be able to acquire information and not knowledge, following Hayek’s (1937) distinction. On the other hand, Simon defines as bounded and procedural rationality, which we can call ‘human rationality’, the kind of rationality that actually features in real decision-making, which is characterized by uncertainty and where the decision maker is assumed to be capable of acquiring and using knowledge. The strong contrast that Simon draws between the nature of economic rationality and the nature of human rationality blurs another relevant distinction: that between bounded and procedural rationalities. This difference seems instead to be the place where it is possible to find a large part of Simon’s new and topical contribution. His references to knowledge mainly regard the procedural dimension, which has less diffusion than the bounded one. Consequently, Simon’s considerations on knowledge come out only later (Rizzello, 1997). An attempt to underline the differences between bounded and procedural rationality seems useful in order to clarify the presence, even if implicit, of a Hayekian idea of knowledge in Simon’s work. Similarly, a way to understand this difference – and specifically the relevance that Simon attributes to the procedural dimension of decision making – seems to be a deep analysis of the role of knowledge in Simon.

146

Handbook of knowledge and economics

The first section of this chapter is devoted to analyzing the theoretical debate that during the 1950s caused the decline of rational choice theory and the rise of psychology-based decision theory. This analysis revolves around the concept of uncertainty, which is strictly related to another concept (knowledge) since the time when their role in economics was discovered during the 1920s and 1930s. In particular, it shows that during the 1950s uncertainty becomes the central problem faced by economists (Section 7.2.1), who propose different kinds of solution: on the one hand, some economists try to strengthen rational choice theory, and, on the other, some support a psychology-based decision theory, setting the basis for a cognitive approach (Section 7.2.2). It is shown how the problem related to uncertainty and the attempts to find its solution spawn another problem: the relationship between the nature and the capabilities of human rationality and those assumed in economic models. The problem of the nature of human rationality requires a comparison between the perspective that tries to support rational choice theory and the one that tries to improve a psychology-based approach. For the latter perspective, we focus on Simon’s contribution, trying to clarify the relations among absolute, bounded and procedural rationality (Section 7.2.3). We then turn to the role of knowledge during the 1950s, and discuss the different ways to conceive of it in the two perspectives analyzed (Section 7.2.4). Elaborating on the theoretical dimension defined in the second section, the third section explores the relation between Simon’s concepts of bounded and procedural rationality. This analysis, underpinned by a comparison of the differences but also the similarities in the positions  of  Herbert Simon and Milton Friedman (Section 7.3.1), clarifies on one hand the agreement between the concept of bounded rationality and the economic models based on risk and information (Section 7.3.2), and on the other it explains the central role played by the procedural dimension in Simon’s thought, in the evolution of the notions of knowledge and uncertainty and in the conception of economics (Section 7.3.3). The fourth section considers some confirmations of the procedural dimension of Simon’s work in the application of the concept of rationality to interaction situations, by means of a comparison between Simon’s idea of game theory and prospect theory. Finally, we highlight the reasons why the relationship between uncertainty and knowledge becomes decisive precisely when the goal is to compare the features of human rationality and those of economic rationality, and focus on the role of knowledge in Simon’s work and discuss its relevance.

The knowledge–rationality connection in Herbert Simon 147

7.2 KNOWLEDGE, UNCERTAINTY AND RATIONALITY: THE DEBATE OF THE 1950S Knowledge and uncertainty are fundamental concepts for the emergence of a psychology-based decision theory as an alternative to rational choice theory and, in general, for the resulting diversification of the perspectives on the idea of economics. Psychology-based decision theory emerges during the 1950s, as a result of several contributions that give life to a debate focused on knowledge, uncertainty and rationality. In particular, the central issue in that period is the uncertainty that characterizes decision making. The solutions to this problem provided by economists led some of them to investigate another problem related to uncertainty: the nature of human rationality and the ways in which it copes with uncertainty. Among the contributions that tried to study it, Herbert Simon’s work seems to be one of the most interesting: he distinguished between economic and human rationality. Knowledge and uncertainty seem to be strictly related since the acknowledgment of their relevance for economics, during the 1920s and the 1930s, but during the 1950s’ debate this link appears richer, also as a result of the advances in psychology. That debate holds the roots of cognitive-based analysis of decision making, as afterwards pointed out by means of the complementarity between the work of Simon and Hayek (Rizzello, 1997). Moreover, another element that reinforces the link between knowledge and uncertainty during the 1950s is the relationship both have with the notion of rationality. Considering the differences between economic and human rationality, and specifically between bounded and procedural rationality, the concept of knowledge can enlighten the importance of the procedural nature of human rationality and of decision making. 7.2.1

Knowledge and Uncertainty

Even if knowledge and uncertainty became central in decision theory during the 1950s, as a result of the advancements in psychology that helped decision analysis, their relevance for economics had already emerged earlier. Indeed, the relevance of uncertainty was pointed out during the 1920s by Knight (1921), who distinguished it from risk, underlining that it is more widespread than risk in the real world. He also noted how uncertainty is often analyzed in economic models with instruments meant to analyze risk. The importance of knowledge for economics emerged with Hayek’s work. In ‘Economics and knowledge’ (1937), he emphasizes the central role played by the processes of formation and use of knowledge to understand

148

Handbook of knowledge and economics

economic performance and introduces the important distinction between information and knowledge. According to Hayek, economic subjects’ knowledge is not enough to explain the functioning of markets. In fact, it depends on personal interpretation that each subject realizes on the basis of information through his personal interpretation. Information is made up of external data and is objective, whereas knowledge is the product of a reprocessing work operated on external data through cognitive patterns, so it is subjective and idiosyncratic. In order to explain economic performance, what matters is not how much information an economic subject can collect, but the knowledge he is able to build on that information. It is a result of the complex interaction of external information, mental structures and their modifications as a consequence of past information. Hayek builds his analysis of knowledge on a model of mind that he elaborated several years earlier.1 The idiosyncratic nature of knowledge entails a heterogeneity in economic subjects that is not compatible with market equilibrium. But it does not prevent subjects from sharing several norms, which emerge spontaneously as a consequence of their free action and lead towards a spontaneous order. Hayek provided an explanation of this process in ‘Economics and knowledge’, but he explored it more deeply in another essay, ‘The use of knowledge in society’ (1945), in which he highlights the connection that can be established between the structure of human mind and institutions, and underlines that institutions are based on subjects’ shared knowledge, generated through a continuous interaction of subjects with the external environment. Knight’s description of the relations that connect available knowledge and uncertainty supports the idea according to which the latter is a consequence of the partiality of the former. In this sense, Knight anticipates the analysis of Hayekian models of incomplete knowledge (Rizzello, 1997). In particular, Knight’s (1921) and Hayek’s (1937) contributions share precisely the analysis of the relationship between knowledge and uncertainty, but Hayek’s position appears more sophisticated: indeed, he underlines that the partial, imperfect and idiosyncratic nature of knowledge is the origin of uncertainty but also, at the same time, an instrument to cope with it, through the shared dimension of knowledge. The importance of the connection between knowledge and uncertainty for economics, therefore, emerges together with the relevance of both concepts. It will remain central in Hayek’s subsequent works. In general, as we try to illustrate, the relation between knowledge and uncertainty represents a privileged and exhaustive perspective to understand both concepts, to analyze roots, developments and nature of cognitive economics, and to fully appreciate the contributions made by one of its founding fathers: Herbert Simon.

The knowledge–rationality connection in Herbert Simon 149 Another crucial event occurred between the rise of the importance of the concepts of knowledge and uncertainty during the 1920s and 1930s and its diffusion during the 1950s: the introduction of expected utility theory, by von Neumann and Morgenstern (1944), during the 1940s.2 It explains economic choices through a formalized model based on a description of probabilistic calculations that allow economic subjects to associate a utility level with a possible future event. Even if expected utility tried to analyze the behavior of economic decision makers in a context characterized by uncertainty, in fact it considered contexts of risk, more than 20 years after Knight (1921).3 The work of von Neumann and Morgenstern called economists’ attention to uncertainty, even if it was not the authors’ main purpose. An important characteristic of expected utility theory can be found in its mathematical foundations, which made it possible to make more specific assumptions and draw more pointed implications with respect to the previous rational choice theory. Moreover, it faced increasingly difficult economic questions. Indeed, as expected utility uses became common, it lent credit to the beliefs about the strength of mathematical coherence as a useful instrument with which to face economic complexity. As a consequence, the complication of algorithms also grew and raised the problem of the real nature of economic models’ behavior: could it be considered just normative, as an instrument used by economists to analyze some features of decision making? Or should it be considered equivalent to real human economic behavior, as a positive description? 7.2.2

Uncertainty and Rational Choice

A way to evaluate the correspondence between rational choice in economic models and the decision making of real subjects could rest in a comparison between the capabilities in problem solving that each requires. The formalization of expected utility facilitated this comparison, since it made precise, and thus verifiable, assumptions and implications of rational choice. In particular, it made it possible to analyze the above comparison through some experimental works (Allais, 1953; Ellsberg, 1961; Markowitz, 1952). Experimental evidence showed several anomalies in the description of economic behavior based on expected utility and rational choice theory; moreover, also on the basis of experimental results, many criticisms were directed to the epistemological assumptions of economics and, especially, to the foundations of rational choice. It is the explanation of the contemporary rise and fall of rational choice theory thanks to its broad diffusion, first, and then because of the observation of numerous anomalies (Egidi, 2005).

150

Handbook of knowledge and economics

In particular, Maurice Allais showed how a decision maker, facing a setting characterized by risk, systematically violates the expected utility axiom of independence (Allais, 1953). The anomalies exposed by his work showed that human capacities do not correspond with those assumed by rational choice theory. Allais was mainly oriented toward a critique of the claim of a descriptive value for expected utility theory, in other words, a critique of its ability to provide a realistic representation of economic behavior. This critique focused on the expected utility failure to include the psychological explanation for the distorted perceptions of risk. More specifically, it analyzed the abandonment of the independence assumption and it found support in experimental findings. Its results, known as the Allais paradox, showed that economic subjects often make inconsistent choices and therefore violate the independence assumption, invalidating expected utility theory. Through his work, Allais showed that the problems related to expected utility concerned not only uncertainty but also risk. By means of the individuation and the descriptions of anomalies, he showed the incapacity of rational choice theory to describe the real behavior of economic subjects. He also found the source of this inability in a distorted perception of risk and, in this way, he ascribed to psychological insights a relevant role in decision making. The experiments also showed that these observations cannot be described with the instruments of expected utility, such as mathematical formalization. Therefore he showed that employing the calculus of probability on an ordinal utility function, which characterized expected utility, is not enough to overcome the problems related to the psychological origin of some decisional processes, such as individual wishes and beliefs. These elements show that economists were again facing the same problems that characterized the previous ordinal utility theory and made it clear that expected utility theory was not the right tool to approach them. The problems with expected utility, exposed by the experiments, stimulated several attempts at preserving formalization (Savage, 1954). Nevertheless, the differences between the capacities exhibited by common decision makers and those required by assumptions of rational choice theory induced numerous economists to focus their attention on the nature of human rationality. On one side, we find an attempt to solve this problem, trying to justify the presence of full rationality (or some close approximation) in the real world, explaining its way to work with an evolutionary explanation: this perspective includes Friedman’s contribution (1953), founded on the results of Alchian’s work (1950). This contribution, which will be extensively discussed below, is based on a perspective that tries, in spite of Katona’s work, to move the empirical verification again toward the aggregate level only. On the other side, we see an attempt to

The knowledge–rationality connection in Herbert Simon 151 analyze uncertainty in its own context, decision making, and to consequently study the source of this context, the mental processes, their cognitive limits, including but not limited to uncertainty: this second perspective includes contributions that represent the basis for the cognitive approach and, among them, the most exhaustive answer to the problem of human rationality nature came from Herbert Simon. 7.2.3

The Problem of the Nature of Human Rationality

Along different paths, the relation between human and economic rationality brings us to Herbert Simon. It becomes the central point of his research and, in the attempt to solve the problems connected to this relation, he finds some answers that will serve as founding stones for the rise of cognitive economics. Moreover, the important role he covered both in the rise of cognitive economics and in the foundation of the Carnegie-Mellon School proves the interdisciplinary nature of his approach and his will to analyze human rationality in depth, using whatever tool or approach may be necessary. The birth of the cognitive approach is often associated with psychology more than any other field, but it embraces all the fields that study mental processes, including economics. The origins of cognitive economics are often considered as an import into economics from cognitive psychology. A deeper analysis, however, shows that the cognitive approach is the result of a shared investigation among many relevant fields that tried to solve a particular kind of problem, leading to the same central question: how does the human mind work? As a consequence, many multilateral links among the interested fields emerged. In particular, the link between economics and psychology is characterized by a two-way reading: on one hand, the results from psychology had great relevance for economics’ advances; on the other hand, economics’ contribution to these results was also significant: let us remember that most anomalies and paradoxes were exposed by experiments realized in economics departments. They also stimulated the rise of the advanced psychological theories that will be the ground for cognitive sciences. The role of economics in the transition from behavioral to cognitive psychology4 can be clarified, as well as through experimental results, also with another important element: the way in which the problems related to the nature of human rationality were studied at the Carnegie-Mellon School. During the 1950s, a group of economists spent long periods of time in Pittsburgh: Herbert Simon, Robert Lucas, Franco Modigliani, Richard Cyert and James March, joined by the younger John Muth and Oliver Williamson. Among them, Simon exhibited the most promise. In his graduate student Williamson’s (2002, p. 1) words, ‘of the many distinguished faculty at Carnegie, none cast a

152

Handbook of knowledge and economics

larger shadow than Herbert Simon. Everything seemed to be within his purview.’ At Pittsburgh, rationality was the central theme and its analysis spawned two different ways of studying it (Sent, 1999). On one side, it led to rational expectations theory, by Muth (1961) and Lucas (1981), and to several related works such as Modigliani’s. On the other side, it raised the problem of the bounded nature of human rationality, the new results coming, by means of a comparison, from psychology and the other cognitive sciences. This second perspective, whose champion and herald was Simon, leads him to elaborate the concepts of bounded and procedural rationality, through several works he produced in the field of the theory of organizations together with Cyert and March. In particular, Simon’s works are oriented toward a distinction between the nature of economic rationality, that assumed by economic models and defined by Simon as full rationality, and the nature of human rationality, that exhibited by human decision makers, in turn separated by Simon in bounded rationality and procedural rationality. As commented by Sheffrin (1996), the birth of two perspectives so different within the same school seems to be not only a problem of explaining how this happened, but also of how one failed to prevent the other from emerging.5 Considering that the same intellectual place generated different, often opposite, answers to the human rationality problem, the genuineness of the argument appears clarified. Moreover, it explains the role of economics in giving problems a formulation able to clarify the relevance of a new psychological theory for their solution. 7.2.4

Knowledge in the 1950s

In the 1950s, contextually to the investigations into uncertainty, a series of studies on knowledge and on its relevance to the analysis of economic phenomena emerges. As with Knight’s (1921) and Hayek’s (1952) contributions, we note that the link between knowledge and uncertainty is also present in this debate, when the efforts to cope with the problem of uncertainty and with another connected problem concerning the nature of rationality imply reference to the role of knowledge, though not always explicit. In particular, Simon’s contribution to the role of rationality is characterized by a remarkable complementarity with Hayek’s ideas on knowledge. It is based on a conception of rationality and knowledge both conceived as limits to the achievement of maximising behavior (by following the parameters of the neoclassical theory) but, at the same time, as tools to obtain statisficing results. However, the complementarity between Hayek’s and Simon’s contributions, along the cognitive dimension of Hayek’s thought, was recognized in the literature (Rizzello, 1997) only

The knowledge–rationality connection in Herbert Simon 153 much later, and this relevant part of Hayek’s thought is to date not fully acknowledged in the historical description on the origins of behavioral economics. Accordingly, in the 1950s we have an imbalance in the recognition of the relevance of uncertainty and knowledge in economics, to the detriment of the latter, which lies behind that debate, although it was the focus of Hayek’s thought. By considering its underlying reasons, besides the historical ones, we can detect others with a narrower theoretical nature. In particular, one important reason may explain the delay in acknowledging the complementarity between Hayek and Simon. This reason relies on a more thorough analysis of Simon’s ideas of human rationality, different from economic rationality and defined in subsequent steps as bounded and as procedural. This latter dimension, in particular, is the central element of Simon’s thought in connection with the role of knowledge. However, procedural rationality became less diffused than bounded rationality. The reason why this happened can be found in the trend to identify human rationality and, more generally, Simon’s contribution only with bounded rationality, which is easier to manage when it is taken into account separately from procedural rationality. But, as we shall try to demonstrate, it is precisely in the concept of procedural rationality that lies the most innovative, relevant and modern component of Simon’s thought.

7.3 PROCEDURALITY, KNOWLEDGE AND UNCERTAINTY The complementarities between Simon’s and Hayek’s thought pointed out the procedural dimension that characterized both authors and allowed a better understanding of the role played by knowledge in Simon’s thought. In this way it is possible to pinpoint the differences between the concepts of bounded rationality and procedural rationality. Towards this goal, it seems especially interesting to compare Herbert Simon’s and Milton Friedman’s positions on the relationship between economic and human rationality. This may seem a problematic relationship, from which we can show that they initially shared the same idea of how to address the problem, but also that afterwards they offered very different answers. In addition to the well-known disagreement between them, these answers also point out an interesting comparison between the concept of bounded and procedural rationality. In particular, it reveals the compatibility of the concept of bounded rationality with the models based on risk and on the acquisition of information, and on the other hand the centrality of the procedural dimension in Simon’s thought, in the development of the concepts

154

Handbook of knowledge and economics

of rationality, knowledge and uncertainty and with respect to the conception of economic science. 7.3.1

Friedman on the Nature of Human Rationality

As illustrated above, during the 1950s economists became aware of the problem of uncertainty in economic decisions. The solutions proposed can be divided into two different camps: on the one hand, some lean toward the upholding of the validity of rational choice theory by trying to subsume uncertainty into mathematical formalizations; on the other hand, others try to explore real human cognition by highlighting the related and more general problem of the nature of human rationality. In this dichotomy, Milton Friedman holds a special position. Even though his goal is to defend the theory of rational choice, instead of getting rid of the gap between behavioral predictions and actual behaviors as a mistake, he acknowledges it as a point of departure for his analysis (Friedman, 1953). Indeed, his position, which will later characterize the entire Chicago School, rests on the belief that it is impossible to assign to the common decision-makers those high calculative skills postulated by the theory of rational decisions and, in particular, of its application to complex choices. In this way, Friedman appears closer to Simon’s position than to Savage’s, with the exception of the kind of answer he offers to explain the difference between the capabilities ascribed to decision makers in the theoretical frameworks and the actual rational capabilities they possess. This affinity with Simon in addressing the problem makes Friedman’s answer more sophisticated. Nevertheless the differences between human rationality and economic rationality do not invalidate the efficacy of the theory because behaviors assumed by decision makers, bereft of sophisticated computational tools, accomplish the same outcomes achievable by rational individuals. Friedman works out his position with regard to the nature of rationality, to which he refers by ‘as if ’, on a different side from the analytical one proposed by Savage (1954) with specific reference to the epistemological dimension. This attention to the epistemological side spurs Friedman to put the descriptive capacities of a theory on the back burner and to support the idea that its effectiveness does not rely on the realism of the hypothesis, but depends on the effectiveness of the final results. With a subsequent step Friedman shows how the effectiveness of the results achieved by the neoclassical–deductive theory relies on the fact that, although the assumptions of economic models do not reflect the nature of human rationality, it is anyhow appropriate to take them as starting points, because actual decision makers’ actions bring about the

The knowledge–rationality connection in Herbert Simon 155 same results as would be achieved by means of rational decisions if those assumptions were true. This explains why he succeeds in supporting the neoclassical approach despite his acknowledgment of the gap between economic and actual rationality. In advancing this position, Friedman supplies an account of the mechanisms leading decision makers to their choices that would require complicated calculations, but that they need not perform. Even from this point of view, Friedman’s position seems to call to mind that of Simon when he claims that economic agents can make highly rational choices, despite their bounded abilities. However, even in this case, Friedman and Simon propose very different accounts to explain the same phenomenon. Friedman proposes an evolutionary explanation by affirming that those who make choices assume efficient strategies, which allow them to maximize utility, so that, although these choices are not based on the assumption of perfect rationality, they can be perfectly captured through them. On the other hand, those who make choices that do not tie in with assumptions of economic rationality fail to maximize utility. The evolutionary character of Friedman’s proposition relies on the fact that non-maximizing behaviors are selected against (see Egidi, 2005). In this way the worst-performing agents learn the right behavior and adjust their actions in the direction of the maximization of utility. Behavior does not concern the assumptions of rationality, which remain unknown to the decision maker, but simply the optimizing behavior displayed by those who maximize utility.6 7.3.2

Bounded Rationality, Information and Risk

Bounded rationality, regarded as a theoretical tool and not as a description of human cognitive abilities, allows us to take another step in the comparison between Friedman’s position and Simon’s. As emphasized above, even if he defends the claim of the neoclassical approach, Friedman does not deny the gap between the rationality characterizing economic decision makers in their real actions and that form of rationality that theoretical models require of them. In this way his position appears closer to Simon’s than Savage’s (Section 7.2.2). We have just pointed out that in this period Simon’s answer rests on an exclusively theoretical level, as does Friedman’s, although with a very relevant difference with reference to the empirical verification. This difference, although it refers to a fundamental aspect, claims the compatibility of the assumptions of the limits of rationality with the analytical tool of utility maximization. In Simon’s perspective, indeed, in the models of optimization, the crucial role is played by the auxiliary assumptions of the limits to impose on rationality,

156

Handbook of knowledge and economics

and not by the general assumption about individual rationality (Simon, 1987a, p. 17). Simon carries out an analysis of the hypotheses related to the utility function, trying to demonstrate the compatibility of the models of maximization with altruism. He considers it a misunderstanding that in the utility function only egoistic behavior is considered as a characteristic feature of human nature. The hypothesis of utility maximization by individuals or by organizations simply entails the idea that they are coherent and not necessarily egoist. Consequently, altruism is compatible with the utility function, provided that the latter includes someone else’s well-being in its elements (Simon, 1993). This compatibility can be extended, with some specifications, to the cognitive limits. These specifications concern the fact that, with reference to the cognitive limits, what is taken into account are not the hypotheses concerning the utility function, but the auxiliary assumptions concerning the limits to impose on rationality. Simon illustrates how some accounts of economic phenomena are not absolutely supported by the hypothesis of rationality, because the object under examination is socially determined. However, they can be rationalized by means of some auxiliary hypothesis. These auxiliary assumptions play the crucial role. This means that it will be decisive to establish when these assumptions cease to be valid, or rather when the limits of rationality take over. It becomes evident that Simon acknowledges that traditional theory takes into consideration human cognitive limits, even if it considers them simply ad hoc or casual deviations from the model of expected utility, whereas from his point of view bounded rationality must necessarily come from the detailed and systematic empirical study of human behavior in the process of formation of decisions (Simon, 1987b, p. 30). The principle that determines the rationality of a choice remains valid, that is, the coherence in the decisional process, and it survives in the transition from the substantive neoclassical rationality to Simon’s bounded rationality. Coming back to the comparison with Friedman’s position, it emerges that, at this point, both detect the problem of rationality but they provide two different answers, in diverse plausible ways: the fundamental difference lies in the strong empirical base of Simon’s answer, which is entirely lacking in Friedman’s perspective. Following Simon, probably the classic version of the subjective expected utility theory does not intend to offer an account of the process that human beings follow in taking a decision. It is rather a system to forecast a choice, by conceiving it as the objectively optimal answer to some situation, by affirming that people choose as if they maximized expected utility (Simon, 1987b, p. 28). Until now, however, neither of them is able to invalidate the other.

The knowledge–rationality connection in Herbert Simon 157 7.3.3

Procedural Rationality, Knowledge and Uncertainty

A sudden change occurred when Simon contrasted the theory of procedural rationality to Friedman’s ‘as if ’ theory: both aim at why and how economic decision-makers exhibit a high degree of rationality, although they posses only bounded cognitive abilities. By assuming an evolutionary–adaptive explanation of rationality, Milton Friedman does not offer an appropriate answer, because in this way he clearly separates the actual behavior of decision makers from the analytical tools employed to study it. Following an alternative path, mainly based on studies in the field of psychology, which stressed the relevance of process in making decisions, Herbert Simon works out the concept of procedural rationality (Newell and Simon, 1972). In particular he elaborates some models of problem-solving and decision-making that are coherent with the advances of cognitive psychology by demonstrating how individuals solve problems by formulating heuristics. With regard to the theory of decision, the opposition between the neoclassical idea – portrayed by Friedman – and the cognitive one – portrayed by Simon – becomes complete with the introduction of the concept of procedural rationality because, differently from bounded rationality, it invalidates the efficacy of optimization as a useful tool to evaluate the outcomes of decision processes. In this way, Simon identifies an alternative analytical tool, consistent with the analysis of the procedures that lead to choice. This tool opposes to the optimization process arguments of a functional nature, synthesized in the definition of the ‘satisficing approach’, adopted by Simon since 1956, when he described it by affirming that a decision maker who chooses the best valid alternative is optimizing; a decision maker who chooses an alternative equalling or exceeding specifying criteria, though with no guarantee that it is the only one or the best one, is satisficing his own wishes (Simon, 1987c). To find the level at which the criteria for evaluation are set, Simon borrows from psychology the concept of levels of aspiration.7 By means of feedback they lead the individual to move towards accessible levels of aspiration. The difference between optimizing and satisficing is evident: besides the reduction of complexity of required computations with respect to optimization, the satisficing approach, as described by Simon, provides an explanation of the procedures leading to the elaboration of a final decision, that is, the choice. Moreover, it stresses the crucial role played by the decisional process and its relevance in the evaluation of the results accomplished by means of adaptation to the level of aspirations. Following Simon, the relationship between bounded and procedural rationality lies in the fact that bounded rationality refers to the many

158

Handbook of knowledge and economics

cognitive limits that prevent economic agents from behaving in a way that approximates the conjectures of neoclassical theory (including uncertainty), whereas procedural rationality refers to the description of the dynamics underlying decisional processes that lead to choice. This description includes of course cognitive limitations, that is, bounded rationality, which have a crucial role within decisional dynamics. Therefore Simon starts from the elaboration of a conceptual tool, bounded rationality, which can in a way be considered interchangeable with subjective expected utility theory. This interchangeability derives from the fact that Simon at first conceives of bounded rationality not as ‘descriptive’ of the actual agents’ behavior, but as a possible tool: one could therefore reckon subjective expected utility and bounded rationality as alternative conceptual tools, according to whether the goal is to evaluate the outcome of some choices or to analyze the side of decisional processes, disregarding choice outcomes. However, Simon later elaborates (1957b) and then makes clear (1976) that bounded rationality, as he conceived of it, is not only an analytical tool but also and, above all, the actual description of the limitations manifested by economic agents. This specification is improved by the elaboration of the concept of procedural rationality, which describes, in this instance and not only theoretically but also effectively, the role of limited human abilities, exploited in the decisional process and in the achievement of the final outcome. The consistency with Simon’s entire theoretic path is nevertheless present since the elaboration of the concept of bounded rationality, which differs from subjective expected utility theory by reason of the presence of empirical foundations, completely absent in the latter. Empirical foundations still survive in procedural rationality and in the satisficing approach.

7.4 PROCEDURALITY, EMPIRICAL FOUNDATIONS AND INTERACTION SITUATIONS Bounded rationality and procedural rationality, in spite of their similarities, also present a major difference: both are conceptual and descriptive instruments, but bounded rationality can also be considered as a just conceptual one. The latter is, for example, Friedman’s interpretation (see Section 7.3.1). We know that Simon regards bounded rationality as a description of the actual limitations of human cognitive skills: not only does he mention this explicitly (Simon, 1976), but he also confirms it indirectly, in that his entire theoretical approach (bounded rationality, procedural rationality and the satisficing approach) is empirically founded. Indeed, if we consider the concept of bounded rationality as empirically grounded, it must have a descriptive nature.

The knowledge–rationality connection in Herbert Simon 159 The importance of empirical foundations for Simon finds yet more support in his idea of the connection between individual decision making and social interactions. In the introduction to the second edition of Administrative Behavior (1957a) Simon compares the behavioral perspective and game theory, in von Neumann and Morgenstern’s formulation:8 according to Simon, they shared an intrinsically critical stance toward accounts of isolated decision-making processes and believed instead in the importance of interaction among economic agents. Most importantly, they also shared a constituent element, the idea that human behavior can be described using a decision-tree, but with a key difference: Simon believes that game theory lacks an empirically grounded description of human behavior. He writes: ‘the theories of human choice and of organization . . . rest on a very different description of rational man than does game theory . . .’ (1957a, p. xxix). Simon maintains that game theory, in von Neumann and Morgenstern’s formulation, is not the best way to understand the interactive dimension of decision-making processes because it lacks empirical foundations. A few years later he will write that a good approach to this goal is that by Kahneman and Tversky, with their works on ‘prospect theory’ (1979, 1992). Simon appreciates Kahneman and Tversky’s contribution for their analysis of the nature and importance of interaction, for their idea of interaction as the moment when economic agents create a mental representation of alternative options and states of the world, and this moment deserves to be singled out in the process of decision making (Simon, 1988, 1989, 1994). Since the basis for an empirically grounded analysis in Simon’s work rests in procedurality, for the sake of our analysis it is helpful to consider its role under each of these perspectives of interactive behavior. In the first, von Neumann and Morgenstern’s game theory, there is no place for a procedural idea of rationality. We have a separation of the different steps of a decision, but in each of them we have a decision based on expected utility and maximization, where bounded rationality may find a place, but not procedural rationality. It is a theoretical perspective similar to Friedman’s (see Sections 7.3.1 and 7.3.2), according to which Simon’s work can be subsumed within neoclassical models, and bounded rationality can be regarded as one of the many limitations of full rationality. Certain later developments in game theory clearly undertook a path for which Simon’s ideas are fundamental (e.g. Aumann, 1997): here bounded rationality is not only theoretically compatible, as in von Neumann and Morgenstern’s and Friedman’s accounts, but explicitly employed to strengthen the neoclassical approach (Sent, 2004, 2005). However, the separation of bounded rationality from procedural rationality may also

160

Handbook of knowledge and economics

compromise the empirical grounding. In this way bounded rationality remains a purely conceptual tool, compatible with neoclassical economics. The closing of the gap between neoclassicals and Simon, which Simon himself believed was impossible, however, would weaken the novelty of many of Simon’s ideas. In the second perspective, if we consider the literature on Kahneman and Tversky’s contribution, we find some difficulties in defining the role of procedurality. Kahneman and Tversky’s approach explores the limitations of expected utility theory, specifically by exposing human failures in computing probability-weighted utility.9 They empirically show systematic violations of the principle of invariance. The reason why the role of procedurality is difficult to capture in Kahneman and Tversky may be that their work has been interpreted by economists as a way to include some results of behavioral economics in the theoretical models of the mainstream. Such interpretation places Simon’s research at the highest degree of contrast with the mainstream, while Kahneman and Tversky’s is a more reasonable reinterpretation that takes into account economics’ need to quantify.10 While it is true that they refer to Simon’s bounded rationality and that he acknowledges their contributions, we have already seen how bounded rationality may find a place in the neoclassical method (Section 7.3.2). With this interpretation, however, Kahneman and Tversky’s research becomes a mere extension of the interchangeability of bounded and full rationality or satisficing and optimization, in the direction of more formalization. There is no room for procedurality. Kahneman and Tversky’s work, however, relates to Simon’s for more than the possibility to formalize bounded rationality. They share the distinction between mental representation and evaluation (or decision); hence the idea that choice is the conclusion of a process of problemsolving. In this process the limits of human rationality certainly play a central role, but so does the relationship between actor and environment. Both these elements, already emphasized in the 1950s by Simon and Hayek, are revived in Kahneman and Tversky’s empirical investigations and then implemented in their theoretical constructions. Mental representation indeed rests on those perceptual mechanisms studied by cognitive sciences. In other words, it depends on the way in which each agent interprets external data and this in turn depends on his available cognitive structures, which are limited and subject to change as a consequence of previous experience. In the light of this discussion, it becomes evident that Kahneman and Tversky’s main contribution does not rest on the possibility to formalize human behaviors that depart from the canons of expected utility theory. It rather consists in their emphasizing the importance of

The knowledge–rationality connection in Herbert Simon 161 subjective mental representation of a decision context and the cognitive mechanisms that constitute the basis for the outcome of that very decision. Therefore, while under the first interpretation, Kahneman and Tversky’s contribution rather resembles an extension of Friedman’s than of Simon’s approach, the second interpretation acknowledges its continuity with Simon and Hayek. In the first interpretation, the concept of rationality offers its flank to the criticism we directed at Friedman’s (Section 7.3.1) and the concept of knowledge fails to capture the distinction between knowledge and information (Section 7.2.1): in this sense, procedurality is left out of the picture. The second interpretation, instead, acknowledges the empirical–descriptive foundations of human rationality and its procedural nature, determined by the continuous interaction of subject and environment and the continuous changes in this interaction. Knowledge is central to this interaction. More generally, the second interpretation of Kahneman and Tversky’s work explicitly encompasses an empirical approach to interaction.

7.5 CONCLUDING REMARKS The relevance of Simon’s contributions in highlighting the role of rationality in economics is largely acknowledged in literature. What most economists mainly seem to have in mind is the role played by the concept of bounded rationality in understanding non-optimal decision making. Yet a deeper analysis shows that Simon’s theory of rationality is the outcome of a more complex process, which leads to a new model of problem solving and decision making. For this process, bounded rationality simply represents a (certainly fundamental) starting point, while most seminal and innovative results are the outcome of the second part of Simon’s analysis: the concept of procedural rationality. The differences between bounded rationality and procedural rationality do not only concern their conceptual nature, but also and mainly their analytical paths. Bounded rationality emerges in a theoretical context characterized by the centrality, for economists too, of several new concepts appropriated from cognitive sciences. In this perspective, it may be regarded as a criticism of rational choice theory. Instead, procedural rationality stems from a course that has its theoretical necessary starting point in bounded rationality and aims at solving the problem of uncertainty. In this view, there are several components of the theory of procedural rationality: learning, perception, representation, acquired knowledge, path-dependence, interaction, emotions, to name a few. Many of these elements are also crucial components of human knowledge.

162

Handbook of knowledge and economics

Consequently, knowledge, in Simon’s work, is strictly connected with rationality in its procedural dimension. This consideration has relevant implications. Most importantly, as this chapter has tried to show, it helps to explain why economists working on the topic of the theory of rationality in the same ‘location’, simply theoretical or even geographical, gave different explanations. Specifically, we could say that, in addressing the limitations of full economic rationality, Friedman gives a strong (but exclusively theoretical and epistemological) theory, whereas Simon is more accepting of the cognitive nature of those limits, takes up the challenge to address the problem of uncertainty, and develops a theory of decision making grounded on the concept of ‘pragmatic’ rationality, which is built around human knowledge.

NOTES 1. 2.

3. 4. 5.

6.

7.

8. 9. 10.

Hayek’s model of mind results from his neurobiological studies. They merged into The Sensory Order, published in 1952 but conceived, at least in part, during the 1920s. Here it seems important to remember that, during the second half of the 1940s, Katona proved the relevance of individual preferences and their empirical verifiability, through the ‘survey method’, which he described in the attempt to show its theoretical strength (Katona, 1951). The importance and verifiability of preferences undermines an important epistemological Walrasian foundation, precisely the assumption of their irrelevance and of the impossibility to verify them only at the individual level. Nevertheless, in the same period the efforts to defend the scientific nature of economics found strong support in expected utility theory. Savage (1954) tries to include uncertainty within subjective expected utility theory. For a general reconstruction on the origins of cognitive sciences, see Gardner (1985). ‘Although the development of the diverse doctrines of bounded rationality and rational expectations by collaborators could be viewed as just a historical coincidence, it is more likely that an intense preoccupation with a set of problems led to two researchers on different paths in search of a solution’ (Sheffrin, 1996, p. 1). The evolutionary–adaptive explanation supplied by Friedman was supported by Alchian’s work (1950), from which arises the branch of evolutionary economics. This branch will develop in the 1980s (Nelson and Winter, 1982) and successively in the 1990s (Dosi and Nelson, 1994; Nelson, 1995). The process of aspiration levels can be described as follows: if the settled levels of aspiration are too easily reached, the individual is likely to raise them in the following action. The levels of aspiration are frequently lowered or raised, according to different experiences. If the environment is perceived as favorable and offers several positive alternatives, the levels of aspiration are raised; in an unfriendly environment such levels are lowered. It is fair to recall that nowadays there are new and different approaches to game theory. This is in line with Allais (see Section 7.2.2). Camerer writes about Kahneman and Tversky: ‘This sort of psychology provided a way to model bounded rationality which is more like standard economics than the more radical departure that Simon had in mind.’ And, more gerenally, ‘much of behavioral economics consists of trying to incorporate his kind of psychology into economics’ (Camerer, 1999).

The knowledge–rationality connection in Herbert Simon 163

REFERENCES Alchian, A. (1950), ‘Uncertainty, evolution and economic theory’, Journal of Political Economy, 58, 211–21. Allais, M. (1953), ‘Le comportment de l’homme rationnel devant le risque: critique des postulats et axiomes de l’école americaine’, Econometrica, 21, 503–46. Aumann, Robert J. (1997), ‘Rationality and bounded rationality’, Games and Economic Behavior, 21, 2–14. Camerer, C.F. (1999), ‘Behavioral economics’, CSWEP Newsletter, winter, www.cswep.org/ camerer.html. Dosi, G. and Nelson, R. (1994), ‘An introduction to evolutionary theories in economics’, Journal of Evolutionary Economics, 4 (3), 153–72. Egidi, M. (2005), ‘From bounded rationality to behavioral economics’, Social Science Research Network, El. Paper Collection, http://papers.ssrn.com/sol3/papers.cfm? abstract_id=758424. Ellsberg, D. (1961), ‘Risk, ambiguity, and the Savage axioms’, Quarterly Journal of Economics, 75, 643–69. Friedman, M. (1953), Essays in Positive Economics, Chicago, IL: University of Chicago Press. Gardner, H. (1985), The Mind’s New Science, New York: Basic Books. Hayek, F.A. (1937), ‘Economics and knowledge’, Economica, 4 (13), 96–105. Hayek, F.A. (1945), ‘The use of knowledge in society’, American Economic Review, 35 (4), 519–30. Hayek, F.A. (1952), The Sensory Order: An Inquiry into the Foundations of Theoretical Psychology, London: Routledge & Kegan Paul. Kahneman, D. and Tversky, A. (1979), ‘Prospect theory: an analysis of decision under risk’, Econometrica, 47, 263–91. Katona, G. (1951), Psychological Analysis of Economic Behavior, New York: McGraw-Hill. Knight, F.H. (1921), Risk, Uncertainty and Profit, Boston, MA: Houghton Mifflin. Lucas, R.E. Jr (1981), ‘Optimal investment with rational expectations’, in R.E. Lucas and T.J. Sargent, Rational Expectations and Econometrics, Minneapolis, MN: University of Minnesota Press, pp. 55–66. Markowitz, H. (1952), ‘The utility of wealth’, Journal of Political Economy, 60, 151–8. Muth, J.F. (1961), ‘Rational expectations and the theory of price movements’, Econometrica, 29, 315–35. Nelson, R.R. (1995), ‘Recent evolutionary theorizing about economic change’, Journal of Economic Literature, 33 (1), 48–90. Nelson, R.R. and Winter, S.G. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Neumann, J. von and Morgerstern, O. (1944), Theory of Games and Economic Behavior, Princeton, NJ: Princeton University Press. Newell, A. and Simon, H. (1972), Human Problem Solving, Englewood Cliffs, NJ: Prentice-Hall. Rizzello, S. (1997), L’economia della Mente, Roma–Bari: Laterza. English translation, The Economics of the Mind, Cheltenham, UK and Lyme, USA: Edward Elgar. Savage, L.J. (1954), The Foundations of Statistics, New York: Dover Publications. Sent, E.-M. (1999), ‘Sargent versus Simon: bounded rationality unbound’, Cambridge Journal of Economics, 21, 323–38. Sent, E.-M. (2004), ‘Behavioral economics: how psychology made its (limited) way back into economics’, History of Political Economy, 36 (4), 735–60. Sent, E.-M. (2005), ‘Simplifying Herbert Simon’, History of Political Economy, 37 (2), 227–32. Sheffrin, S.M. (1996), Rational Expectations, 2nd edn, Cambridge: Cambridge University Press. Simon, H.A. (1957a), ‘Introduction to the Second Edition’, in H.A. Simon, Administrative Behavior, New York: Macmillan, pp. ix–xli.

164

Handbook of knowledge and economics

Simon, H.A. (1957b), Models of Man, New York: Wiley. Simon, H.A. (1976), ‘From substantive to procedural rationality’, in S. Latsis (ed.), Method and Appraisal in Economics, Cambridge: Cambridge University Press, pp. 129–48. Simon, H.A. (1979), ‘Rational decision making in business organizations’, American Economic Review, 69, 493–512. Simon, H.A. (1987a), ‘Behavioral economics’, in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of Economics, Vol. I, London: Macmillan, pp. 221–5. Simon, H.A. (1987b), ‘Bounded rationality’, in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of Economics, Vol. I, London: Macmillan, pp. 266–7. Simon, H.A. (1987c), ‘Satisficing’, in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of Economics, Vol. IV, London: Macmillan, pp. 243–4. Simon, H.A. (1988), ‘Why economists disagree’, Journal of Business Administration, 18 (1,2), 1–19. Simon, H.A. (1989), ‘The state of economic science’, in W. Sichel (ed.), The State of Economic Science: Views of Six Nobel Laureates, Kalamazoo, MI: W.E. Upjohn Insitute for Employment Research, pp. 97–110. Simon, H.A. (1993), ‘Altruism and economics’, American Economic Review, 83 (2): 156–61. Simon, H.A. (1994), ‘Preface’, in H. Gabrié and J.L. Jacquier (eds), La Théorie Moderne de l’Entreprise: L’Approche Institutionelle, Paris: Economica, pp. 7–12. Tversky, A. and Kahneman, D. (1992), ‘Advances in prospect theory: cumulative representation of uncertainty’, Journal of Risk and Uncertainty, 5, 297–323. Williamson, O.E. (2002), ‘Empirical macroeconomics: another perspective’, in M. Augier and J. March (eds), The Economics of Choice, Change and Organization, Aldershot, UK and Brookfield, US: Edward Elgar, pp. 419–41.

PART II ECONOMICS, KNOWLEDGE AND UNCERTAINTY

8

A note on information, knowledge and economic theory Giovanni Dosi

This chapter addresses the question of what economic theory has to do with knowledge, in general, and more specifically, with the interpretation of the contemporary so-called ‘knowledge-based’ economy. In fact, one may reasonably come up with two opposite answers. The first one is that in one sense, which I shall specify shortly, economic theory is intrinsically about knowledge-based economies. The opposite answer, which I consider at least equally true, is that most strands of current theory have very little to say by way of an analysis of the nature of the particular form of economy that one observes nowadays and its relations with the transformation in its knowledge bases. Some words on the first point might help to clarify the second one. One of the central objects of inquiry of economic theory since its origin as a discipline has been precisely the interactions among a multitude of decentralized agents and the ensuing collective outcomes. (Everyone has heard of Adam Smith’s ‘invisible hand’ conjecture on the properties of decentralized markets . . .) But in an essential sense, asking how a decentralized economy works is equivalent to asking how socially distributed knowledge is collectively put to work in ways that are not socially detrimental and, possibly, increase the welfare of everyone. Adam Smith’s conjecture (subject to several qualifications, many of which have been missed by later theorists) was indeed that markets are able to elicit private knowledge, propelled by the pursuit of self-interest, and yield orderly outcomes, superior – in terms of ‘welfare’ – to, say, an autarkic system of production and consumption. The point that an economy is basically a system of distributed, diverse, pieces of knowledge has been emphasized, among others, by von Hayek. And, of course, this is also a way of reading the most rigorous formalization of the economy as an interdependent system, namely general equilibrium (GE) analysis as put forward in the 1950s and 1969s by Arrow, Debreu, Hahn and McKenzie (noting at the same time that ‘knowledge’ takes a much narrower meaning under the GE ‘information’ heading: see below). The existence theorems, there, are a way of saying that, among all the imaginable worlds, one can also coherently conceive of an economy wherein every selfishly motivated agent, by 167

168

Handbook of knowledge and economics

making the best use of its own information, contributes to ‘sharing its use’ with all other agents in ways that are mutually consistent and also mutually beneficial. (I am provisionally using ‘information’ and ‘knowledge’ here as equivalent concepts.) So, yes, in this general and rather abstract sense, economic theory has always been about interdependencies in knowledge-intensive systems. However, it is enough to check the long list of assumptions that one has to make in the canonical GE model in order to fully appreciate the distance between what it says and the interpretative requirements of any one historically observed economy. (Incidentally, note also that the very pioneers of the theory are well aware of this, unlike many of the following believers: compare the writings of Kenneth Arrow or Frank Hahn with any random sample of articles from the Journal of Economic Theory or Econometrica. Indeed, when I see works on empirically applied GE models, I must confess that I have the same feeling I had when I saw long ago at UC Berkeley the announcement of a seminar on ‘Applied Heidegger’!) The long list of restrictive assumptions is also an indicative proxy for the phenomena that economic theory is unable to account for (at least in that analytical format); for the progresses (and regresses) that have been recently made; for the humility that economists should, but generally do not, put into their policy prescriptions; and, last but not least, for the healthy amount of scepticism that non-economists should retain when listening to economists’ wisdom.

8.1 INFORMATION, KNOWLEDGE AND ECONOMIC THEORY As mentioned, GE is a very elegant, institutionally very parsimonious, representation of how agents best use the available information and interact with each other accordingly. But ‘information’ is not an ordinary good that can be treated, say, like a machine tool or a pair of shoes (again, on the economic characteristics of information, Arrow is a pioneering reference). Shoes wear out as one uses them, while information typically has a high up-front cost in its generation but can be used repeatedly without decay thereafter, or there might even be learning-by-using type phenomena (as from the first to the nth time one applies Pythagoras’ theorem in high school). Moreover, information might be appropriable, in the sense that other agents might have significant obstacles to accessing it (ranging from legal protections, such as patents, all the way to the sheer difficulty of fully appreciating what a particular piece of information means: see also below). But information as such typically entails a non-rival use

A note on information, knowledge and economic theory 169 (in the sense that it can be utilized indifferently by one or one million people, which, again, is not the case of ordinary commodities like shoes or machine tools). In my view, some of the most important advances of the theory over the last two or three decades have concerned precisely the economic consequences of these features of information. Without entering into any detail, recall the wide literature on the ‘economics of information’ and of ‘information processing’ (see, in primis, the pioneering works of Kenneth Arrow, Herbert Simon, Richard Nelson, George Akerlof and Joseph Stiglitz); on ‘principal–agent’ models, most often studying the incentive implications of imperfect, asymmetric information, on the grounds of otherwise quite orthodox assumptions; on the organizational implications of information-related transaction costs and collective rents (see, e.g., the works of Oliver Williamson and Masahiko Aoki); and on ‘new growth models’ explicitly incorporating the generation of technological information (see the contributions of Philip Aghion, Paul Romer, Elhanan Helpman, Peter Howitt and many other colleagues). For our purposes here, let me just recall three major implications of even the most rudimentary accounts of the specificity of information for economic theory. First, the ‘invisible hand’ properties of the canonical GE model do not generally carry over to economic models where the most restrictive informational assumptions are relaxed (e.g. on the perfect access to information by all agents and on the fact that information itself drops freely from the sky). So the theory may easily predict equilibria and growth paths that are socially sub-optimal, systematic divergences between rewards and marginal products, and also the possibility of long-term unemployment. Second, the social distribution of information, and thus the institutional architectures of the system, matters a lot in terms of microeconomic incentives and aggregate performance. Third, by adding the highly plausible assumption of ‘locality of learning’, one easily obtains ‘path-dependent’ models of development – at the levels of individual firms, technologies, industries and whole countries (see the contributions of Paul David, Brian Arthur, Cristiano Antonelli, Richard Nelson, Sidney Winter, Dosi, and many other ‘evolutionary’ models of economic change). Impressionistically, ‘locality’ stands for the fact that you most probably learn by building upon what you already know (so that pushing the examples to the caricature, it is much easier to learn differential equations after having taken the course of calculus than without it; or, even at an aggregate level, the probability that the next generation of microprocessors will be invented in the USA, conditional on the past innovative performance in the field, is much higher than in

170

Handbook of knowledge and economics

Burkina Faso). And locality/path-dependence stands also for the relative incremental coherence in the domains of exploration that individuals, organizations and possibly countries may attain (so that, for example, becoming a great economist does not make it easier for you to become a good football player, being a competitive textile manufacturer is not likely to help in competing in bioengineering etc.). On the contrary, one often finds relatively coherent ‘trajectories’ in technological learning (cf. Dosi and a few other contributors to the economics of innovation). Incidentally, note also that path-dependence in learning is likely to entail tricky dilemmas between ‘exploitation’ and ‘exploration’ – in the terminology of James March – that is, between allocations of efforts aimed at improving what one is already good at doing versus activities of search for uncertain novelties. Putting it somewhat bluntly, even simple accounts of some essential characteristics of information analytically shake the naive and Panglossian belief that unhindered market mechanisms yield the best of all possible worlds. To use a term that I rather dislike, ‘market failures’ are generally associated with the production and use of information. Intuitively, for this to happen it is sufficient to acknowledge the properties mentioned above concerning: (i) increasing returns; and (ii) non-rivalry in the use of information.The former obviously tend to conflict with the idea that pure competition is normatively the best form of market organization, and also with the idea that competition can sustain itself as a viable market structure. The latter decouples the costs of generation and the benefits of use of information: after all, one could say that the cost of production of, say, Pythagoras’ theorem was entirely borne by Pythagoras himself, while all subsequent generations benefited from it for free. Relatedly, such a decoupling might induce underinvestment in information generation (and attempts to tackle the problem via an increased appropriability of its benefits might even have perverse outcomes). Moreover, as is well known in the theory, necessary conditions for some close link to hold between marginal productivities of inputs, relative prices and distributive shares are decreasing returns with respect to the use of the inputs whose productivity we are measuring (even neglecting the paramount difficulties involved in the measurement itself). Again, the acknowledgement of the role of information as a ‘factor of production’ breaks that link because of the increasing returns and the externalities associated with its generation and use. Has anyone ever tried to measure the ‘marginal productivity’ of Fermi and Oppenheimer within the Manhattan Project? Link them to their relative price? Account for their inputs into subsequent ‘atomic bomb production functions’? Well, it follows from the economics

A note on information, knowledge and economic theory 171 of information that similar overwhelming difficulties apply to the GM, or Microsoft or Boeing ‘production functions’, and, more so, to their aggregation, such as the ‘US production function’. I would like to emphasize that all the arguments so far can comfortably rest upon rather conventional assumptions regarding in particular the ‘rationality’ of agents – at least in their ability to make the best use of the information they access (whatever that means) – and on collective ‘equilibrium’ set-ups (which is a very very strong assumption on the collective consistency of individual plans). Some economists, notably those with ‘evolutionary’ and ‘institutionalist’ inclinations, depart even further from the canonical assumptions and suggest the following points, admittedly more controversial among practitioners.

8.2 INFORMATION AND KNOWLEDGE A distinction ought to be drawn between information and knowledge. The former entails well-stated and codified propositions about ‘states of the world’ (e.g. ‘it is raining’), properties of nature (e.g. ‘A causes B’) or explicit algorithms on how to do things. On the other hand, knowledge, in the definition I am proposing here, includes: (i) cognitive categories; (ii) codes of interpretation of the information itself; (iii) tacit skills; and (iv) problem-solving and search heuristics irreducible to well-defined algorithms. So, for example, the few hundred pages of demonstration of the last Fermat theorem would come under the heading of ‘information’. Having access to that, some dozen mathematicians in the world will have the adequate knowledge to understand and evaluate it. Conversely, a chimpanzee facing those same pages of information might just feel like eating them, and the majority of human beings would fall somewhere in between these two extremes. Similarly, a manual on ‘how to produce microprocessors’ is ‘information’, while knowledge concerns the pre-existing abilities of the reader to understand and implement the instructions contained therein. Moreover, in this definition, knowledge includes tacit and rather automatic skills like operating a particular machine or correctly driving a car and overtaking another one (without stopping first in order to solve the appropriate system of differential equations involved!). And, finally, it includes ‘visions’ and ill-defined rules of search, like those involved in most activities of scientific discovery, and in technological and organizational innovation (for example, proving a new theorem, designing a new kind of car, figuring out the behavioural patterns of a new kind of crook that has appeared on the financial market). In this definition, knowledge is partly tacit, at the very least in the sense that the agent itself, and even a

172

Handbook of knowledge and economics

very sophisticated observer, would find it very hard to explicitly state the sequence of procedures by which information is coded, behavioural patterns are formed, problems are solved and so on. This is certainly a major admission of ignorance on the part of the analyst, but there are good – almost ‘ontological’ – reasons for this. After all, as Ken Arrow himself pointed out long ago, if an innovation is truly an innovation it is impossible for a cognitively bounded observer to precisely forecast it. And, indeed, there are powerful uncomputability theorems that confirm this intuition. But ‘tacitness’ – some of us suggest – extends also to domains where little invention is involved (as mentioned above, driving cars, operating machine tools, debugging computer programs and, even more so, efficiently running production flows, interpreting market trends etc.).

8.3 THE ROLE OF CORPORATE KNOWLEDGE In modern economies, firms are major, albeit by no means unique, repositories of knowledge. Individual organizations embody specific ways of solving problems that are often very difficult to replicate in other organizations or even within the organization itself. In turn, organizational knowledge is stored to a large extent in the operating procedures (the ‘routines’) and the higher-level rules (concerning for example ‘what to do when something goes wrong’, or ‘how to change lower-level routines’) that firms enact while handling their problem-solving tasks in the domains of production, research, marketing and so on. Dynamically, technological knowledge is modified and augmented partly within individual firms, and partly through interaction with other firms (competitors, users, suppliers etc.) and other institutions (universities, technical societies etc.). In fact, over the last three decades, at least, a great deal of effort – within the broad field of the economics of innovation – has gone into a better understanding of the variety of processes by which knowledge is augmented and diffused in the economy (major contributions in this area include those by Christopher Freeman, Keith Pavitt, Richard Nelson and Nathan Rosenberg, among others). A first broad property – probably not surprising to non-economists, but with important analytical and normative implications – is the diversity of learning modes and sources of knowledge across technologies and across sectors. For example, in some activities, knowledge is accumulated primarily via informal mechanisms of learning by doing, learning by interacting with customers and suppliers and so on, while in others it involves much more formalized activities of search (such as those undertaken in R&D

A note on information, knowledge and economic theory 173 laboratories). In some fields, knowledge is mostly generated internally and is specific to particular applications. In others, it draws much more directly upon university research and scientific advances. I am mentioning all this because recent research suggests that this diversity of learning modes is also a major determinant of the diverse patterns of evolution in industrial structures (e.g. in terms of distribution of firm sizes, natality and mortality of firms, corporate diversification etc.). Moreover, the identification of sectoral specificities in the forms of knowledge and in learning patterns bears straightforward normative consequences (e.g. R&D policies or policies aimed at speeding up the diffusion of innovations are likely to have quite diverse effects in the textile industry versus bioengineering). In a related way, an important step in the understanding of the ‘anatomy’ of contemporary systems of production and knowledge accumulation has involved taxonomic exercises (Keith Pavitt’s taxonomy is probably the most famous one), trying to map ‘families’ of technologies and sectors according to their sources of innovative knowledge and their typical innovative procedures. Finally, note that building upon the considerations mentioned so far on the nature of technological learning and on the ways organizations incorporate knowledge, a few scholars have started to explore an explicitly coevolutionary view, whereby the accumulation of technological knowledge demands and possibly triggers changes in corporate organizations and broader institutions. To sum up, it seems to me that various strands of research within the fields of the economics of information, the economics of innovation and organizational theory have recently contributed a great deal to our understanding of how knowledge-rich economies work (and, equally important, of how they do not work!). However, the thrust of most of the works that I have discussed so far is a microeconomic one. This does not mean to say that they are void of macroeconomic content. On the contrary, it turns out to be relatively easy and highly promising to incorporate some of the mentioned findings on the economics of information, knowledge and learning into macroeconomic models. So, for example, self-sustained growth can be shown to be a general property of knowledge-based economies, even independently from capital accumulation (of course, in less abstract models, knowledge accumulation and capital accumulation are intertwined, and self-propelled dynamics apply). Moreover, the introduction of asymmetric information into simple macro models generally yields ‘Keynesian’ outcomes, such as persistent involuntary unemployment, credit rationing and so on (cf. the ‘New Keynesian’ contributions pioneered by Stiglitz and colleagues).

174

Handbook of knowledge and economics

From a complementary angle, an expanding family of evolutionary models, microfounded in a multitude of heterogeneous agents that imperfectly learn and are selected in various markets, is proving capable of accounting for a wide set of aggregate regularities, ranging from the patterns of international growth of incomes and productivities all the way to ‘meso’ phenomena such as size distributions of firms and their persistent asymmetries in efficiency (cf. the works spurred by Richard Nelson and Sidney Winter’s evolutionary theory of economic change, and also a few works of ours from the ‘Pisa team’). All this notwithstanding, it seems to me equally true that there is still an enormous gap between the wealth of microeconomic findings, on the one hand, and the understanding that we have of how knowledge is distributed in the economy as a whole and the ways this affects its performance and dynamics, on the other. This holds at analytical level and bears all its consequences at a normative one. For example, the theory is still ill equipped to tackle questions such as the conditions under which ‘technological unemployment’ emerges, the effects of particular patterns of technical change on growth, or the collective impact of specific institutional arrangements. Correspondingly, it is particularly weak in answering policy questions such as those concerning unemployment in knowledgebased economies. I now briefly turn to these issues.

8.4 FROM MICRO TO MACRO . . . It is interesting to note that within the economic discipline, the progressive attention, over the last three decades, to the intricacies of the generation and use of knowledge in an economy has been paralleled, within a good deal of macroeconomic theory, with a movement in the opposite direction. It is impossible here to enter into the finer details of macroeconomic controversies, and, even less so, their sometimes bizarre epistemological justifications. As a first and rough approximation, note that, as mentioned above, most advances in the interpretation of the role of knowledge in economic coordination and change might be understood, with reference to a canonical GE model, as more or less radical departures from its most demanding assumptions regarding, for example, the institution-free environment, the information available to individual agents, their basic homogeneity (apart from differences in their preferences and initial endowments), their rational ability to understand the world they live in, to exploit the opportunities it provides and to forecast the future. Well, the trend in much current macro theory has been, if anything, toward assumptions of

A note on information, knowledge and economic theory 175 even greater homogeneity among them. As a rough but vivid illustration of this statement, it is revealing to compare any sample of an intermediateto-advanced macro textbook of, say 40 years ago, with what is mostly taught nowadays (parallel comparisons of state-of-the-art publications would only reinforce the argument). In the former, one finds a good deal of macro statements based upon comparative-static exercises involving relationships among aggregate entities (e.g. the ‘aggregate propensity to consume’ the ‘multiplier’, the ‘accelerator’, ‘IS–LM curves’ etc.). And they also displayed the most rudimentary forms of ‘informational imperfection’ and ‘bounded rationality’, namely – most often – crude adaptive expectations, ‘money illusions’ and the like. Obviously, a way forward could have been a much greater refinement of the microeconomic foundations, interactive dynamics, information processes, learning mechanisms, institutional assumptions and so on. Unfortunately, what has happened in the mainstream of the discipline has been the opposite (for reasons – partly internal to the sociology of the discipline itself, and partly due to a broader Zeitgeist – that I do not have space to discuss here): the ‘rational expectations’/‘new classical economics’ paradigm is an extreme example of this tendency. So, most often, the enormous gap between the assumptions implied in the GE model of economic coordination, on the one hand, and observed behavioural traits and institutional conditions, on the other, is written away with an act of faith, and a more elegant account of macrodynamics is derived from the optimizing behaviours of a ‘representative’ agent. (This notwithstanding much handwaving concerning, for example, the derivation of ‘representative agents’ themselves from a GE set-up – see, e.g., the profoundly disruptive observations of Alan Kirman, among others – or the general impossibility of generating models whereby even fully forwardlooking, ‘representative’, agents can learn their equilibrium behaviour.) Moreover, as regards the ‘rationality’ attributed to the agents, 40 years ago they were assumed to be able to take moving averages and recognize the sign of derivatives; nowadays they ought to be able to solve complicated intertemporal optimization problems (or, at least, behave in equilibrium as if they did). I mention all this for two reasons. First, from a theoretical point of view, if one were to accept such a macroeconomic view, it would be an idle waste of time to discuss issues such as ‘the implications of a knowledgebased economy’. Simply put, no matter how high the level of knowledge incorporated into any one economy, if agents fully mastered it, and if we also ruled out the specificities of information and knowledge discussed in the previous section, no problem would arise. Indeed, one could think of a macrodynamic summarizing of a sequence of optimal adjustments by fully

176

Handbook of knowledge and economics

rational agents to exogenous shocks all the way from the Stone Age to the Microprocessor Age (in this respect, readers not too familiar with the esoteric debates of economists are invited to check the interpretations that the professional community takes seriously of, e.g., the Great Depression of the 1930s or even the current depression!). Second, and relatedly, a good part of policy discussion draws rather closely on the agenda set by macroeconomic theory. As a consequence, in the current agenda there is very little room for questions concerning, for example, the specificities of particular forms of socially distributed knowledge and their effects on unemployment, income distribution, growth and so on. At the same time, there seems to be a dangerous tendency to derive policy prescriptions from the original acts of faith built into the theory regarding the self-adjusting properties of the economy. To caricature only slightly, no matter what the policy problem at hand, one often hears the answer ‘just let the market work’. But the right questions are: precisely how do markets work? How are they affected by different informational structures and mechanisms of knowledge generation? And, indeed, we still know very little about the answer. In brief, my view is that a major and urgent task ahead is a sort of reconstruction of macroeconomic theory building upon the rich insights into knowledge, corporate organizations and institutions briefly reviewed above (and of course drawing upon the quite a few existing macro models that already try to do it). Short of that, I shall just put forward some scattered remarks, without any claim to coherence. Going from detailed micro descriptions of ‘knowledge-intensive’ economies to necessarily more concise aggregate accounts requires also important commitments on the mechanisms of coordination and adjustments among agents who are diverse in terms of the knowledge that they embody and the institutional positions they occupy. One way out, clearly, is to assume some implicit GE and get rid of the job. However, all that has been said so far makes that assumption particularly doubtful. As an illustration, consider the following. Start as a reference, again, from a GE. There, the intuitive image of how coordination occurs is a multitude of agents bringing their goods to the square of the village and trading with each other; ‘adjustments’ occur via the way people ‘go up’ supply curves and ‘go down’ demand curves as notional prices change; and, finally, at the end of the day everything that there is to know is summarized by the ensuing prices. Moreover, with the appropriate modifications, one may extend the same image to a GE with production (with people also buying and selling inputs) and to economies where people think of what they might want tomorrow (technically, things are much more complicated than that, but for our purposes this metaphor is

A note on information, knowledge and economic theory 177 sufficient). In this view, the first task that economic interactions address is to convey information about the relative availability of goods and services and their prices, while the flows of ‘technical’ knowledge remain ‘blackboxed’ into constructs such as ‘production functions’. Conversely, a knowledge-based view is much more ‘Hayekian’ in spirit. People might still meet in the village square, but their purpose is not only to trade goods but ‘to do things’ on the grounds of their diverse pieces of knowledge (someone is good at designing engines and someone else at selling them etc.). As we know, ‘trading knowledge’ is difficult because one cannot fully appreciate its value before having applied it. In any case it would be hard to price it due to increasing returns (and focusing on the ‘trading services’ that incorporate knowledge does only little to mitigate the problem). Also, incentive compatibility problems generally emerge. Moreover, this is likely to be a world of complementarity rather than substitution (design and marketing knowledge are only useful together). And, finally, people might augment their knowledge just by talking to each other. One can clearly see that, in such a world, ‘going up and down demand curves’, alone, is not likely to do the trick of coordination: one will require some further specifications on the way people get together, talk to each other, organize what they do. That is, in order to understand how that system coordinates and changes over time, one will need to know much more about its institutional architecture and about the patterns of learning. Moreover, all this would apply if one were to abandon the metaphor of the village square and instead assume that agents are also physically dispersed and interact with only a sub-set of the population. Unfortunately, current economic theory – even in its ‘evolutionary’ and ‘institutionalist’ versions – still falls short of providing comprehensive taxonomies of coordination and learning mechanisms that could then be ‘reduced’ into tractable macro models. So, in the above illustration, one would like to have some sort of archetypal patterns of the way people share their knowledge, sell their services, organize their production activities and so on, and then study the collective dynamic properties of different institutional set-ups. Promising theoretical attempts are there, but we are still quite far from the goal. The following are three rather different examples in this direction. First, we have started to see exercises of ‘comparative institutional analysis’, which continue to share with the GE world the focus upon equilibrium situations and also the assumption that agents are entirely capable of making the best use of the information they get, but the interest of the exercise rests precisely in allowing different systems to distribute differently the information – therefore also providing different incentive structures – and

178

Handbook of knowledge and economics

also to socially distribute different menus of available courses of action (works like those by Akerlof, Stiglitz or M. Aoki head in this direction). Second, a forthcoming generation of agent-based models – which are more ‘bottom-up’, in the sense that they explicitly represent a multitude of agents who interact with each other without any prior commitment to any collective equilibrium – seems well suited to handle thought experiments concerning the aggregate effects of different distributions of knowledge and different interaction mechanisms. Third, one finds in the institutionalist macro literature – including the French ‘regulation approach’ and the ‘variety of capitalism’ investigations – various attempts to identify and sometimes formalize a sort of historical taxonomy of ‘regimes’ governing the interaction mechanisms in the various markets, for example products, labour, finance and so on (e.g. the works of Michel Aglietta, Robert Boyer, Benjamin Coriat, Peter Hall, David Soskice, Wolfgang Streeck, among others). In that, the ‘regulation’ approach is rather ‘top-down’ in the sense that it often starts from daring assumptions on functional relationships among aggregate variables (e.g. wages and productivity, income growth and productivity growth etc.), but in fact, even beyond its contribution to historical analysis, it might turn out in the end to be a complement and a challenge to more ‘bottom-up’, behaviourally richer, models.

8.5 A FEW MORE CONCLUDING QUESTIONS If I were to end here, I would simply summarize this quick overview of the contribution of economic theory to the understanding of knowledgebased economies with qualified optimism on the ability of the discipline to shed some light on some of their important aspects. In particular, I have argued, recent developments in the economics of information and of innovation have brought important insights into the processes of generation and diffusion of knowledge, and their economic consequences, although many streams of macroeconomic analysis are lagging behind in taking them on board.In the whole foregoing discussion, the emphasis has been placed on the toolkit of analytical categories, models and conjectures that economists have to offer in general rather than on the interpretation of specific contemporary trends. In fact, it follows from the perspective that I have tried to outline here that, in an essential sense, all economies that we know are profoundly knowledge-based: they were so a century ago, and they are now. However, with an adequate toolkit one might be able to identify what distinguishes the contemporary role of knowledge from that, say, observed by Marshall or Schumpeter. A few crucial questions (to

A note on information, knowledge and economic theory 179 which I shall not attempt any answer), directly based on the interpretative categories introduced above, illustrate the point: ● ● ●

● ●

● ●

How have the sources and procedures of knowledge accumulation changed? Have new relationships emerged between accumulation of knowledge and accumulation of physical capital? Is it true that the balance between economically useful tacit knowledge and codified information is shifting in favour of the latter? And with what consequences? What are the patterns in the social (and also international) distribution of knowledge? How does all this affect market interactions? If new modes and directions of knowledge accumulation are identified, what are their implications in terms of corporate organization and strategies? What kind of new institutional arrangements have emerged, if any? What are the implications of all this in terms of employment, growth and income distribution?

Note that it is an improved theoretical toolkit that allows us to pose with precision these very questions (even if we are still far from satisfactory answers!). A general conjecture here is that we are currently witnessing a secular technological transformation that is affecting the basic economic mechanisms of demand formation, accumulation, employment generation and, together, the very fabric of society. The basic message of this short chapter is that, yes, economic theory can contribute to its understanding, but there is still a long way to go. There are major analytical issues to which the economics discipline can potentially offer a great deal, but to large extent has not yet delivered the goods. Consider, just as examples, the question of the ‘compensation effects’ of technical progress (i.e. under what circumstances are the employment-destroying effects of innovation compensated by employment creation of equal or greater magnitude?); or the stabilizing/destabilizing effects of faster/wider access to information on market dynamics (e.g. what is the impact of new information technologies on financial markets and their broader consequences in terms of real aggregate variables and policy making?). There are other major questions with respect to which economic theory can be only part of wider interdisciplinary endeavours. For example, I do not think it is an exaggeration to say that the very structure of a democratic society rests upon forms of knowledge distribution that are not too asymmetric, allow sufficient mobility, and imply a reasonable ability of

180

Handbook of knowledge and economics

all citizens to understand the content of collective decisions. In turn, an urgent issue concerns precisely the maintenance of these conditions in the coming ‘information society’. Economists can contribute to the understanding of all this, but only together with sociologists, political scientists and so forth. So these conclusions are also a plea for scientific humility, or, to put it more vividly, given the current state of the art of the discipline, do not believe any economist who comes to you with simple answers and magic bullets! This applies even more so at the policy level. After all, among the few things that we do know is the fact that knowledge-based economies are always likely to embody unexploited opportunities not only for technological, but also for organizational and institutional, innovation. And it is with respect to the exploration of these opportunities that a fruitful dialogue can be established between economists and policy makers.

SELECTED READING Economic Properties of Information (and Some Economic Implications) Aghion, P. and Howitt, P. (1998), Endogenous Growth Theory, Cambridge, MA: MIT Press. Akerlof, G. (1984), An Economic Theorist’s Book of Tales, Cambridge: Cambridge University Press. Arora, A., Fosfuri, A. and Gambardella, A. (2001), Markets for Technology: Economics of Innovation and Corporate Strategy, Cambridge, MA: MIT Press. Arrow, K.J. (1962), ‘Economics of welfare and the allocation of resources for invention’, in R. Nelson (ed.), The Rate and Direction of Inventive Activity, Princeton, NJ: Princeton University Press, pp. 609–25. Arrow, K.J. (1962), ‘The economic implications of learning by doing’, Review of Economic Studies, 29 (3), 155–73. Arrow, K.J. (1974), The Limits of Organisation, New York: Norton. Arrow, K.J. (1996), ‘Technical information and industrial structure’, Industrial and Corporate Change, 5 (2), 645–52. March, J. (1994), Primer on Decision Making: How Decisions Happen, New York: The Free Press. Nelson, R. (1959), ‘The simple economics of basic scientific research’, Journal of Political Economy, 67 (3), 297–306. Romer, P. (1990), ‘Endogenous technological change’, Journal of Political Economy, 98 (5), Part 2, S71–S102. Shapiro, C. and Varian, H. (1999), Information Rules: A Strategic Guide to the Network Economy, Cambridge, MA: Harvard Business School Press. Simon, H.A. (1969), The Sciences of the Artificial, Cambridge, MA: MIT Press. Simon, H.A. (1982/1997), Models of Bounded Rationality, 1–3, Cambridge, MA: MIT Press. Spence, M.A. (1974), Market Signaling, Cambridge, MA: Harvard University Press. Stiglitz, J. (1974), ‘Information and economic analysis: a perspective’, Economic Journal, 95 (Supplement), 21–41.

A note on information, knowledge and economic theory 181 Stiglitz, J. (2000), ‘The contributions of the economics of information to twentieth century economics’, Quarterly Journal of Economics, 115 (4), 1441–78.

Knowledge, Innovation and Economic Evolution Arthur, B.A. (1994), Increasing Returns and Path Dependence in the Economy, Ann Arbor, MI: Michigan University Press. Cowan, R., David, P.A. and Foray, D. (2000), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 13, 211–53. David, P.A. (1985), ‘Clio and the economics of QWERTY’, American Economic Review Papers and Proceedings, 75 (2), 332–7. Dopfer, K. (2005), The Evolutionary Foundations of Economics, Cambridge: Cambridge University Press. Dosi, G. (1998), ‘Sources, procedures and microeconomic effects of innovation’, Journal of Economic Literature, 26 (3), 1120–71. Dosi, G., Freeman, C., Nelson, R.R., Silverberg, G. and Soete, L. (eds) (1988), Technical Change and Economic Theory, New York: Pinter Publishers. Dosi, G., Marengo, L. and Fagiolo, G. (2005), ‘Learning in an evolutionary environment’, in Dopfer (2005). Dosi, G. and Nelson, R.R. (2010), ‘Technical change and industrial dynamics as evolutionary processes’, in B.H. Hall and N. Rosenberg (eds), Handbook of the Economics of Innovation, Vol. I, Burlington, VT: Academic Press, pp. 51–128. Freeman, C. and Soete, L. (1982), The Economics of Industrial Innovation, London: Pinter. Freeman, C. (1994), ‘The economics of technical change’, Cambridge Journal of Economics, 18 (5), 463–514. Von Hayek, F. (1983), Knowledge Evolution and Society, London: Adam Smith Institute. Lundvall, B.Å. (1992), National Systems of Innovation, London: Pinter Publishers. Nelson, R.R. (1993), National Innovation Systems, Oxford: Oxford University Press. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: The Belknap Press of Harvard University Press. Pavitt, K. (1991), ‘What makes basic research economically useful?’, Research Policy, 20 (2), 109–19. Pavitt, K. (1999), Technology, Management and Systems of Innovation, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Rosenberg, N. (1990), Inside the Blackbox. Technology and Economics, Cambridge: Cambridge University Press.

From Decentralized Information to Aggregate Outcomes Kirman, A. (1989), ‘The intrinsic limits of modern economic theory: the emperor has no clothes’, Economic Journal, 99 (Supplement) (395), 126–39. Kirman, A. (1992), ‘What or whom does the representative individual represent?’, Journal of Economic Perspectives, 6 (2), 117–36.

Information, Knowledge and the Firm Aoki, M. (1988), Information, Incentives and Bargaining in the Japanese Economy, Cambridge: Cambridge University Press. Dosi, G. and Marengo, L. (1993), ‘Toward an evolutionary theory of organizational competences’, in W. England (ed.), Evolutionary Concepts in Contemporary Economics, Ann Arbor, MI: Michigan University Press.

182

Handbook of knowledge and economics

Dosi, G., Nelson, R.R. and Winter, S. (eds) (2000), The Nature of Dynamics of Organisational Capabilities, Oxford: Oxford University Press. Levinthal, D.A. (1997), ‘Adaptation on rugged landscapes’, Management Science, 43 (7), 934–50. Marengo, L. and Dosi, G. (2005), ‘Division of labor, organizational coordination and market mechanisms in collective problem-solving’, Journal of Economic Behavior and Organization, 58, 303–26. Teece, D. and Pisano, G. (1994), ‘Dynamic capabilities’, Industrial and Corporate Change, 3 (Special issue 3). Williamson, O. (1985), The Economic Institutions of Capitalism, New York: Free Press. Winter, S.G. (1988), ‘On Coase, competence, and the corporation’, Journal of Law, Economics and Organization, 4 (1), 163–80.

. . . and Some Hints at Different Institutionalist Approaches to Macroeconomics and Comparative Systems Aglietta, M. (1982), Regulation and Crisis of Capitalism, New York: Monthly Review Press. Aoki, M. (2001), Toward a Comparative Institutional Analysis, Cambridge, MA: MIT Press. Boyer, R. (1988a), ‘Technical change and the theory of “Régulation”’, in G. Dosi, C. Freeman, R.R. Nelson, G. Silverberg and L. Soete (eds) (1988), Technical Change and Economic Theory, New York: Pinter Publishers, pp. 67–94. Boyer, R. (1988b), ‘Formalizing growth regimes’, in G. Dosi, C. Freeman, R.R. Nelson, G. Silverberg and L. Soete (eds) (1988), Technical Change and Economic Theory, New York: Pinter Publishers, pp. 608–30. Boyer, R. and Hollingsworth, J.R. (1977), Contemporary Capitalism: The Embeddedness of Institutions, Cambridge: Cambridge University Press. Coriat, B. and Dosi, G. (1998), ‘The institutional embeddedness of economic change. An appraisal of the “evolutionary” and the “regulationist” research programme’, in K. Nielsen and B. Johnson (eds), Institutions and Economic Change, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 3–32. Dosi, G., Fagiolo, G. and Roventini, A. (2010), ‘Schumpeter meeting Keynes: a policyfriendly model of endogenous growth and business cycles’, Journal of Economic Dynamics and Control, 34 (9), 1748–67. Hall, P.A. and Soskice, D. (2001), Varieties of Capitalism. The Institutional Foundations of Comparative Advantage, Oxford: Oxford University Press. Stiglitz, J.E. (1994), Whither Socialism?, Cambridge, MA: MIT Press. Stiglitz, J.E. (2001), ‘Rethinking macroeconomics: what failed and how to repair it’, Journal of European Economic Association, 9, 591–645. Streek, W. and Yamamura, K. (eds) (2001), The Origins on Non Liberal Capitalism: Germany and Japan, Ithaca, NY and London: Cornell University Press.

9

The cognitive explanation of economic behavior: from Simon to Kahneman Massimo Egidi

9.1 INTRODUCTION: ARE ‘RATIONAL AGENTS’ CONSCIOUSLY RATIONAL? The ‘marginalism’ debate emerged in the late 1930s after several studies – one by the Oxford Research Group (see Hall and Hitch, 1939) and another by R.A. Lester (1946) had argued that, empirically, it was not apparent that entrepreneurs followed the marginalist principles of profit maximization/cost minimization in running their firms. In particular, they found that many firms [were] conducting ‘full cost’ pricing rules and routines and [moreover] that predicted falls in employment as a result of higher wages were not evident. They consequently questioned the relevance of the profit-maximization assumption in Neoclassical theories of the firm. (History of Economic Thought website, http://www.newschool.edu/nssr/het/ essays/product/Maxim-2.htm#old)

The debate elicited by the Oxford Research Group was related to a conceptual difficulty of marginalistic approach: the question if it is reasonable to believe that entrepreneurs, and more generally economic subjects, follow the profit-maximization principle in running their affairs, provided that maximization may require specific knowledge and high computational complexity. We shall briefly sketch this debate later, but it is useful to anticipate an essential point raised by Harrod (1939) in his defence of marginalism. He responded to the criticisms by claiming that profit maximization was not observed in many firms partly because the information necessary for such calculations was hard to obtain. But, he added, the ‘best’ decisions would nonetheless arise from a process of ‘natural selection’. In this way Harrod introduced an evolutionary justification of rationality, to overcome the conceptual difficulty of attributing to economic actors the ability to dominate the complexity required to achieve the optimal economic decision. The gap between the subjects’ competence and the complexity of the economic problem (Heiner, 1983) was recognized and a solution was tentatively advanced. But for many decades the problem remained as one of the less clarified within the marginalistic approach and it is still partially opaque. 183

184

Handbook of knowledge and economics

The best-known evolutionary explanation of rational behavior, after Harrod’s proposal, was provided by Milton Friedman. According to Friedman, although individuals do not possess the formal tools with which to calculate the optimum adequately, they behave as if they do – like bicycle riders who keep themselves in dynamic equilibrium even though they are unaware of the complex equations of the dynamics of motion, or billiard players who accomplish complex trajectories with their billiard balls although ignorant of the laws of rational mechanics. Friedman expressed skepticism about the possibility of discovering how business decisions are made through observation of individuals’ explicit thought and behaviors, suggesting that individuals might actually not be aware of the mental processes involved in their actions. This observation was related to the idea that a part of relevant knowledge may be tacit, and it led Friedman to the extreme position of prescribing that the individual expression of preferences must be disregarded. The ‘as if ’ hypothesis was then supplemented with the further assertion that individual preferences are not observable, and indeed that they are irrelevant to proving the validity of an economic theory (Friedman, 1953). According to the ‘as if ’ hypothesis, the theory of rational choice indicates to economic actors how best they can achieve their goals, and it is assumed that those who fail to conform will be gradually excluded by a process of selection that permits only ‘fully rational’ operators to survive. On this view it is therefore both pointless and uninteresting to investigate the psychological aspects of decision making, because a low degree of awareness of individuals was not supposed to be incompatible with full rationality. In fact Friedman’s evolutionary explanation, if successful, had permitted the conduct of empirical study of human behavior disregarding any resort to a psychological experimentation because the explanation of rationality should be fully attributed to the economic forces of selection and competition. But his approach reveals many conceptual failures, partially due to extrapolation of Harrod’s view beyond the limits of his original proposal; in particular, we shall see that Friedman’s justification of optimal behaviors is not appropriate, because an evolutionary approach may consistently explain the emergence of the ‘relatively best’, generally sub-optimal, behaviors. This means that an evolutionary explanation of economic behavior  and  of  the cognitive gap may be based on a bounded rationality approach to decision making. For this reason we shall explore the line of thought that moves from bounded rationality to cognitive economics, partially crossing the evolutionary approach. As we shall see, the

The cognitive explanation of economic behavior 185 solution offered along this line is related to the question of consciousness, and ultimately to the distinction between tacit and explicit knowledge. This distinction may in fact be cognitively grounded by referring to the distinction between conscious and automatic reasoning. In problem framing and solving human beings use both these form of reasoning, the conscious and effortful process of explicit reasoning and the automatic and effortless process of intuitive reasoning. Automaticity and the use of procedural memory govern unconscious reasoning processes, as Luchins (1942), Schneider and Shiffrin (1977), Shiffrin and Schneider (1977), Bargh and Chartrand (1999), Reber (1996) and other cognitive psychologists have clarified; therefore the interaction of automatic and deliberate thinking may provide a cognitive foundation of the March–Simon distinction between routinized and innovative behavior, as suggested by Daniel Kahneman in his Nobel Lecture. According to Kahneman, the crucial element in understanding how automatic processes interact with deliberate mental processes is ‘accessibility’, an element that can help to respond to the problem of the gap between the limited human mental skills and the high complexity required by economic decisions. The following pages are dedicated to exploring the possibility of a cognitive foundation of economic behavior, to explain the competence gap and the connections with an evolutionary approach to the problem.

9.2 RATIONAL CHOICE AND PSYCHOLOGY OF CHOICE: AN UNRESOLVED DUALISM While the question of relation between rational action and behavior dates to Marshall, to observe this debate at his full maturity we must return to the ‘golden age’ of standard rationality theory: the 1950s. The success achieved by linear and dynamic programming in those years seemingly justified unlimited faith in the possibility of optimization models to explain all economically significant forms of behavior. There was a widespread conviction that it was invariably possible and justifiable to reduce macrophenomena to rational forms of behavior and to represent rational forms of behavior as problems of constrained maximization. Yet, as the model of rational decision making became increasingly well defined, so there was a corresponding extension of its domain of application – an extension that led to a growth of computational complexity and to advancements in the creation of new, sophisticated optimization algorithms.

186

Handbook of knowledge and economics

This raised the problem of whether it was legitimate to ascribe to individuals the ability to perform extremely complex decision-making processes, resolving the problems connected with them by means of highly time-consuming and sophisticated algorithms, or whether models of rational behavior should only be interpreted in a normative sense as techniques aiding decision making and suitable for use by experts, not by common decision makers. It was this dilemma that prompted Simon to advance his hypothesis of ‘bounded rationality’ and to dispute the idea of perfect and allencompassing rationality. However, in those years a different solution of the dilemma was proposed by Milton Friedman, a solution that was highly successful and provided a (fallacious) point of reference for mainstream theory.

9.3 THE SEPARATION OF PSYCHOLOGY AND ECONOMICS The normative approach to decision-making theory, considered by Friedman as fully compatible with the ‘as if ’ hypothesis, was limpidly expounded in Lionel Robbins’s Essay on the Nature and Significance of Economics (1932), in which he defined economics as the ‘science of choice’. In this approach, ‘calculation’ is therefore totally independent from individual mental activities, and it takes place irrespective of the mental processes of single individuals. The role of rational decision-making theory is viewed as being fundamentally normative – a view shared by the vast majority of economists for just under a century on the assumptions and the definitions provided by Robbins. He codified a view in which economics and psychology are fully autonomous disciplines with independent scientific statutes. As pointed out by Schumpeter, this separation came about only after many decades of heated debate: In principle, utility, be it total or marginal, was considered a psychic reality, a sensation that became evident from introspection, independent of any external observation . . . with directly measurable proportions. I believe this was Menger and Böhm-Bawerk’s opinion. (Schumpeter, 1954, p. 1060)

As is well known, the essential step to transform utility theory into a normative one was made by Pareto, who propounded an axiomatic foundation of decision making. His approach gained general consensus, and in a few years most economists shared the opinion that utility theory ‘has

The cognitive explanation of economic behavior 187 a much better claim to being called logic of choice than a psychology of value’ (ibid., p. 1058). In the Theory of Games and Economic Behavior, published in 1944, von Neumann and Morgenstern took a further step forward. They set out an axiomatic approach to the theory of decision making in conditions of uncertainty, by formalizing the expected utility hypothesis two centuries after Bernoulli’s original definition of it. Debate on the notion of expected utility ensued and lasted for over a decade. It encompassed a number of controversies connected with the confusion generated by the epistemology of the Chicago School. In 1952, Friedman and Savage published their famous study on expected utility in which they constructed an expected utility curve that, they claimed, provided a reasonably accurate representation of observable behavior at the aggregate level. As said, Friedman and Savage’s approach considered the individual’s expression of preferences to be irrelevant. Consequently, their method did not suggest empirical control for individual preferences: on the contrary, it imposed a priori restrictions on the expected utility function based on characteristics relating to the behavior of large aggregates of individuals. For instance, the fact that numerous middle-to-low-income citizens are ready to risk small sums of money on gambling implies that they are risk takers; analogously, the fact that those same citizens take out insurance means they are risk averse. The former property requires a convex utility function, while the latter requires concavity. In order to account for both these features of the population’s behavior, Friedman and Savage suggested that the (aggregate) expected utility curve must have an ‘S’ shape for middle-to-low-income values. Yet Friedman constructed a general shape for the curve without testing the characteristics on a real population: in fact, he neither considered empirical data on insurance, nor made reference to reliable data on gamblers’ incomes, which did not even exist at the time. Friedman and Savage’s study seemed to be a considerable theoretical achievement in regard to definition of the notion of utility. Yet this advance was based on an untenable general epistemological approach that was unfortunately successful and for long remained an unquestioned dogma for a vast number of economists. As we shall see in the next sections, both the ‘as if ’ assumption and the methodological prescriptions on how to connect axiomatic decision theory with empirical data were successfully challenged by the new approach that emerged with Allais’s criticisms.

188

Handbook of knowledge and economics

9.4 MAURICE ALLAIS’S FALSIFICATIONIST APPROACH TO AXIOMATIC DECISION THEORY Maurice Allais’s research pointed to conclusions the reverse of those obtained by Friedman and Savage’s approach. He carried out experiments on individual preferences – using an ingenious falsificationist method – that showed systematic failures in the theory’s predictions. In 1952, at a symposium held in Paris, Allais presented two studies in which he criticized the descriptive and predictive power of the ‘American School’s’ choice theory, and especially Friedman’s version of it (Allais, 1953). He submitted experiments in which subjects faced with alternative choices in conditions of risk systematically violate the assumptions of the expected utility theory. His investigation methodology overturned the prescriptions imposed by the Chicago School because it was founded on observation of an individual’s behavior and introduced an experimental method whereby the inherent difficulty of direct observation of individual preferences could be overcome by cross-checking alternative choices. The experiments proposed by Allais had the following two distinctive features. First, the properties of choice that characterize the expected utility function must be identified in axiomatic form; these properties are: completeness, transitivity, continuity and independence. Second, subjects are presented with pairs of binary choices selected in such a way that one combination of the answers involves the violation of at least one of the axioms. On this ground Allais showed that a large part of the subjects exposed to binary choices violated some axiom of expected utility. An outline of one of the best-known experiments follows (see Box 9.1). Note that if Savage’s postulate is correct, the preference A > B (‘A > B’ means ‘A is preferred to B’) should entail C > D.1 But the experiment contradicts this prediction: What one finds, however, is that the pattern for most highly prudent persons, the curvature of whose satisfaction curves is not very marked, and of who are considered generally as rational, is the pairing A > B and C > D. This contradicts Savage’s fifth axiom. (Allais, 1979, p. 89)

Violations like the one just shown could be interpreted as signaling inconsistency in the system of individuals’ preferences. A natural reaction to the discovery of these violations is to suppose that inconsistencies are not systematic; and it was perhaps for this reason that the initial reaction to Allais’s experiment results was lukewarm. Many believed that his example

The cognitive explanation of economic behavior 189

BOX 9.1 ALLAIS PARADOX Do you prefer Situation A to Situation B? Situation A Certainty of receiving 100 million (francs) Situation B A 10% chance of winning 500 million, an 89% chance of winning 100 million, a 1% chance of winning nothing Do you prefer Situation C to Situation D? Situation C An 11% chance of winning 100 million, an 89% chance of winning nothing Situation D A 10% chance of winning 500 million, a 90% chance of winning nothing was an extreme case, not a systematic one, in view of the particularly large sums at stake. Only later, after repeated experiments by Allais with actual modest sums being given to players, did the phenomenon emerge once again and thus had to be recognized as being systematic in character (Camerer, 1995). Since experiments showed a violation of the expected utility theory axioms, it was only natural to suspect that this violation depended on overly stringent characteristics imposed on the definition of the expected utility function.2 Reaction to Allais’s experiments led in fact to the proposal of more sophisticated versions of utility theory in conditions of uncertainty that modified or moderated certain axioms, or generalized their characteristics. Many proposals were put forward, especially from the mid-1970s onwards, and all of them attempted to relax or slightly modify the original axioms of expected utility theory. The most widely known of them are perhaps the ‘weighted utility theory’ (Chew and MacCrimmon, 1979), which assumed a weaker form of the independence axiom; the ‘regret theory’ proposed by Loomes and Sugden (1982); and the ‘disappointment theory’ propounded by Gul (1991). None of them had statistical confirmation over the full domain of applicability (Hey, 1991, par. 6.5; Tversky and Kahnemann, 1987, p. 88).

190

Handbook of knowledge and economics

Therefore this response to Allais’s criticism did not prove successful; rather, it confined the expected utility problem to a very specialized sector and limited its impact on microeconomics. In any case, in scientific circles, both in the area of probability theory and in the field of economic theory of choice, the scope of Allais’s work was not appropriately valued. Arrow (1978) justifies this underestimation by saying that if his study had been published by some of the most important American journals, future developments would have been achieved 30 years earlier. But this did not happen. Only gradually did economists come to recognize the systematic discrepancy between the predictions of expected utility theory and economic behavior; this raised a dramatic and still unsolved question: how to model human behavior in economics in a more realistic way. Modern mainstream economic theory is largely based on an unrealistic picture of human decision making. Economic agents are portrayed as fully rational Bayesian maximizers of subjective utility. This view of economics is not based on empirical evidence, but rather on the simultaneous axiomization of utility and subjective probability. In the fundamental book of Savage the axioms are consistency requirements on actions with actions defined as mappings from states of the world to consequences (Savage 1954). One can only admire the imposing structure built by Savage. It has a strong intellectual appeal as a concept of ideal rationality. However, it is wrong to assume that human beings conform to this ideal. (Selten, 1999, p. 2)

9.5 THE EVOLUTIONARY JUSTIFICATION OF RATIONALITY Maurice Allais’s experiments have two distinct aspects that it is useful to illustrate. On the one hand, his experiments showed that individuals’ behaviors cannot be fully explained in a consistent manner on the basis of expected utility theory. On the other hand, he pointed out the weaknesses of Chicago ‘positive economics’. Friedmann’s ‘as if ’ hypothesis was based on two main pillars: the full rationality of agents and the efficacy of a selection mechanism based on competition. The first was definitely challenged by Allais; in this section we shall see that the second cannot be grounded on an evolutionary modeling, because formal models of evolutionary processes show that evolution by adaptation via trial and error, under competitive pressure, (normally) leads to sub-optimal behavior. This means that an evolutionary approach can assume the process of competition as a mechanism of selection of the ‘fittest’ behavior, without pretending to attribute optimality (full rationality) to this behavior. This in fact was the cautious position of Harrod’s proposal.3 As we

The cognitive explanation of economic behavior 191 recalled in the introduction, the debate on this point arose after the Oxford Research Group had argued that the empirical evidence did not show that entrepreneurs followed the marginalist principles of profitmaximization in running their firms (Hall and Hitch, 1939, pp. 18–19). Harrod responded to these criticisms by providing a first sketch of evolutionary justification, claiming that the ‘best’ decisions would arise from a process of ‘natural selection’. New business procedures would then be analogous to new mutations in nature. Of a number of procedures, none of which can be shown either at the time or subsequently to be truly rational, some may supplant others because they do in fact lead to better results. Thus while they may have originated by accident, it would not be by accident that they are still used. For this reason, if an economist finds a procedure widely established in fact, he ought to regard it with more respect than he would be inclined to give in the light of his own analytic method. (Harrod, 1939, p. 7)

Fritz Machlup (1946, 1947) and George J. Stigler (1946) joined the debate to defend the marginalist principle. But the best-known evolutionary view was put forward by Armen Alchian (1950, 1953), who argued that the neoclassical theory of the firm was not about firms as such but industries. Alchian maintained that individual firms essentially followed routinized procedures (as Harrod claimed), but it was the industry that adhered to the marginalist principles. Therefore Harrod’s suggestion prudently maintained that evolution would select the (local) best, not necessarily the optimal behavior, while Friedman’s ‘as if ’ approach claimed an evolutionary justification of full rationality. An important contribution to the evolutionary approach was made by Nelson and Winter, who deepened the distinction between routinized and innovative behavior and, by resurrecting Harrod’s view, argued that the evolution of organizations does not necessarily lead them to optimality (Nelson and Winter, 1982; Winter, 1964, 1975, 2005). Their position was reinforced by the formal demonstration that evolution by mutation produces sub-optimal configurations in which a system may stay locked, provided by S. Kauffman in his NK model (1989). Kauffman’s basic idea was that evolution is a process of collective problem solving undertaken by organisms in their environment: the evolution of an organism, or in general of a complex biological system, is guided by its ‘fitness’, that is, by its reproductive success in the environment. The characteristics that determine the fitness of an organism can be represented in a discrete space because they are a set of ‘traits’ that can assume different values.

192

Handbook of knowledge and economics

In Kauffman’s original approach ‘traits’ can be proteins or genes, each of which can assume different ‘configurations’ or ‘values’ (alleles). An organism is characterized by N traits, each of which assumes a given value. A mutation is nothing other than a change in the value of a trait (allele). Consequently, to explore the effect of single mutations on the organism’s fitness we must change the values of the traits one at a time. A crucial property of the traits is ‘epistasis’: when a mutation is introduced, it normally happens that the effect on the organism’s fitness depends on the values of other traits. Call K the average number of genes that contribute to the fitness variation of the organism when a mutation occurs, that is, the average number of epistatic interactions. K may vary from K = 0 (total independence) to K = N−1 (total interdependence). In the former case (K = 0), the effect of a mutation on the fitness depends solely upon the single gene that is affected by the mutation; therefore by comparing the different effects of different mutations on the same gene, we can discover the allele that produces the higher increase in the fitness. If we sequentially make the same comparison for all the genes, we can discover for every gene the alleles that make the best contribution to the fitness. We can thus increase the fitness of the organism until its maximum value by acting on each gene independently. This means that an organism with zero epistatic interactions may achieve an optimal configuration in response to a sequence of random mutations. The assumption that each gene contributes to overall fitness independently of all other genes is clearly an idealization. In a genetic system with N genes, the fitness contribution of one or another allele of one gene may often depend upon the alleles of some of the remaining N−1 genes. Such dependencies are called epistatic interactions. (Kauffman, 1989, p. 539)

Kauffman shows that as the epistatic interaction grows, the number of local optima increases, and an organism affected by mutations may remain trapped once it has reached a local optimum. Again we have an explanation of why complex systems, which can metaphorically represent individual or organizations, may remain trapped in sub-optimal configurations (see also Frenken, 2006). Kauffman’s model therefore provides important support for Nelson and Winter’s views in so far as they emphasize that market mechanisms may not be able to select the best organizational structures and, again, that inefficient firms may survive in the long run. So far we have achieved two distinct strong criticisms of the ‘as if ’ hypothesis: on the one hand the falsifiability of the axioms, which implies serious ‘débacle’ in the predictive power of the theory (Allais). On the other, the persuasion that evolution does not necessarily select optimal behavior: adaptation via mutations

The cognitive explanation of economic behavior 193 by trial and error can be seriously considered as the candidate process to explain economic behavior, but not to explain fully rational economic behavior. Adaptation is on the contrary coherent with a bounded rationality approach. With Allais’s experiments and Kauffman’s modelization of evolutionary processes, we have both empirical evidence and formal models to show that in a large set of conditions there is a systematic discrepancy between real economic behavior and the prescriptions of the traditional theory. Therefore we must explore a parallel route: the evolutionary explanation not based on the theory of rational decision making but on the theory of bounded rationality. The most promising direction along this path is developed within the cognitive approach to economics, an approach that substitutes pure rationality with cognition and examines in depth the psychological and cognitive aspects of decision making; Simon can be considered the initiator of this view, so we turn to his analytical position.

9.6 A PARALLEL CRITICISM: SIMON’S BOUNDED RATIONALITY In the same period of Allais’s criticism of rationality the decision-making model was about to be seriously questioned from another viewpoint and a different context: that of administrative and managerial behavior, which had hitherto defied economic analysis, despite the fact that rational planning analysis within organizations was highly developed. The success attained by optimization methods in the 1950s brought out two critical aspects: on the one hand the extreme sophistication of many optimization models made it impossible for them to be applied in most organizations; on the other, it was becoming clear that the amount of calculation needed to obtain an optimal solution could in some cases be insurmountably high. It was within this context, and in light of empirical observations on how organizations function, that the limits to the individual ability to make rational calculations became evident. The theory of bounded rationality can be traced back to Herbert Simon’s work at Carnegie-Mellon with Dick Cyert, Jim March and Harold Guetzkow at the beginning of the 1950s. Their research program dealt with realistic issues of economic organizations in a period when the conceptual apparatus available was entirely inadequate to the purpose. The group examined methods for the control of decisional processes within companies. It did so from within the selfsame companies analyzed and was thus able to conduct in-the-field appraisal of the behavior of managers and employees. The radical revision of the two notions of rationality

194

Handbook of knowledge and economics

and organization that characterizes Simon’s theory, compared with the neoclassical tradition, originated from the extraordinary interaction of this group. In Models of my Life (1991) Simon recalls that the group gradually altered the language of discussion by introducing ‘semantic changes’ that shed new light on the themes discussed: the notions of ‘bounded rationality’, ‘satisficing’ and ‘problem solving’ were thus developed in a context of highly interdisciplinary analysis. The limitations of the theory available to the group at the time were evident: the Weberian analysis of rationality and bureaucracy that had enjoyed such prolonged success revealed its shortcomings when applied to the behavior of managers founded on the ability to solve problems and innovate in ever-changing situations. The traditional approach, as said, was centered on decision making as a choice; construction of the decision context was considered secondary, and so was the discovery of alternative strategies. The shift of attention was due in part to the effects of consumption theory, for in this theory consumption alternatives are assumed to be normally known by subjects; the only significant problem is the choice of the consumption plan that maximizes expected utility, bearing in mind the limits on the availability of funds. Matters change entirely, however, when the same scheme is applied to contexts of production and organization. In this case, decisions are taken in an environment where it is extremely difficult, and at times impossible, to evaluate all the available alternatives and their consequences. Exploration of this world reveals that the decision is nothing but a final act of a complex process that precedes it, and through which the relevant information is gathered and the appropriate knowledge is structured. By introducing the notion of bounded rationality, Simon picked up on both of these properties of the decision-making process. He argued that the real restriction on a rational decision was the need to construct the context of the decision. To do so, individuals must search for all the relevant information and then construct a ‘mental model’ representing the decisional context. The difficulty of fully representing the latter and of organizing an appropriate mental representation of it marks the bounds of rationality. Simon initiated a new research strategy in order to uncover the secrets of human cognition. He took up one of Turing’s central statements – if a problem can be clearly described with appropriate language, then it can be transferred into a form computable by a machine – and began to build artifacts of artificial intelligence. The artificial reproduction of certain aspects of human problem solving was a new strategy with which to understand the human mind; and the writing of computer programs that made it possible opened the doors to ‘artificial intelligence’. Simon worked in parallel on giving strong impetus to the empirical analysis of cognitive

The cognitive explanation of economic behavior 195 processes. His starting point was analysis of the game of chess, which Simon explored extensively from a theoretical as well as experimental viewpoint. His examination provides a striking example of the complexity involved in constructing good, or ‘satisficing’, strategies. The game of chess focused researchers’ attention on the question of complexity and on the limits of mental calculation. But Simon moved beyond the notion of calculation by first introducing the idea of ‘symbolic manipulation’ and then directly considering the determinants of cognitive processes (reasoning, categorizing, chunking etc.). Experiments applying ‘protocol analysis’ were carried out: verbal ideas expressed during the game by players were taken down and analyzed in order to understand the cognitive processes involved. Problem solving emerged as one of the crucial aspects of the players’ mental activity. This analysis pioneered by Simon showed that players’ mental activity systematically violates rational choice: chess strategies are intertemporal decisions that require players to elaborate and re-elaborate their analyses; their decisions are based on a process of learning and mental model building repeatedly at odds with perfect rationality. Moreover, by opening up the new field of artificial intelligence, problemsolving theory and the connected experiments using protocol analysis made it clear that, despite the great progress achieved, the limit of the ‘artificial’ imitation of the players’ mental activity was that it captured only the ‘explicit thinking’ (i.e. the deliberate mental processes) accessible through introspection. The 1960s were therefore the years of the greatest challenge to the axiomatic foundations of rational choice. On the one hand, Allais’s critique aroused renewed interest in psychology; on the other, Simon made clear that if human intelligence was to be thoroughly understood it had to be ‘decomposed’ into its many complex processes and elements: induction, reasoning and problem solving were, in Simon’s view, the true protagonists in comprehension of human bounded rationality, and consequently yielded a more realistic picture of economic and organizational phenomena.

9.7 THE 1980S: REVISING THE PARADIGM Kahneman and Tversky’s approach differed crucially from the attempts of many scholars to respond to Allais’s experiments by slightly modifying the axioms of expected utility theory; rather, they restructured the problem by concentrating on the mental processes involved. The approach fit coherently within the analytical frame of bounded rationality, as the two authors explicitly acknowledged.

196

Handbook of knowledge and economics

BOX 9.2 FRAMING EFFECT (1) Problem 1 Assume you are $300 richer than you are today. Choose between: A – the certainty of earning $100 B – a 50% probability of winning $200 and a 50% one of not winning anything Problem 2 Assume you are $500 richer than today. Choose between: C – A sure loss of $100 D – A 50% chance of not losing anything and a 50% chance of losing $200

In 1987, Tversky and Kahneman moved beyond Allais’s experiments to show that when individuals take risky decisions, they exhibit a systematic inconsistency related to the framing of the decision. Tversky and Kahneman ran the following experiment, which clearly elicited this effect (see Box 9.2). Readers responding to the two problems will probably opt for the adverse risk option in Problem 1, and therefore choose an earning that is assured (Answer A). And this was the choice made by the vast majority of participants in the experiment. Instead, the answer preferred in Problem 2 will probably be the one in favor of risk taking, and therefore Answer B. It was noted that the majority – who picked answers A and D – violated the theory of expected utility (the independence axiom of the theory), as in Allais’s experiments. Simple reflection shows that, in terms of expected utility, the two problems are the same problem; in fact, the entity’s available wealth was considered after the choice had been made (see Box 9.3). Therefore a large majority of individuals behave as risk takers when confronted by a problem presented in terms of loss (Problem 2) while they behave as risk averse when the same problem is presented in terms of gain (Problem 1). This behavioral inconsistency is called the ‘framing effect’ and demonstrates that the representation (framing) of a problem may be crucial in ‘ordering’ the preferences. Moreover, Tversky and Kahneman observed that the path of preferences observed in the two problems are of particular interest as they violate not only the theory of expected utility, but practically all choice models based on other normative theories. It must be noted that these data are

The cognitive explanation of economic behavior 197

BOX 9.3 FRAMING EFFECT (2) Problem 1 Case A 400 with prob = 1 Case B 300 with prob = 0.5 or 500 with prob = 0.5 Problem 2 Case C 400 with prob = 1 Case D 300 with prob = 0.5 or 500 with prob = 0.5

not consistent with the ‘Regret’ model proposed by Bell (1982) and Loomes and Sugden (1982) and axiomatized by Fishburn (1982). (Tversky and Kahneman, 1987 p. 75)

We again have confirmation that extended utility axiomatic theories – which explain violations of the expected utility theory by introducing new axioms or reducing the original ones – cannot receive experimental confirmation in all spectrums of experiments conducted to date. On the contrary, numerous experiments have confirmed the abovedescribed framing effect. Kahneman and Tversky suggest that in order to understand a decision one must thoroughly analyze the cognitive processes that underlie it. It is thus necessary to examine how people represent problems; how the complex process of editing is carried out; and how mental models are built in order to make a particular decision. A suggestion closely related to Simon’s analysis of the decisional problem considers the decision to the final act in a problem-solving process. Problem solving is the core of a subject’s activities in taking a decision, and it is described as a calculation made to find a (satisficing) strategy. This ‘calculation’ is carried out under strong restrictions imposed by the cognitive limitations of individuals and which may generate systematic biases: that is, systematic deviations from the results that would be obtained by a ‘hyperrational’ subject, an omniscient calculator of unlimited power. The ‘guidelines for future researches’ emerging from the bounded rationality approach and from Kahneman and Tversky’s results suggest that an explanation for biases and deviations from ‘Olympian’ rationality should be sought by conducting deeper exploration of the cognitive processes lying behind decisions. This approach opens up a broad field of new experimental and theoretical analysis that will be briefly sketched later.

198

Handbook of knowledge and economics

9.8 LOCK-IN: THE PERSISTENCE OF MANY SUB-OPTIMAL SOLUTIONS IN HUMAN PROBLEM SOLVING The cognitive explanation of the limits of rationality originates from the pioneering analysis of strategic situations conducted by Simon in the 1950s. In 1956, Cyert, Simon and Trow carried out an empirical study of managerial decisions that revealed an evident ‘dualism’ of behavior: on the one hand, there was behavior guided by a coherent choice among alternatives typical of structured and repetitive conditions; on the other, behavior characterized by highly uncertain and ill-defined conditions, where the predominant role was played by problem-solving activities.4 Decisions in organizations vary widely with respect to the extent to which the decision-making process is programmed. At one extreme we have repetitive, well-defined problems (e.g., quality control or production lot-size problems) involving tangible considerations, to which the economic models that call for finding the best among a set of pre-established alternatives can be applied rather literally. In contrast to these highly programmed and usually rather detailed decisions are problems of non-repetitive sort, often involving basic long-range questions about the whole strategy of the firm or some part of it, arising initially in a highly unstructured form and requiring a great deal of the kinds of search processes listed above. (Cyert et al., 1956, p. 238)

The core of the decision-making process is therefore the activity of searching. The conditions for application of standard choice theory are largely lacking because preference orderings are highly incomplete, decisions are simultaneously inconsistent, and choices are largely ineffective in relation to the goals pursued. The most important part of the process is driven by the ability of the subjects to formulate and solve new, unexpected and illdefined problems. The notion of bounded rationality therefore assumed increasing importance in field studies on team decisions under ill-defined conditions within organizations. These studies induced March and Simon to radically rethink the traditional idea of ‘planning’. They turned the notion of planning, based on the notion of optimal intertemporal decision making, upside down and substituted it with the notion of ‘organizational learning’, a process of collective problem solving that essentially involves the revision and correction of the procedures to accomplish goals. The most important feature of organizational problem solving is the ability of teams to revise their solutions and remedy the errors that they may have committed. This is the most revolutionary aspect of the recasting of the traditional view of organizational activities put forward in March and Simon’s celebrated book Organizations

The cognitive explanation of economic behavior 199 (1958). Here, the notion of organizational learning is expounded with clarity, and the description of organizational decisions is realistically rooted in the notion of organizational conflict; the conflicting views of individuals in the same organization are considered to be the engine of the organizational learning. A few years later, in A Behavioral Theory of the Firm (first published 1963), Cyert and March showed that organizational learning is a highly path-dependent process, and moreover that it is strongly biased by the constant emergence of erroneous beliefs and solutions. when an organization discovers a solution to a problem by searching in a particular way, it will be more likely to search in that way in future problems of the same type; when an organization fails to find a solution by searching in a particular way, it will be less likely to search in that way in future problems of the same type. Thus, the order in which various alternative solutions to a problem are considered will change as the organization experiences success or failure with alternatives. In a similar fashion, the code (or language) for communicating information about alternatives and their consequences adapt to experience. Any decision-making system develops codes for communicating information about the environment. Such a code partitions all possible states of the world into a relatively small number of classes of states. Learning consists in changes in the partitioning. In general, we assume the gradual development of an efficient code in terms of the decision-making rules currently in use. Thus, if a decision rule is designed to choose between two alternatives, the information code will tend to confine all possible states of the world into two classes. If the decision rules change, we assume a change in the information code, but only after a time lag reflecting the rate of learning. The short-run consequences of incompatibilities between the coding rules and the decision rules form some of the most interesting long-run dynamic features of an organizational decision-making model. (Cyert and March, 1992, p. 174)

To some extent, the emergence of erroneous behaviors is taken to be a natural outcome of the ‘imperfections’ and limits of human rationality. Throughout their analysis, March and Simon maintain that shortcomings and errors in organizational planning are embodied in the nature of human decision making – a view that induces them on the one hand to explore the limits of individual rationality, and on the other to find evidence of organizational errors.

9.9 THE EXPLANATION OF BIASES AS RESULTING FROM CONSTRAINTS ON SEARCH PROCESSES DUE TO COMPUTATIONAL COMPLEXITY In a series of papers now considered classics (1950–79), Simon explored the decision-making process from both the theoretical and experimental

200

Handbook of knowledge and economics

viewpoints. Taking the behavior of chess players as a benchmark for understanding the limits on the capacity of humans to discover strategies, he modelled a player’s search for a solution as exploration within the tree of alternatives: errors in this context are generated by the need to simplify the search by pruning most of the branches of the tree, so that the number of game configurations requiring exploration can be drastically reduced. Players need to simplify the game’s strategic representation in order to dominate the problem mentally. ‘Simplifying’ the search by deleting large parts of the tree a priori generally leads to discovery of a non-optimal strategy. This strategy would lose against a perfectly rational opponent, but because perfect rationality is computationally unachievable (the number of states to be explored to find the winning strategy exceeds the available memory of any human being), both players construct an imperfect, sub-optimal strategy, and both commit systematic errors. The winner is the player whose errors are less important than those of his opponent. In a more general context of problem solving under conditions of complexity, decomposition of problems into sub-problems is one of the heuristics most widely used by human beings to achieve a solution. The decomposition is recursively applied to each sub-problem until elementary and easily solved sub-problems are identified. The simplicity of an elementary sub-problem may enable the player to discover the optimal strategy with ease. This is apparently the key to reducing the complexity of the original problem, and it is frequently used by problem solvers. Unfortunately, however, the discovery of all optimal solutions to the sub-problems does not yield the optimal global solution to the original problem. In fact, it is possible to show that the conditions under which the global solution (composed by the optimal sub-problems solutions) is optimal are very restrictive, and consequently most decompositions patterns lead to sub-optimal global solutions (Egidi, 2004). The origin of this unexpected property can be understood if we consider puzzle solving. Here, players must build a simplified representation of the game in order to discover a (boundedly rational) strategy. The optimal solution to puzzles consists in the shortest path from the starting configuration to the goal. Given the enormous number of game configurations that must be analyzed to get an optimal solution, in order to obtain a simple representation of the solution, players classify the states of the puzzle into a relatively small number of categories: large sets of game configurations are therefore aggregated into few categories. These categories are the result of a process of abstraction and classification based on the salience of symbolic features of the configurations of the game. In the case of Rubik’s Cube, for example, the arrangement of the colors

The cognitive explanation of economic behavior 201 of the tiles along one, two or more corners are salient elements with which to categorize classes of configurations that are supposed to come progressively closer to the final configuration. These categories allow the identification of sub-goals into which the original puzzle is decomposed. For example, one of the most popular strategies for solving the cube is based on a procedure by which players as a first step must form a cross on the top of the Rubik’s Cube so that the colors of the edge cubes match the colors of the center cubes. To make the cross, players must sequentially move a first corner, the central tile on the top face, the other three corners facing the same top face, into the right positions. Each sequential position of the corners is mentally represented as a ‘class of configurations’ because it is defined by the positions of a limited number of tiles, while the positions of all the other tiles are irrelevant. In other words, players consider classes of configurations as elementary building blocks during the search for a strategy. By discovering a procedure, they keep in mind a sequence of actions that connects classes of configurations together, assuming an order among them. Consequently, by simplifying the representation of the game through categorization, players build up an ‘aggregated game’ in which the configurations are aggregated states. They must conjecture the distance from the states to the target: that is, they must conjecturally order the categories in relation to their distances to the final goal. Frequently, however, the order that players conjecture does not hold for all elements of the categories: for example, when solving the Rubik’s Cube puzzle, players may ‘naturally’ believe that the class of configurations ‘four corners in the right place’ is closer to the goal than the class of configurations ‘three corners in the right place’, but this is not necessarily true for all elements of these two categories. Players thus produce relatively simple and abstract representations of the strategy by categorization, but their evaluation of the order among the categories may be inaccurate. Owing to this distortion, the players do not always approach the goal along the shortest path; on the contrary, at least for some configurations, they achieve the goal by following a tortuous path that, in some steps, gets further away from the goal. The most intriguing aspect of this situation is that, while the paths to the goal generated by a single procedure are optimal for some of the elements (the ‘right’ ones) but not for all of them, when players solve the puzzle for the ‘right’ elements, the optimality of their decisions will be confirmed. They consequently cannot easily perceive that they have made a wrong classification and modify their decomposition pattern accordingly. It is evident that there are different categorizations and decomposition patterns to each problem. Some of them maintain the features of

202

Handbook of knowledge and economics

the original problem, but the vast majority do not, and this generates decision biases. The set of all different decomposition patterns may be represented in a ‘landscape’ with different levels of ‘fitness’, that is, with different degrees of sub-optimality (see Section 9.7). In other words, when a player constructs one simplified representation of the problem by identifying some basic categories by which to describe the game, he discovers one point in the landscape. To move to a different point, that is, to discover a different representation of the problem, the player must redefine some of the categories, or some relations among them. Learning a new game strategy is a process of redefining categories, and it is consequently a cognitively effortful process. This explains why it is so difficult to discover more efficient strategies of a game, and why players may remain locked into a sub-optimal strategy. Experiments in puzzle solving (Egidi, 2004) confirm the stability of sub-optimal representations, giving further support to Cyert and March’s original explanation of biases in organizational behaviors.

9.10

THE MECHANIZATION OF THOUGHT

So far, we have seen that experimental data on puzzle solving show that, once most individuals have identified a strategy, they are likely to remain anchored to it even though it is not optimal. The first experiment in this domain dates back as far as 1942, when Luchins (1942) and Luchins and Luchins (1950) ran experiments with subjects exposed to mathematical problems for which there were different solutions with different levels of efficiency. The authors showed that when subjects had identified the optimal solution for a problem in a given context, they ‘automatically’ and systematically transferred it to a different context where it was sub-optimal. This process was called ‘mechanization of thought’. Experiments with Target the Two (Cohen and Bacdayan, 1994; Egidi and Narduzzo, 1997) confirm that a similar process is involved in behavior by a team, and in an even more evident and persistent manner: groups of subjects jointly engaged in solving a shared problem may remain even more stably ‘trapped’ in sub-optimal solutions than single individuals. In fact, whilst difficulties encountered by a single subject when solving a problem in a new way depend on whether a new solution can be discovered, and are influenced by cognitive limitations on individual learning, this is all the more the case for a group, because it must find new ways to cooperate in problem solving by devising and adopting an alternative solution jointly.

The cognitive explanation of economic behavior 203

9.11 TACIT VERSUS EXPLICIT KNOWLEDGE A relevant (and only partially explored) aspect of cognitive traps is their ‘stability’: in many contexts, errors and violations of rationality are systematic and persistent. A number of experiments (Camerer et al., 2004a; Tversky, 1977) show, in fact, that when subjects are made aware of biases connected to their choices, they only minimally adjust their behavior. Tversky (1977, p. 201) comments on these findings thus: Daniel Kahneman and I have studied the cognitive processes underlying the formation of preference and belief. Our research has shown that subjective judgments generally do not obey the basic normative principles of decision theory. Instead, human judgments appear to follow certain principles that sometimes lead to reasonable answers and sometimes to severe and systematic errors. Moreover, our research shows (Tversky and Kahneman, 1974; Kahneman and Tversky, 1979) that the axioms of rational choice are often violated consistently by sophisticated as well as naive respondents, and that the violations are often large and highly persistent. In fact, some of the observed biases, such as the gambler’s fallacy and the regression fallacy, are reminiscent of perceptual illusions. In both cases, one’s original erroneous response does not lose its appeal even after one has learned the correct answer.

We have thus far reviewed the explanation of biases and errors in puzzles and games playing based on incomplete representation of the problems, that is, on purely cognitive features of human behavior, and examined the available evidence. As we have seen, Luchins and Luchins show that when subjects have identified the optimal solution of a task in a given context, they automatically transfer it to contexts where it is sub-optimal. Luchins and Luchins’s experiments demonstrate that, once a mental computation deliberately performed to solve a given problem has been repeatedly applied to solve analogous problems, it may become ‘routinized’. Its routinized use enables individuals to pass from deliberate effortful mental activity to partially automatic, unconscious and effortless mental operations. This routinization of thinking is the emergent part of a more complex and general question: to what extent are our actions the effect of a deliberate mental computing activity and, conversely, to what extent is such mental activity accessible through introspection? This was a core question addressed by Kahneman’s Nobel Lecture (2002), where he distinguishes (see Figure 9.1) two modes of thinking and deciding: what he calls ‘intuition’ and ‘reasoning’. Kahneman notes that there is considerable agreement among psychologists on the characteristics that distinguish these two cognitive processes. Following Stanovich and West (2000), he calls them respectively System 1 and System 2.

204

Handbook of knowledge and economics

CONTENT

PROCESS

PERCEPTION

Figure 9.1

INTUITION

REASONING

SYSTEM 1

SYSTEM 2

Fast Parallel Automatic Effortless Associative Slow-learning

Percepts Current stimulation Stimulus-bound

Slow Serial Controlled Effortful Rule-governed Flexible

Conceptual representations Past, present and future Can be evoked by language

Kahneman’s description of cognitive processes

The operations of System 1 are fast, automatic, effortless, associative, and difficult to control or modify. The operations of System 2 are slower, serial, effortful, and deliberately controlled; they are also relatively flexible and potentially rule-governed.

As indicated in Figure 9.1, the operating characteristics of System 1 are similar to the features of perceptual processes. On the other hand, as Figure 9.1 also shows, the operations of System 1, like those of System 2, are not restricted to the processing of current stimulation. Intuitive judgments deal with concepts as well as with percepts, and can be evoked by language. The distinction between System 1 and System 2 was first drawn by Schneider and Shiffrin (1977), who called them respectively ‘automatic’ and ‘controlled’ processes; since then many analogous two-system models have been developed under different names, as discussed in Camerer et al. (2004a, 2004b). The question raised by Luchins and Luchins fully matches this distinction because it shows how a process of reasoning – typically composed of slow, serial and effortful mental operations – comes to be substituted by an effortless process of automatic thinking. Of course, understanding the inverse process is matter of equal importance: the question is how it happens that automatic mental activities are accessible to conscious thinking, and to what extent they may be used in deliberate cognitive activities. While some aspects of these questions have been clarified by neurophysiology, particularly through the use of brain

The cognitive explanation of economic behavior 205 imaging techniques, traditional psychological experiments, too, have furnished a great deal of important information. The crucial element in understanding how automatic processes interact with deliberate mental processes is ‘accessibility’. As Kahnemann emphasizes in his Nobel Lecture, accessibility is a continuum, not a dichotomy: experimental evidence shows that the more a person acquires information and competence in a particular domain, the more he becomes able to recall and use the automatic part of his knowledge. This implies that accessibility has different levels and that some operations demand more mental effort than others. The acquisition of skill selectively increases the accessibility of useful responses and of productive ways to organize information. The master chess player does not see the same board as the novice, and the skill of visualizing the tower that could be built from an array of blocks could surely be improved by prolonged practice. (Kahneman, 2002, p. 453)

The concept of accessibility is related to the notions of stimulus salience, selective attention, and response activation or priming. Following Kahneman again, Physical salience also determines accessibility: if a large green letter and a small blue letter are shown at the same time, ‘green’ will come to mind first. However, salience can be overcome by deliberate attention: an instruction to look for the smaller letter will enhance the accessibility of all its features. (Ibid.)

In parallel with the two types of cognitive processes, psychologists define two types of memorization. On the one hand there is ‘procedural memory’, which is automatic, unconscious or non-conscious, and is reflected in our actions. Well-known examples of the use of procedural knowledge are driving a car and acquiring grammatical competence: in grammar acquisition, using plurals and the past tense in accordance with grammatical rules is automatic. The same happens in the repeated playing of games, where routinization leads to procedural knowledge (Cohen and Bacdayan, 1994). On the other hand, there is ‘declarative memory’, which is effortful and open to conscious inspection, and it requires symbols. The most evident examples are mathematical operations and symbolization. The distinction between explicit/controlled and automatic/unconscious mental processes has a parallel in the distinction between the ‘constructivist’ and ‘ecological’ orders suggested and analyzed by Vernon Smith. The former order, the ‘constructivist’ one, is related to declarative, controlled processes, while the latter refers to automatic processes. There is an obvious relationship between this view and Hayek’s idea of institutions

206

Handbook of knowledge and economics

as an order not generated by intentional design. I refer to Vernon Smith’s Nobel Lecture, which shows in masterly manner how markets and other economic institutions are pervaded by the interaction between constructivist and ecological rationality (Smith, 2002). The distinction between the two cognitive systems helps us move toward a new, more complex explanation of systematic deviations from pure rationality. Without describing all the consequences of this new view, here I merely point out that recent evidence from the neural sciences further emphasizes the importance of emotional processes (Camerer et al., 2004b). The discussion in the previous sections relates to the limits of rationality due to constraints in System 2, the controlled process of reasoning (based on symbolic manipulations). The processes of editing and building a representation of a problem are seemingly related to the threshold between the two systems, and to their dynamic interaction. In fact, on the one hand we have the ‘routinization’ of thought; on the other, the emergence or ‘elicitation’ of unconsciously memorized items used as elementary building blocks in deliberate reasoning. A clear example is provided by chess. Chess grandmaster performances are based on ability to process and deliberately use more than 20 000 chessboards stored in the procedural memory. To some extent, these items can be considered the basic building blocks in representation of the game and discovery of a satisficing strategy. The difficulty of substituting some or all of these basic items with new ones can, as we have seen, explain the general sub-optimality of strategy building.

9.12 CONCLUDING REMARKS The economic methodology inherited from the Chicago School becomes untenable when it is demonstrated formally that evolution by adaptation leads generally to sub-optimal stable solutions: the ‘as if ’ hypothesis can no longer be considered as an evolutionary justification of perfect rationality; on the contrary, an evolutionary explanation of boundedly rational economic behaviors can still be considered relevant not merely as metaphor but as a useful conceptual approach; one of the interesting features of this view is that the evolution of behaviors can be easily modeled on the basis of the dualistic interpretation of human behavior (routinized and innovative); this interpretation, initiated by Menger, emphasized by Schumpeter and more recently developed by many scholars, may be grounded on the basis of Simon’s approach, which distinguishes routinized thinking from problem-solving activity – a distinction based on strong evidence from experimental psychology (Luchins, 1942; Kahneman, 2002; Reber, 1996).

The cognitive explanation of economic behavior 207 The distinction between two types of cognitive processes – the effortful process of deliberate reasoning on the one hand, and the automatic process of unconscious intuition on the other – can provide a different map with which to explain economic behavior by attributing to human beings cognitive skills as the basic foundation of their boundedly rational behavior. How human actions arise as the joint effect of these two cognitive processes – deliberate and unconscious reasoning – and how institutions function as the joint effect of tacit and explicit knowledge, are becoming to be more tractable problems in the light of the cognitive approach. Moreover, this distinction allows us to explain why individuals can be only partially aware of the ‘rationale’ of their actions, and why a permanent gap exists between consciously elaborated problems and the global knowledge needed to solve it. This leads to a better understanding of why, as suggested by Hayek, institutions are partially the ‘results of human action but not of human design’ (Hayek, 1967).

NOTES 1. Proof: If A is preferred to B, then U(100) > 0.10U(500) + 0.89U(100) + 0.01U(0). Rearranging this expression gives 0.11U(100) > 0.10U(500) + 0.01U(0); and adding 0.89U(0) to each side yields 0.11U(100) + 0.89U(0) > 0.10U(500) + 0.90U(0), which means that C is preferred to D. 2. A similar reaction was provoked by the Ellsberg paradox (Ellsberg, 1961). 3. Even though the most influential evolutionary approach to justifying rationality is Friedman’s, the evolutionary justification was put forward by many important authors before and after Friedman. The evolutionary approach was first formulated by Alfred Marshall, who in articles published in the 1870s sketched a model of the mind and used it to describe the processes by which routines arise, and the mechanism of innovation and creativity within organizations (Rizzello, 2003). However, although Marshall’s evolutionary view attracted many admirers among economists, it remained as a pure metaphor for half a century until the emergence, in the late 1930s, of debate on the realism of marginalist principles. 4. In this last set of conditions, not only must subjects gather information; they must also be able to select the information and knowledge that is effectively relevant to their purposes and to assimilate it into the system of knowledge that they already possess. To do so, they must have a ‘level of competence’ adequate to the situation of their choice; they must, that is, implement skills of learning and problem solving.

REFERENCES Alchian, A.A. (1950), ‘Uncertainty, evolution and economic theory’, Journal of Political Economy, 58, 211–21. Alchian, A.A. (1953), ‘Biological analogies in the theory of the firm: comment’, American Economic Review, 43 (4), 600–603.

208

Handbook of knowledge and economics

Allais, M. (1953), ‘Le comportement de l’homme rationnel devant le risque: critique des postulats et axiomes de l’Ecole Américaine’, Econometrica, 21, 503–46. Allais, M. (1979), ‘The foundations of a positive theory of choice involving risk and a criticism of the postulates and axioms of the American School’, in M. Allais and O. Hagen (eds), Expected Utility Hypotheses and The Allais Paradox: Contemporary Discussions of Decision under Uncertainty with Allais’ Rejoinder, Dordrecht, Boston, MA and London: Reidel. Reprinted in Egidi and Rizzello (2004). Arrow, K.J. (1978), ‘The future and the present in economic life’, Economic Inquiry, 16 (2), 157–70. Bargh, J.A. and Chartrand, T.L. (1999), ‘The unbearable automaticity of being’, American Psychologist, 54 (7), 462–79. Bell, D.E. (1982), ‘Regret in decision making under uncertainty’, Operations Research, 30, 961–81. Bernoulli, D. (1738), ‘Specimen theoriae novae de mensurea sortis’, Commentarii Academiae Scientiarum Imperialis Petropolitanae, 5, 175–92. English translation in L. Sommer (1954), ‘Exposition of a new theory one the measurement of risk’, Econometrica, 22, 23–36. Camerer, C. (1995), ‘Individual decision making’, in J.H. Kagel and A.E. Roth (eds), Handbook of Experimental Economics, Princeton, NJ: Princeton University Press, pp. 587–703. Camerer, C., Loewenstein, G. and Prelec, D. (2004a), ‘Neuroeconomics: why economics needs brains’, Scandinavian Journal of Economics, 106 (3), 555–79. Camerer, C., Loewenstein, G. and Prelec, D. (2004b), ‘Neuroeconomics: how neuroscience can inform economics’, Journal of Economic Literature, 43 (1), 9–64. Chew, S.H. and McCrimmon, K.R. (1979), ‘Alpha-nu choice theory: an axiomatization of expected utility’, Working Paper 669, University of British Columbia Faculty of Commerce. Cohen, M.D. and Bacdayan, P. (1994), ‘Organizational routines are stored as procedural memory: evidence from a laboratory study’, Organization Science, 5 (4), 554–68. Cyert, R.M. and March, J.G. (1992), A Behavioral Theory of the Firm, Englewood Cliffs, NJ: Prentice-Hall. Cyert, R.M., Simon, H.A. and Trow, D.B. (1956), ‘Observation of a business decision’, Journal of Business, 29, 237–48. Egidi, M. and Narduzzo, A. (1997), ‘The emergence of path dependent behaviors in cooperative contexts’, International Journal of Industrial Organization, 15 (6), 677–709. Egidi, M. (2004), ‘Decomposition patterns in problem solving’, Social Science Research Network, http://www.ssrn.com/. Egidi, M. and Rizzello, S. (eds) (2004), Cognitive Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Ellsberg, D. (1961) ‘Risk, ambiguity and the Savage axioms’, Quarterly Journal of Economics, 75 (4), 643–69. Frenken, K. (2006), Innovation, Evolution and Complexity Theory, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Fishburn, P.C. (1982), ‘Nontransitive measurable utility’, Journal of Mathematical Psychology, 26, 31–67. Friedman, M. (1953), Essays in Positive Economics, Chicago, IL: University of Chicago Press. Friedman, M. and Savage, L.J. (1952), ‘The expected-utility hypothesis and the measurability of utility’, The Journal of Political Economy, 60 (6), 463–74. Gul, F. (1991), ‘A theory of disappointment in decision making under uncertainty’, Econometrica, 59 (3), 667–86. Hall, R.L. and Hitch, C.J. (1939), ‘Price theory and business behavior’, Oxford Economic Papers, 2, 12–45. Harrod, R.F. (1939), ‘Price and cost in entrepreneurs’ policy’, Oxford Economic Papers, 2, 1–11. Hayek, F.A. (1967), ‘The results of human action but not of human design’, in F.A. Hayek

The cognitive explanation of economic behavior 209 (ed.), Studies in Philosophy, Politics and Economics, London: Routledge & Kegan Paul, pp. 96–105. Heiner, R.A. (1983), ‘The origin of predictable behaviors’, American Economic Review, 73, 560–95. Hey, J.D. (1991), Experiments in Economics, Oxford: Basil Blackwell. Kahneman, D. (2002), ‘Maps of bounded rationality: a perspective on intuitive judgment and choice’, Nobel Prize Lecture, 8 December, http://www.nobelprize.org.nobel_prizes/ economics/laureates/2002/kahneman-lecture.html, pp. 449–89. Kahneman, D. and Tversky, A. (1976), ‘Prospect theory: an analysis of decisions under risk’, Econometrica, 47, 263–91. Kauffman, S.A. (1989), ‘Adaptation on rugged fitness landscapes’, in D.L. Stein (ed.), Lectures in the Sciences of Complexity, Santa Fe Institute Studies in the Sciences of Complexity, I, Redwood City, CA: Addison Wesley, pp. 527–712. Luchins, A.S. (1942), ‘Mechanization in problem-solving’, Psychological Monographs, 54, 1–95. Luchins, A.S. and Luchins, E.H. (1950), ‘New experimental attempts in preventing mechanization in problem-solving’, The Journal of General Psychology, 42, 279–91. Loomes, G. and Sugden, R. (1982), ‘Regret theory: an alternative theory of rational choice under uncertainty’, Economic Journal, 92, 805–25. Machlup, F. (1946), ‘Marginal analysis and empirical research’, The American Economic Review, 36 (4), 519–54. Machlup, F. (1947), ‘A rejoinder to an anti-marginalist’, The American Economic Review, 37 (1), 148–54. March, J.G. and Simon, H.A. (1958), Organizations, New York: John Wiley. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: The Belknap Press of Harvard University Press. Reber, A.S. (1996), Implicit Learning and Tacit Knowledge: An Essay on the Cognitive Unconscious, Oxford: Oxford University Press. Rizzello, S. (2003), ‘Introduction’, Cognitive Developments in Economics, London: Routledge. Robbins, L. (1932), An Essay on the Nature and Significance of Economic Science, New York: Macmillan (2nd edn 1935). Savage, L. (1954), The Foundation of Statistics, New York: John Wiley and Sons. Schneider, W. and Shiffrin, R.M. (1977), ‘Controlled and automatic human information processing: I. Detection, search and attention’, Psychological Review, 84 (1), 1–66. Schumpeter, J.A. (1954), History of Economic Analysis, London: George Allen and Unwin. Selten, R. (1999), ‘What is bounded rationality?’, SBF Discussion Paper B-454. Shiffrin, R.M. and Schneider, W. (1977), ‘Controlled and automatic human information processing: II Perceptual learning, automatic attending and a general theory’, Psychological Review, 84 (2), 127–90. Simon, H.A. (1991), Models of my Life, New York: Basic Books. Smith, V.L. (2002), ‘Constructivist and ecological rationality in economics’, Nobel Prize Lecture 2002; also in The American Economic Review, 93 (3), 465–508. Stanovich, K.E. and West, R.F. (2000), ‘Individual differences in reasoning: implications for the rationality debate’, Behavioral and Brain Sciences, 23, 645–726. Stigler, G.J. (1946), ‘Professor Lester and the marginalists’, The American Economic Review, 37 (1), 154–7. Tversky, A. (1977), ‘On the elicitation of preferences: descriptive and prescriptive considerations’, in D. Bell, R. Keneney and H. Raiffa (eds), Conflicting Objectives in Decisions, New York: Wiley, pp. 209–22. Tversky, A. and Kahneman, D. (1974), ‘Judgement under uncertainty: heuristics and biases’, Science, 185, 1124–31. Tversky, A. and Kahneman, D. (1987), ‘Rational choice and the framing of decisions’, in R.M. Hogart and M.W. Reder (eds), Rational Choice: The Contrast between Economics and Psychology, Chicago, IL: University of Chicago Press, pp. 67–94. Previously published in The Journal of Business, 59 (4), S251–S278.

210

Handbook of knowledge and economics

Von Neumann, J. and Morgenstern, O. (1944), Theory of Games and Economic Behavior, Princeton, NJ: Princeton University Press. Winter, S.G. (1964), ‘Economic “natural selection” and the theory of the firm’, Yale Economic Essays, 4, 225–72. Winter, S.G. (1975), ‘Optimization and evolution in the theory of the firm’, in R.H. Day and T. Groves (eds), Adaptive Economic Models, New York: Academic Press, pp. 73–118. Winter, S.G. (2005), ‘Developing evolutionary theory for economics and management’, WP 2005-01 Working Paper of the Reginald H. Jones Center, The Wharton School, University of Pennsylvania.

10 Towards a theoretical framework for the generation and utilization of knowledge Pier Paolo Saviotti

10.1 INTRODUCTION Modern industrialized societies seem to be evolving towards the knowledgebased society, a society involving growing knowledge intensity. In fact, this is not a new phenomenon. The origin of the knowledge-based society can be found in the second half of the nineteenth century with the advent of the modern university system and with the institutionalization of industrial R&D (Murmann, 2004; Freeman and Soete, 1997). In other words, a knowledge-based society differs from pre-existing ones because it uses knowledge provided by institutions specialized in its creation and diffusion. In spite of the growing importance of knowledge for economic development, we still have a very limited understanding of processes of knowledge creation and utilization. A theoretical structure allowing us to represent, model and measure knowledge is required. This is not to deny that many valuable contributions to this theme have been made. On the contrary, such contributions exist, starting from the pioneering work of Hayek, Machlup, Simon and so on (see in Foray, 2000). Furthermore, most of the literature on innovation covers the same subject, though under a different name. The point to be made here is that knowledge has rarely been studied directly. Usually some phenomena involving or requiring knowledge, such as innovation, have been studied, giving us very important insights into knowledge itself. Yet the studies have somehow remained without proper foundations since an adequate characterization of processes of knowledge generation and utilization was missing. The object of this chapter is precisely to start to define such a characterization. We can notice right away that this characterization has to be applicable to both fundamental, or basic, knowledge and to the most applied types of industrial knowledge. In other words, it cannot be a characterization of knowledge that satisfies only the historian of science and the epistemologist, or alternatively the economist and the historian of business. Not only it is becoming increasingly clear that advances in industrial applications require advances in basic knowledge, but the degree of interaction between the two is becoming more intense and frequent, thus making 211

212

Handbook of knowledge and economics

these very distinctions increasingly irrelevant. Scholars of innovation have even started defining a second mode of knowledge generation and utilization, called Mode 2, in which fundamental and applied knowledge would continuously interact, thus becoming difficult to separate both chronologically and institutionally (Gibbons et al., 1994). Without entering into this problem we can simply point out here that an adequate theoretical representation of processes of knowledge generation and utilization must be applicable to both fundamental and applied types of knowledge. This approach constitutes the opposite of the process of demarcation in epistemology: while epistemologists in the past were attempting to distinguish true, or scientific, knowledge from less rigorous types, the approach developed in this chapter aims at finding the common properties of types of knowledge ranging from scientific to craft based. In what follows, a general characterization of knowledge usable for the above purposes will first be introduced. This will then lead to an analysis of processes of knowledge production based on well-established economic concepts, such as division of labour, coordination and competition. Finally, processes of knowledge creation and utilization within firms will be shown to be analysable with a conceptual framework compatible with the one described above for knowledge in general.

10.2 SOME CONSIDERATIONS ON THE NATURE OF KNOWLEDGE In this section, two properties of knowledge that are considered to be very general and applicable to any type of knowledge, from scientific to more empirical and craft based, are described. These two properties do not constitute a complete description of knowledge, but, as will be seen, they provide a surprisingly powerful basis to analyse processes of knowledge creation and utilization. The two properties are: (P1) Knowledge is a co-relational structure. (P2) Knowledge is a retrieval or interpretative structure. For brevity, in what follows I shall refer to the external environment as ExtEnv. 10.2.1

Knowledge as a Co-relational Structure

In this chapter it is assumed that a reality independent of human observers exists in the sense that a number of entities constitute the ExtEnv that

Theoretical framework for the generation & utilization of knowledge 213 cannot be modified at will. However, what follows is compatible with a wide variety of positions. The observables and variables that will be the objects of co-relations are not assumed to be ‘true’ entities, but only our mental representations. Depending on the basic assumptions of different scholars, such observables and variables may be real objects in themselves or simply intellectual constructs useful to create theories, but which find no counterpart in nature. In other words, the representation of knowledge described here is compatible with basic assumptions ranging from those of Berkeley (see Losee, 1997, pp. 165–7), who maintained that ‘material substances do not exist’, of Mach (see ibid., pp. 168–70), who shared Berkeley’s conviction that it is a mistake to assume that the concepts and relations of science correspond to that which exists in nature, or with the most naïve realist assumption that observables and variables are real entities. The focus of this chapter is on the production of knowledge as a collective enterprise, and this focus is compatible with the conception of observables and variables both as real entities and as mental representations. In fact, these different positions relate more to what we could call variation, the production of new ideas, than to selection, the testing of these ideas. However, selection will eliminate a number of potential observables, variables and connections. Whatever assumption one makes about the ontological character of observables and variables, it is impossible to formulate mental representations of the ExtEnv at will. The ExtEnv constitutes both a set of resources and a set of constraints for the activity of human beings. Such constraints mean that the ability of human beings to modify the ExtEnv in order to survive is in principle considerable, but limited. There is no magic wand that allows us to obtain particular outcomes at will. Human beings interact with this ExtEnv by means of their sense organs and a series of enhanced sense organs and tools. Initially human beings had to rely only on their sense organs for any observations on the ExtEnv. In the course of human evolution they developed enhanced sense organs (e.g. telescopes, measuring devices, scanners etc.) that allowed them to access parts of the ExtEnv not directly accessible through their primary sense organs. Furthermore, they developed a series of tools (e.g. axes, hammers etc.) that allowed them to modify purposefully their ExtEnv, tools that Georgescu-Roegen (1971) called exosomatic organs. We find here already the basic distinction between two distinguishable but intimately interconnected activities. On the one hand there is a need to observe and to know; on the other hand there is a need to modify the ExtEnv. The activities corresponding to the two needs are clearly separable, at least conceptually, but closely related because it is easier to modify the ExtEnv if we know its structure and properties. In fact, these two needs and the related activities correspond to what we currently call science and technology. Science is

214

Handbook of knowledge and economics

the activity that understands and knows our ExtEnv and technology is the activity that modifies the same environment. Examples of the relatedness of these two activities can be found from very ancient times, for example in the field of navigation. However, until the second half of the nineteenth century such relationships were more occasional than systematic. The situation is very different today in societies characterized by a high intensity of R&D. The distinction between science and technology corresponds roughly to that between ‘to know what’ and ‘to know how’ (Loasby, 1999). To know how allows us to modify our ExtEnv. To know how can be made easier by knowing what happens in the particular subset of the ExtEnv that we intend to modify. However, the knowledge of what of the particular subset is not always available; thus sometimes knowledge of how has to be developed without the knowledge of what of the subset considered. In the extreme case in which no knowledge of what was available the search for appropriate forms of modification of the ExtEnv could follow a trialand-error method. Such a method would be very costly in search activities. However, the number of trials required to design appropriate strategies for the modification of the ExtEnv could be drastically reduced if we had a sound knowledge of the nature of the ExtEnv. Further analysis of this point will be undertaken later in this chapter. Lundvall et al. (2002) use a more sophisticated classification of knowledge types, including ‘know why’ and ‘know who’. In this chapter the distinction between know what and know how is considered more fundamental than the other two. Know why and know who are two categories that, while very useful, are in principle derivable from know what and know how. This close connection between science and technology is considerably enhanced by the truth criterion we normally use. A theory is considered to be ‘temporarily’ correct if its predictions correspond to empirical observations of the ExtEnv. This close connection between the nature of theories and the structure of the ExtEnv lies at the roots of our ability to use theoretical knowledge to modify our ExtEnv. I now proceed to explain what is meant by co-relational structure. We can identify in our ExtEnv a number of observables, that is, of entities responsible for observed phenomena. To each observable we can associate one or more variables that represent and measure different aspects of the observable. As previously pointed out, in this chapter no particular assumption is made about the truthfulness of the observables and of the variables representing them. In other words, we are not assuming the observables to be real entities that can be observed in an unbiased way by human observers. Observables and variables are mental representations (Loasby, 1999) that allow us to explore the ExtEnv and to establish in it a

Theoretical framework for the generation & utilization of knowledge 215 series of constituting entities and structures. All theories are conceived in the space of mental representations. Of course, our mental representations and the theories that are based on them can be generated at will, but not all of them can pass the required tests: correspondence with experimental results differentiates ‘good’ from ‘bad’ theories. In other words, the variety of mental representations that can be constructed is much greater than that of the mental representations surviving empirical tests. The presence of co-relation, or connection, between two variables means that, at least for a number of the properties of the systems of which they represent observables, their behaviour is not independent but linked. We can see that if variables were not linked, our knowledge of the ExtEnv would be far more costly to acquire. If all the entities that constitute the ExtEnv were completely independent, our knowledge would be the sum of what we knew about each entity. The presence of co-relations, or connections, allows us to deduce values of unknown variables from those of known and related ones. Thus the existence of connections reduces our information costs. Typically we detect co-relations by detecting correlations in the behaviour of different variables. The two terms co-relation and correlation are not synonyms: the former indicates a link between two variables within the structure of the ExtEnv while the latter describes the statistical correlation between the properties of the same two variables. In a sense, co-relation, or connection, is the more fundamental of the two terms, because it pertains to the structure of the ExtEnv, but the existence of a co-relation is usually detected by the presence of correlation. We now consider an example of knowledge as a co-relational structure. Example 1. The law of ideal gases PV = nRT

(10.1)

If P is the pressure of a gas, V its volume, n the number of moles of the gas and T its temperature, with R a general constant, equation (10.1) tells us that all these variables are correlated in such a way that if we raise the temperature, the volume and the pressure of the gas have to increase in order for equation (10.1) to continue to be satisfied. In other words, equation (10.1) represents the co-relation of the behaviour of a number of variables of the gas. It is part of a theoretical model, that of ideal gases, in which the atoms or molecules of the gas are considered points occupying zero volume and behaving independently of one another. Such a model is part of a wider theory of gases, and contributes to theories such as thermodynamics or kinetics.

216

Handbook of knowledge and economics

In this case the co-relation takes on a very accurate and quantitative character. In many cases the co-relation provided by a theory can be more qualitative and loose while still being a co-relation. Examples of accurate and quantitative correlations are found mainly in the physical sciences, although they can be found also in the biological and social sciences, even if with a lower frequency. Any scientific law that can be expressed in the form of one or more equations provides examples of an accurate and quantitative co-relation between variables. In spite of the great number of these laws and equations, they do not represent the majority of our knowledge, except in a few fields. A large number of theories of the ExtEnv have to be content with much looser and less accurate co-relations. For example, the so-called Engel’s law (Hirshleifer, 1988, pp. 98–100) states that the share of income spent on basic commodities, such as food, housing and so on, falls as the average income per head increases. Here the co-relation between income per head and expenditures in particular categories of commodities can be detected empirically and measured, but there is no complete theory of Engel’s law. A co-relation (leading to a correlation) is established, but in a less accurate and quantitative way. An even looser, even if very interesting, type of co-relation can be found in Max Weber’s theory linking the ‘Protestant ethic’ and the ‘spirit of capitalism’ (Weber, 1968). The co-relation can be stated in the following form: countries/societies/groups adhering to a Protestant religion have a higher probability of giving rise to a capitalist economic system than nonProtestant ones. It is of no consequence for the objective of this chapter that today many scholars criticize this theory. The point to be made here is not the truthfulness of the theory, but the form in which it is created, and this form is that of a co-relation. The theory correlates in a loose and nonquantitative way religious and cultural beliefs and economic performance. A particular place in this context must be reserved to econometric analysis. Econometric equations provide an example of very accurate correlations, but they do not necessarily allow us to detect co-relations as we can with so-called analytical models. In analytical models of given systems, co-relations, or connections, between entities and variables are assumed as initial hypotheses in order to calculate system properties. The correspondence between calculated and measured values of system properties confirms the existence of the co-relations hypothesized. In econometric models, on the other hand, we find out that some variables are correlated, but we do not attempt to determine the precise nature of the interaction of the basic variables. Thus econometric analysis establishes the presence of a correlation, rather than its precise form (co-relation) as could be done in an analytical model.

Theoretical framework for the generation & utilization of knowledge 217 In summary, the property of being a co-relational structure can be considered a general property of knowledge. This has both limits and many interesting implications. 10.2.2

Knowledge as a Retrieval or Interpretative Structure

According to information theory (Shannon and Weaver, 1949), information does not have meaning, it is purely factual. In this sense information is different from knowledge. Based on the previous characterization, we could say that knowledge detects co-relations between different variables while information is constituted by the numerical values of the variables. However, while information does not in itself carry any meaning, its use requires knowledge of the context in which information was created (see also Cowan et al., 2000). Thus data sets on atomic transition frequencies or on the distribution of some biological populations would not be interpretable by an observer who does not know the relevant theoretical framework. Thus information is not generally interpretable by a nonknowledgeable agent/actor, but any information set requires the knowledge of one or more subsets of the ExtEnv. In this sense knowledge can be considered a retrieval/interpretative structure. The idea of information as totally devoid of meaning is more plausible in cases where the required underlying knowledge is widely available and taken for granted by all members of a community. For example, a train timetable with a list of times of arrival to and departure from particular places is generally interpretable by most of the members of a reasonably well-educated society. However, the smaller the community knowing a particular subset S(ExtEnv), the more non-members of that community will be unable to interpret an information set generated within that community. One of the properties commonly attributed to knowledge is its cumulative character. Human beings cannot learn more advanced parts of knowledge within a given discipline unless they have previously learned the most basic parts of the same discipline. The previous knowledge held by an individual or organization determines the capacity of the same individual or organization to learn any further and more advanced piece of knowledge within the same discipline. Thus knowledge is a retrieval/interpretative structure both for information sets corresponding to a given discipline and for other more advanced pieces of knowledge within the same discipline. The concept of knowledge as a retrieval/interpretative structure bears a considerable resemblance to that of absorptive capacity (Cohen and Levinthal, 1989, 1990), although the latter was formulated with reference to R&D. R&D is not only useful to create new knowledge but it can also

218

Handbook of knowledge and economics

help a firm to learn (absorb) some external knowledge created by another firm or research institution, or simply stored in the scientific and technical literature. The probability that a firm having performed a given type of R&D can absorb some external knowledge depends on the similarity of the internal R&D and of the external knowledge. In turn, the similarity of two pieces of knowledge is the inverse of their cognitive distance (Nooteboom, 1999, 2000). 10.2.3

Knowledge as a Network

The property of knowledge as a co-relational structure implies that (i) the subsets of the ExtEnv studied are systems with components given by observables and by their interactions, and that (ii) we can represent the knowledge of such systems by means of a network in which variables constitute nodes and co-relations/connections constitute links. A network representation of knowledge has a static and a dynamic aspect. The dynamic aspect is captured by the density, or connectivity, of the network, defined as the ratio of the number of actual links to that of possible links for a given number of nodes. In general, for any type of network we can expect both the number of links and the number of nodes to vary in the course of time. In the case of knowledge the number of nodes can vary with the discovery of new observables or variables and the number of links can vary with the number of co-relations, or connections, between variables. We can expect network density to rise when the number of nodes grows faster than the number of links and to fall when the reverse happens. We can expect that observables will be discovered and relevant variables will be defined by a gradual process. Once new variables have been defined they will not immediately be connected to all the pre-existing variables. Using a network analogy, we can expect that the process of creation of new nodes/variables will be faster than the process of establishing links/ correlations between existing nodes/variables. Thus, in periods in which many new variables are introduced, we can expect the average density of linkages in the network of knowledge to fall. Conversely, when no new variables are introduced we can expect the density of linkages/correlations in the network of knowledge to increase (Saviotti, 2005). This dynamic representation of knowledge is compatible with Kuhn’s (1962) analysis of the evolution of science. New observables and variables are likely to be created when new paradigms emerge. In this early phase we can expect new variables to be poorly connected to those existing in the previous network of knowledge. Thus the emergence, or revolutionary, phase of a new paradigm is likely to be accompanied by a falling connectivity of the network of knowledge. On the other hand, we can expect

Theoretical framework for the generation & utilization of knowledge 219 the subsequent phase of normal science to be characterized by a growing number of links and thus by a growing network connectivity. 10.2.4

The Local Character of Knowledge

A further property of knowledge, here called its local character, follows immediately from those described in the previous sections. Even when corelations can be represented by analytical equations, such equations can contain a very limited number of variables. It is possible to write equations containing a large number of variables but it becomes increasingly difficult to solve or to interpret them. Furthermore, correlations can be detected only over limited ranges of values of the relevant variables. To explain this point we refer back to Example 1, the law of ideal gases. The relationship of pressure, volume, temperature and the number of moles of an ideal gas is valid only within a particular range of values of the variables, corresponding to the assumption that the atoms and/or molecules of the gas behave independently. However, at very high pressures and at very low temperatures equation (10.1) gradually becomes inadequate to predict the behaviour of real gases. For high pressures and low temperatures more complicated models are required in order to represent the behaviour of real gases. In general, all models are approximations valid only in particular ranges of the variables included. Any model is then always a simplified analogue of reality, containing a lower number of variables and of interactions with respect to the subset Si(ExtEnv) that it intends to represent. The model will then be a good representation of reality to the extent that the neglected variables and interactions contribute weakly to the behaviour of the system. Such weak contribution is in general unlikely to persist over the complete range of possible values of all the variables and interactions or with respect to all the properties of the system. Thus we can expect that the validity of most models will be limited to some ranges of the values of the variables and interactions of the subset Si(ExtEnv). There is a further way in which knowledge can be considered local. According to the second property of knowledge mentioned above, that of being a retrieval or interpretative structure, any further or more advanced piece of knowledge within a given discipline can only be learned by human actors who already know the fundamental parts of the same discipline. If we now consider that the whole ExtEnv can be subdivided into different subsets, each studied by a discipline, we can see that different disciplines may have very different or very closely related observation spaces. For example, physics and chemistry have very similar and partly overlapping observation spaces, while physics and anthropology have very different

220

Handbook of knowledge and economics

observation spaces. As a consequence, we can expect disciplines to have variable degrees of similarity. Here let us note that the learning ability conferred upon people/agents/actors already holding given pieces of knowledge is not only limited within a discipline. Knowledge of a given discipline A increases the probability of learning a similar discipline B more than that of learning a very different discipline C. Once more, the concept of similarity developed here bears a close though inverse resemblance to that of ‘cognitive distance’ (Nooteboom, 1999, 2000; this Handbook, Chapter 15) in knowledge space. If we consider that disciplines are themselves heterogeneous and that they tend to become increasingly heterogeneous in the course of time, we can give the previous considerations a more general form as follows. The total observation space corresponding to ExtEnv (O(ExtEnv)) can be partitioned into subsets that may cover part or all of O(ExtEnv). In O(ExtEnv) we can imagine being able to measure the distances between any two pieces of knowledge located either within the same discipline or in two different disciplines. The local character of knowledge can then be represented in the following way: The probability that a human actor/agent holding at time t a given type of knowledge (actor’s internal knowledge) can learn another piece of knowledge external to the actor or to the discipline increases in a way inversely proportional to the distance between the internal and the external knowledge. This statement can be expressed concisely by means of the following formula: PKiS(Ki+Ke) ~ 1/D0(Ki, Ke)

(10.2)

where PKiS(Ki+Ke) is the probability that a given actor having internal knowledge Ki at time t can learn external knowledge Ke, and D0(Ki, Ke) is the distance of the internal and external pieces of knowledge in O(ExtEnv). The considerations presented so far were developed based on evidence from the history of science, that is, on the outcome of cognitive processes. It is interesting that the concept of the local character of knowledge finds confirmation in the work of cognitive psychologists. For example, it seems that we remember better items that are similar to what we actually do than items completely new to us (Buenstorff, 2001, p. 3). To the extent that memory influences our selection of problems and interpretation of phenomena, it can be one of the causes of the local character of knowledge. More generally, both the interpretation of new stimuli and the formation of expectations is performed in accordance with existing schemata and

Theoretical framework for the generation & utilization of knowledge 221 categories that tend to be resistant to change. Thus the features of cognitive processes that lead to a local character of knowledge also tend, if left undisturbed, to make our individual mental representations increasingly rigid over time (ibid.). The possibility of representing knowledge as a network reinforces the existence of the local character of knowledge. We can expect to find highly connected variables within the same discipline or a narrow subset of the same discipline and to find much lower densities of connection between very far apart disciplines or subsets of knowledge. An example of this phenomenon can be found in the evolution of physics, chemistry and medicine in the last 200 years. During this period the connections between organs, cells, molecules and atoms was gradually elucidated. In other words, the observables of medicine (organs, cells) were gradually interpreted and explained on the basis of the observables of physics and chemistry. As previously pointed out, although an important objective of knowledge is to create a fully connected network, the connectivity/density of the real network of knowledge is likely to keep fluctuating as a consequence of the balance between the rate of creation of nodes and the rate of creation of links (Saviotti, 2005). We observe here that the local character of knowledge has some explicit and implicit precedents, although they are far less general than the version of the concept presented here. Atkinson and Stiglitz (1969) introduced a production function in which only a limited number of techniques is actually feasible. Improvements in technology do not concern the whole production function, but affect only one or a few techniques. Innovations are concentrated in the technology that is currently in use, while other technologies remain largely unaffected. Thus only a limited number of choices is available to a firm at a moment in time. Nelson and Winter (1982) talked about the local character of search in a similar sense. In their model each firm at a given time can be represented in input factor coefficient space by a point, corresponding to the technique used by the firm (pp. 180–83). According to their model, the probability that as a result of an innovative process a firm ends up with a different ratio of input factor coefficients is inversely proportional to the difference between the initial and the final ratios, or equivalently, to their distance in input factor space. This of course implies that local search involves incremental modifications of existing techniques and that ratios near the initial one are the most probable. As has already been pointed out, two concepts closely related to the local character of knowledge are those of absorptive capacity and cognitive distance (Cohen and Levinthal, 1989, 1990; Nooteboom, 1999, 2000; this Handbook, Chapter 15). A firm can only absorb some external knowledge similar to the one it already holds.

222

Handbook of knowledge and economics

Conversely, a firm can only absorb types of external knowledge having a small cognitive distance with respect to its own internal knowledge. Thus the concepts of absorptive capacity and cognitive distance confirm both the property of knowledge as retrieval/interpretative structure and the local character of knowledge. Summarizing, we can say that knowledge has a local character because: ● ● ●

●

10.2.5

Knowledge can provide co-relations/connections only over a small number of variables at a time. It can provide co-relations only over a limited range of values of the variables considered. The probability that a human actor holding a given internal knowledge Ki learns some piece of external knowledge Ke is inversely proportional to the distance between Ki and Ke in the observable space O(ExtEnv). The creation of new nodes can be expected to precede the creation of links within the new nodes and between the new and the old ones. Theories of Knowledge

The representation of knowledge described so far does not constitute a complete theory of knowledge. However, it is compatible with widely accepted epistemological theories. First, the creation of new observables and variables constitutes conjectures, to be tested by their correspondence to empirical observations. As Popper (1934) demonstrated, existing observables, variables and connections can never be proved to be true or correct. Their validity can only be corroborated by new experiments, but it is limited to the set of experiments carried out up to that point. Even in this limited sense knowledge can be very useful to modify our ExtEnv in the subsets of ExtEnv where the theory has been adequately tested. In a broader sense the representation of knowledge as a co-relational and as retrieval/interpretative structure is compatible with the idea of knowledge as an organized structure, to which both Kuhn’s and Lakatos’s theories belong (Chalmers, 1980). In particular, the representation of knowledge described in this chapter is compatible with some recent structuralist theories of science (Balzer et al., 1987; Franck, 1999) according to which the collection of all empirical science forms a theoretical holon, composed of constellations of elementary theories, theories that would be connected by inter-theoretical links of different types, such as equivalence, specialization, connection and so on. Furthermore, as Polanyi pointed out (1958), the process of creating new knowledge cannot by its own internal rules lead inevitably to true knowledge. There is no such thing as a

Theoretical framework for the generation & utilization of knowledge 223 so-called scientific method. Science does not develop through an algorithmic machine in which such a method is embodied, but requires intuition and imagination. Important generalizations are put forward long before they can be supported or corroborated, and sometimes they are adhered to even in presence of falsifying evidence. Objective knowledge can be created, but only by the critical analysis and the comparison of the ideas of different researchers and groups. Objective features of knowledge emerge in the context of verification and not in that of theory or concept formulation. Thus new knowledge is not created by the combination of individual scientists and the scientific method but by the collective effort of scientists working in institutions where they formulate new ideas, and compare and contrast them with those of their peers until a collectively acceptable theory emerges. The collective character of knowledge production is thus one of its most important aspects. One can then talk of knowledge embedded in social networks (Nightingale, 1998, p. 692). Finally, the representation of knowledge as a co-relational and as a retrieval/interpretative structure can be used for both true and false theories. Different networks of knowledge, consisting of different variables and connections, will exist at different times. The evolution of knowledge will be represented by the transition between these different networks. The nodes and links of a theory that is proposed at a given time, and at later times turns out to be false or incorrect, will be replaced by new nodes and links.

10.3 THE PRODUCTION OF KNOWLEDGE The term knowledge production is deliberately used here to stress the possible similarities with processes of material production. Of course, a priori the production of knowledge cannot be considered identical to that of shoes or of semiconductors. However, there is a level of generality at which some similarities can be identified. Thus we can say that the production of knowledge requires resources and inputs. Furthermore, these inputs are transformed into outputs, consisting of various types of knowledge. Using the concepts previously developed in this chapter, we could say that the detection of observables and the creation of variables and connections are the required outputs. Several economists have used production functions to analyse the production of knowledge (Griliches, 1988; Audretsch, 1998). Like any form of material production, the production of knowledge occurs in specialized institutions, although generally they are not firms (Murmann, 2004; Mokyr, 2002). Thus it seems that some established economic concepts are in principle applicable to the

224

Handbook of knowledge and economics

production of knowledge. In this section we proceed to outline a series of other economic concepts that are in principle applicable to the production of knowledge. 10.3.1

Division of Labour

Since Adam Smith (1776), division of labour has been recognized as an important determinant of economic growth. By reducing the size of the task each worker has to carry out, and by specializing in it, a growing efficiency can be achieved. A finer division of labour is obtained by increasing the number of steps into which each process is subdivided. A growing division of labour can improve efficiency but it is limited by the extent of the market and by coordination costs. Thus, to obtain economic advantages from a growing division of labour, coordination costs must not rise simultaneously. While these concepts have been predominantly discussed in the contexts of firms and of markets, they are equally applicable to the production of knowledge. According to the previous sections, knowledge can be created by discovering new observables, by defining appropriate variables and by finding co-relations between these variables. We can easily understand that to define all the observables and the relative variables for the whole ExtEnv is a task beyond the reach of individuals or of single organizations. In the production of knowledge, division of labour occurred by means of the creation of disciplines, which in turn could be subdivided into sub-disciplines, specialities and so on. Furthermore, an extensive division of labour takes place within each discipline. Bearing in mind that knowledge creation can provide two distinct routes to the adaptation of human beings to their ExtEnv, two alternative processes of discipline formation can be identified: (a) select a subset of ExtEnv, identify within it observables and variables and establish correlations between these variables; (b) select a particularly important human need, such as nutrition, housing and so on, and define it as the task of a particular discipline. Processes of type (a) correspond to our desire to know, or to general knowledge or basic research, while processes of type (b) correspond to the modification of our ExtEnv, or to technology. We can call cognitive the disciplines corresponding to our desire to know what and technological those corresponding to our desire to modify our ExtEnv. These processes can be separated in principle, but will inevitably be mixed in real knowledge-creating organizations. In other words, cognitive and technological disciplines can be separated conceptually but are very difficult to separate in practice. The reason for this inseparability is the advantage given to technological disciplines by the existence of cognitive ones having

Theoretical framework for the generation & utilization of knowledge 225 overlapping observables spaces. We can expect technological disciplines to have an observable space partly overlapping that of one or more cognitive disciplines and partly constituted by variables that are unique to them. For example, civil engineering requires knowledge of mechanics, of materials chemistry and of geology, among others, while having some variables and concepts that are unique to it. Engineering or applied disciplines provide a better way to modify ExtEnv if they are based on a sound knowledge of the subset Si(ExtEnv) of ExtEnv which they intend to modify. The previous concepts can be described in a compact way by means of a formal notation. However, in this chapter a more intuitive style of presentation will be used. Interested readers are referred to Saviotti (2004). The partitions within the overall activity of knowledge generation can be defined at different levels of aggregation. Thus sub-disciplines are subsets of disciplines, specialities subsets of sub-disciplines, theories subsets of specialities, models subsets of theories and so on. This classification could be developed further, but what has previously been introduced suffices for the purposes of this chapter. The main conclusion of this discussion is that the production of knowledge is characterized by a form of division of labour, in which different subsets of the ExtEnv are assigned to different disciplines, the subset of each discipline being further subdivided into specialities, theories, models and so on. An important question, to which only a partial answer will be given in the course of this chapter, concerns the criteria used to define the boundaries of different disciplines, specialities, theories, models and so on. Is there only one possible way of defining those boundaries? No complete answer will be attempted here, although part of the answer will emerge from the following discussion. One possible way of partitioning the whole observation space would be to classify all the possible observables according to their ‘size’, assuming that the smaller ones are ‘contained’ in the larger ones. It would then be possible to consider the discipline studying the subset of the ExtEnv containing the smallest observables as the most fundamental one. The observables of other disciplines would have to be correlated to the smallest ones. Within the natural sciences we could say that physics is the most fundamental discipline because some of its observables (e.g. sub-atomic particles) are the ‘smallest’ ones and because they are basic components of the observables of the other natural sciences. However, this representation would at best be applicable to the disciplines defined by criterion (a) above, that is, by knowledge generation as the main objective. Disciplines created by criterion (b), engineering or applied disciplines, would not fit in this classification. Further dimensions, such as human needs, would have to be added to observable size. Furthermore, in some cases the relevance of the most fundamental observables may be very limited. For example,

226

Handbook of knowledge and economics

although human beings are biological organisms, even today it would not be a very wise strategy to calculate people’s expected behaviour on the basis of their genetic heritage and of a number of phenotypic characteristics. Moreover, the strategy of calculating the expected values of some phenomena in a discipline B based on the smaller and more fundamental observables of another discipline A can only be conceived ex post (Saviotti, 2004). In the middle of the nineteenth century no one could have conceived a research strategy based on the understanding of the genetic basis of diseases, which could sound ex ante reasonable at that time. While today a research strategy aimed at elucidating the molecular mechanisms responsible for particular types of diseases would sound perfectly rational, such a strategy was not available to medical scientists in the middle of the nineteenth century. This is just an example of the separate development of science and technology that was very frequent until the nineteenth century and that is the result of their separate and not easily compatible objectives. This is why Nightingale (1998) says that science answers the ‘wrong’ questions to be used directly in technological development. In reality, the questions answered by science are not wrong, but they are generally different from those that would be required by technology. In general, scientific disciplines and technologies that we realize to be connected developed in an entirely separate way. This was not the result of mistakes made by technologists but of the choice of a research strategy that could produce satisfactory results within the time horizon of the technologists rather than a perfect solution in an infinite time horizon. In other words, many strategies of technological, but also of scientific, development were procedurally rather than substantively rational. The treatment of knowledge strategies is developed to a greater extent in Saviotti (2004, pp. 115–18). We will not here deal with this aspect except for noticing that the boundaries between disciplines, an extremely important feature of the division of labour in knowledge production, developed historically in a haphazard, non-rational and path-dependent way. This path-dependent character may have been enhanced by internal criteria of choice of observables and variables within each discipline. These internal criteria of choice could have made the coordination of different disciplines a difficult or impossible task. 10.3.2

Coordination

As was previously pointed out, the division of labour is required to simplify tasks and to allow processes of knowledge generation to become more efficient. However, the advantages of the division of labour can be obtained only if there is coordination because the different pieces of

Theoretical framework for the generation & utilization of knowledge 227 knowledge created by individual researchers/scientists need to be combined to provide a general and comprehensive knowledge of our ExtEnv. Individual scientists’ conjectures about particular subsets of the ExtEnv must be compared and combined with those of other fellow scientists. Ideas about the same observables and variables put forward by different scientists need to be compared to establish which ones provide the best correlations. Furthermore, conjectures about different subsets of observables and variables need to be combined in order to achieve either a general and comprehensive theory or to apply these conjectures to the modification of our ExtEnv. Thus the collective nature of knowledge creation and utilization involves both division of labour and coordination. Although disciplines can be considered the result of division of labour in knowledge generation, they are themselves complex and aggregated. In other words, a discipline does not contain just a unit of knowledge but a complicated structure that was itself the result of past division of labour and coordination. We can then expect that the need for coordination will arise both within and between disciplines. The partitions of O(ExtEnv) that give rise to the emergence of disciplines and that constitute the division of labour in knowledge generation are not real partitions existing in reality, but are socially constructed. The criteria on which the partitions are based may very well be scientific, but they create boundaries between observables that do not find any correspondence in reality. This has important consequences for the activities aimed at modifying our ExtEnv. Any attempt at modifying our ExtEnv will almost necessarily cross the boundaries between disciplines in the sense of involving observables contained in the ranges of different disciplines. The problem would then arise of coordinating the knowledge created in two or more disciplines in order to understand/correlate the variables corresponding to the subset of the ExtEnv to be modified. The coordination of two or more disciplines would involve the ‘combination’ of pieces of knowledge created within each of the disciplines in order to either understand or modify a subset of ExtEnv that is not contained exclusively within one of the disciplines. As was pointed out during the previous discussion of the network of knowledge, the probability of connections between variables is always limited even within each discipline, and we can expect it to be even lower between variables belonging to different disciplines. In other words, we can expect cognitive distances to be greater between than within disciplines. Perfect coordination would then imply that each variable of the two disciplines must be correlated with all other variables of the other discipline, in addition to being correlated with all the variables of the same discipline. Alternatively, if we refer to the representation of knowledge as a network, we can describe the perfect coordination of two disciplines as involving a

228

Handbook of knowledge and economics

total connectivity of the network obtained by combining the networks of the two disciplines. Of course, even within a single discipline connectivity is never total and it can change in the course of time, falling during certain periods and rising in others. The coordination of two different disciplines is likely to be more complicated because at least a part of the variables of the two disciplines can be different, and even when the variables are the same, it is possible for the concepts and tools of the two disciplines to differ. Thus interdisciplinary coordination is in general likely to be more difficult than intradisciplinary coordination. In network terms, the zones at the boundaries between two disciplines are likely to be less well connected than those at the core of the disciplines. If the modification of a given subset of the ExtEnv requires the co-relation of the variables of two or more disciplines, it is quite likely that basic or theoretical knowledge will be of less help than if the subset of the ExtEnv that one wanted to modify involved only variables internal to a discipline. The problem is magnified by the fact that different and sometimes even neighbouring disciplines developed historically in separate ways, thus making the problem of coordination more difficult. Each discipline defined its own variables, concepts, tools, modes of analysis and so on, and quite often those of two disciplines are of difficult compatibility. A clear example of these specificities is given by the different value attached to quantitative analysis and modelling in economics and sociology, two disciplines that have partly overlapping observable spaces. Of course, one way of overcoming this problem involves defining disciplines whose variables match exactly those of the subset of the ExtEnv that one wants to modify. This is the solution consisting of creating engineering, applied science disciplines, medicine and so on; that is disciplines not defined by the choice of a subset of the ExtEnv to be explained, but based on the choice of a subset of the ExtEnv to be modified. In fact, if we examine existing disciplines we can see that there is a mixture of those defined by knowledge objectives and those defined by modification/technological objectives. Even within the disciplines that seem to be primarily oriented towards knowledge generation per se, an important motivation to the creation of knowledge often came from practical problems. For example, in both astronomy and in physics several important developments were due to very practical problems: navigation, the measurement of time and the development of the steam engine are but some of the examples of practical problems providing powerful inducements to scientific development. According to Popper (1972, p. 258), practical problems are very often the sources of new theories. Nelson (1994) and Nelson and Rosenberg (1993) maintain that the creation of technology-based disciplines is by no means the exception, but it is likely to be a very general phenomenon.

Theoretical framework for the generation & utilization of knowledge 229 Incidentally, the creation of a unified general theory that was at times envisaged would correspond to creating a network of knowledge that was fully connected. The creation of a unified general theory would involve: ● ● ●

A complete coverage of the whole observation space, including every possible observable and variable. A flat distribution of probability of co-relation of different variables, always equal to its maximum possible value. A maximum connectivity of the total network of knowledge.

Let us observe here that the creation of a unified general theory would lead to non-local knowledge, since correlation would extend to any possible pair of variables and to all their values. Alternatively, the span of correlation would be equal to its maximum possible value. This case would correspond to the Laplacian dream. 10.3.3

Codified and Tacit Knowledge

The distinction between tacit and codified knowledge is a subject that has received considerable attention in the literature (see, e.g., Cowan et al., 2000; Lundvall et al., 2002). The considerations of this subsection have the function of proving that the previous framework is capable of providing analytical tools that can be used to study the codification of knowledge. First of all, knowledge production is a collective enterprise, requiring division of labour and coordination. Coordination of the work of different scientists requires communication. When researchers need to compare or combine different conjectures or theoretical structures, they have to be able to communicate. This involves not only sharing a general language (say English) but sharing the knowledge of the building blocks of their theories. For example, according to Cowan et al. (2000), the creation of a dictionary is a component of the process of codification. Existing knowledge is not created in a form that is immediately true and easily transmissible. As Polanyi (1958) and Kuhn (1962) pointed out, creative processes in science are highly subjective and their outcome is not easily communicable. In other words, knowledge is always created in a tacit form (Polanyi, 1958). However, given that (i) tacit knowledge is very difficult to communicate, and (ii) coordination requires communication, knowledge when created in a tacit form must be transformed into an explicit or codified one in order to be easily communicated. Thus the need for codification arises as a consequence of the need for coordination, which itself follows from the collective character of knowledge creation

230

Handbook of knowledge and economics

and utilization. Codification is not an intrinsic property of a given type of knowledge but the result of the process of codification, a process that requires the allocation of resources (Saviotti, 1998). Furthermore, tacit and codified knowledge are not two discrete states of knowledge. On the contrary, different pieces of knowledge have degrees of codification intermediate between the completely tacit and completely codified (Saviotti, 1998). To the extent that codification requires resources, including time, we can expect the most mature parts of disciplines to be more codified than the parts of the same disciplines close to their knowledge frontier. In the meantime, both because they are older and because they are easier to communicate, older parts of a discipline are likely to be more widely shared by the relevant scientific communities and newer parts by much more limited subsets of the same scientific communities (Saviotti, 1998). Furthermore, precisely because it is widely shared, highly codified knowledge is unlikely to provide competitive advantage to any firm using it for its productive purposes. For example, knowledge of classical mechanics, a beautifully codified sub-discipline, will not give any engineering firm a competitive advantage. If anything, it will be a necessary but far from sufficient condition to be competitive. The competitive advantage of any such firm can only come from combination of classical mechanics with other pieces of knowledge involved in an application to some productive problem, knowledge that in general can be expected to be highly tacit. Thus, in general, we can expect an inverse relationship between codification and appropriability (Saviotti, 1998). A further implication of the previous considerations is that very new knowledge at the frontier of a discipline is likely to be tacit and shared by very few people/scientists. During this initial period when it is new, a piece of knowledge will not be easily communicable, it will be difficult to imitate, and it will provide high rewards in term of monopoly profits to those few who manage to develop and acquire it. As the new type of knowledge matures and becomes more codified, a larger number of scientists can acquire it, thus reducing appropriability. However, even codified knowledge can be easily communicated only to scientists knowing the code. This type of scientists is not uniformly distributed in the world economic system. Even highly codified knowledge can lose appropriability within some countries/regions while remaining out of reach for other countries/ regions. Dynamically the distribution of appropriability in the world economic system depends on (i) the distribution of the intrinsic properties of knowledge (age, degree of codification etc.) and (ii) the distribution of human resources required to absorb and utilize the same knowledge. On the other hand, competitive advantage can be acquired either by having exclusive access to highly tacit knowledge, for example by employing the

Theoretical framework for the generation & utilization of knowledge 231 only people who have such knowledge, or by being the only ones to know the code for a new type of highly codified knowledge. As a result of the previous considerations we can expect: ● ●

10.3.4

technologies to be generally more tacit than scientific disciplines; and more science-based technologies to be generally less tacit than more empirically based technologies. World Views, Mental Models, Pattern Recognition and so on

The previous analysis did not take explicitly into account the possibility that observers constructing theories have different mental models (Buenstorff, 2001; Stahl-Rolf, 2000). Mental models are intuitive representations of the ExtEnv, or of a subset of it. Such mental models are present before observers begin observations or experiments. In this sense, world views, paradigms and so on can be considered examples of mental models. Observers having different mental models, even when faced with the same phenomena, may select their observations differently, derive from them different observables and variables, and arrive at different co-relations. In other words, the theoretical analysis developed starting from different mental models may differ in a number of dimensions. The simultaneous presence of alternative mental models is not necessarily a stable situation. If two or more mental models attempting to interpret the same subset Si(ExtEnv) of the ExtEnv are comparable, then they can be considered competitive. Comparability in this case implies the comparison of the prediction of some properties of Si(ExtEnv): in the end the mental model that leads to the most accurate predictions will be selected over the others. In this case we can expect the presence of two or more mental models to be a temporary situation. Furthermore, this would be an example of the competition of different mental models (see the following section). However, mental models are not always comparable. For example, one mental model can lead to better predictions of some properties of Si(ExtEnv) and a second mental model to better predictions of different properties of Si(ExtEnv). In this case the coexistence of different mental models could be expected to continue until a further mental model, superior to all the previous ones and capable of encompassing their observables, variables and interpretations was found. Even if in the end a unique mental model were to prevail and to replace all the pre-existing ones, this process could take a long time. We can expect the time required to achieve a synthesis of competing mental models to increase as their differences become greater. At least during certain periods we can expect two or more different sets

232

Handbook of knowledge and economics

of mental observables, variables and interpretations to correspond to the same set Si(ExtEnv) of the ExtEnv. These different models would then compete for the explanation of phenomena arising from a common observation space. The competitive interactions of different theories or mental models are the subject of next section. A very important property of science and technology that has been observed by many scholars (see, e.g., Nightingale, 1998) is the ability to recognize patterns. Pattern recognition can be analysed in terms of the concepts used in this chapter. A pattern can be considered as a set of observables and of their interactions, or connections. Such observables can be of many types. Typically they are those available to the observer at a given time. Thus they can be very aggregate observables and there is no a priori guarantee that they are good ones, in the sense of leading either to a sound knowledge of the system studied or to the ability to modify it. In general the ability to detect certain kinds of patterns depends on the capacity of observation of the scientist and on his/her previous knowledge. For example, the recognition of new patterns can be determined by their similarity to previously known patterns (Nightingale, 1998). In terms of the network representation of knowledge, patterns can be conceptualized as intermediate states between pairwise co-relations and the whole network of knowledge. A pattern is a limited network whose boundaries may not be completely well defined but whose core can be clearly identified. Within a pattern we can identify observables, or components, and interactions between them, described by connections. Patterns are the result of the selection of a particular subset of the ExtEnv. The selection procedure, which is clearly tacit, is the result of the intellectual apparatus of the observer. On the other hand, with respect to the observation of individual observables and variables, pattern detection involves an operation of synthesis. Patterns are important elements of observation but do not lead directly to an explanation without having been decomposed. Individual observables and variables have to be identified within patterns in order to detect their connections. In general we can expect the knowledge of a particular subset of the ExtEnv to proceed by identifying new observables within those that were used in the previous period and by reconstructing the connections between the new expanded set of observables. In other words, the knowledge of a particular subset of the ExtEnv progresses by gradually disaggregating the subset itself.

Theoretical framework for the generation & utilization of knowledge 233 10.3.5

Inter-theory Competition

The relationship of different theories depends on their observation spaces and on the variables that correspond to them. When two theories have identical observation spaces, corresponding to a given subset Si(ExtEnv), they compete for the explanation of that subset Si(ExtEnv) (Saviotti, 1996). Competition would here occur by means of: (i) the identification of observables and variables within Si(ExtEnv); and (ii) the establishment of connections between the relevant variables. Theories can then differ in the timing of both (i) and (ii). Success in this competition would depend on the differential ability to establish connections leading to accurate predictions of experimental results. The opposite case is that of two theories covering completely separate subsets of the ExtEnv and specializing within each of them. This is the case of theories belonging to different disciplines. An intermediate case is that of two theories covering partly overlapping subsets of the ExtEnv. This case corresponds to a situation of imperfect competition, in which each theory competes for the explanation of the shared parts of the ExtEnv but has a local monopoly in the parts of the ExtEnv that are unique to it. All these cases can be represented graphically in terms of set theory (Figures 10.1–10.3). A more formal representation of this situation is given in Saviotti (2004). Let us analyse in greater detail the situation illustrated in Figures 10.1– 10.3. First, competition occurs when the two theories, Th1 and Th2, have identical observable spaces. They are then competing because they have to provide competing explanations of the same phenomena. A phenomenon is here defined as an event or a set of events that can be explained by means of its connections between underlying observables and variables. On the other hand, when the two theories, Th1 and Th2, have completely separate observable spaces, the phenomena that they try to explain are different. In this case, Th1 and Th2 do not compete but each specializes in explaining a different subset of the ExtEnv. The relationship of Th1 and Th2 can be either of independence or of complementarity. For example, if the two theories were jointly required to explain some more general phenomena at the level of the whole discipline or to perform a modification of the ExtEnv, then they would play a complementary role. Summarizing, we can expect the relationship between different theories, and consequently their patterns of interaction, to depend on the relationship of their observation spaces. Thus a relationship of competition between Th1 and Th2 will exist if they have the same observation space and if they attempt to explain the same phenomena. The use of competition made here can be reconciled with the one commonly found in economics textbooks by referring to the equivalent biological situation.

234

Handbook of knowledge and economics ODi

OTh1 = OTh2

Note: The two theories, Th1 and Th2, are competing for the interpretation of the same subset of the ExtEnv.

Figure 10.1

Inter-theory competition ODi

OTh1

OTh2

Note: Th1 and Th2 specialize in different subsets of the observation space ODi corresponding to discipline Di.

Figure 10.2

Theories with separate observable spaces ODi

OTh2 OTh1

Figure 10.3

Theories with partly overlapping observable spaces

In biology, competition is only one of the possible forms of interaction between species. Two species are competing when they attempt to make use of the same resource, for example two types of birds feeding on the same seeds. In economics, firms try to sell goods and services that consumers want to buy. Consumers are the firms’ resources. Perfect competition requires identical and homogeneous goods because firms try to beat their

Theoretical framework for the generation & utilization of knowledge 235 competitors in selling to a homogeneous population of consumers. In the present case theories Th1 and Th2 use as resources their identical observation spaces. Thus competition in any case involves identity of the resources used. Of course, in these cases resources have to be given a very general interpretation. The intermediate case in which Th1 and Th2 have only partly overlapping observation spaces (Figure 10.2 and Figure 10.3) corresponds to imperfect or monopolistic competition. Th1 and Th2 compete in the shared parts of their observation spaces, but there can still be some form of competition in the parts of the ExtEnv that are not common to both theories depending on the degree of similarity of the relevant observables. Let us here recall the case of imperfect competition in economics. In this case firms produce non-identical but limitedly substitutable outputs. It is possible to imagine that if the degree of substitutability could be measured, for example by means of the distance of different outputs in service characteristics space, competition would be the more intense the closer the products are in service characteristics space. We can similarly say that the intensity of competition between theories will be more intense the more similar their observable spaces. Of course, both in the case of products and services and in that of theories, when the resources are not identical there is a degree of local monopoly, as in monopolistic competition. In this context theories that have completely separate observable spaces may either be independent or complementary, depending on whether their outputs have to be used jointly to produce further scientific explanations or to achieve a modification of the ExtEnv or not. In summary, we can see that processes such as the creation of different disciplines, the setting of their boundaries, the joint use of disciplines in treating a complex problem, and the interactions of different disciplines can be analysed by means of very well-established economic concepts, such as division of labour, coordination and competition, although these concepts might have to be redefined in a slightly more general sense than the one in which they have been used in the past.

10.4 INDUSTRIAL APPLICATIONS The general framework previously described can be used to derive empirical applications. This section will be concerned with the way firms create and use knowledge. The concept of the knowledge bases (KB) of firms will be used here to show how that can be done. The logical extension of the previous analysis would be to identify the variables used by firms at given times and the correlations established between these variables. In order to do empirical research we would need to have access to data sets containing

236

Handbook of knowledge and economics

the variables present in the KB of each organization at different times. Such data sets do not exist and their construction is likely to be very costly. In the previous part of this chapter a framework for the analysis of processes of knowledge generation and utilization was discussed. As was pointed out, the objective of this framework is to enable us to understand how these processes contribute to the economic performance of firms, industries and countries. In this section the analysis will be extended from knowledge generation in general to how knowledge is created and used within firms. In a firm, knowledge is generated in a number of ways and in different subsets of the firm. The subset that receives the greatest attention in the literature is R&D. However, while R&D is probably the most important contributor to knowledge generation in firms, it is not necessarily the only one. In a general way we can divide all economic activities into ‘routines’ and ‘search activities’ (Nelson and Winter, 1982). In this context search activities are all those activities that scan the ExtEnv looking for suitable alternatives or additions to existing routines. Search activities include R&D but also other activities, such as parts of industrial design or marketing. In fact, we can imagine routines and search activities to be the extremes of a range including all other possible economic activities. Within the firm, knowledge is generated in different subsets of the organization. Internal division of labour determines the knowledge-generating activities carried out by each individual or department. However, the final objective of the firm is not knowledge generation per se but the creation of new products and services with which the firm will compete. In other words, we need to understand (i) processes of knowledge generation, and (ii) how these processes contribute to the economic performance of the firm. Given what was said in the previous paragraph, we can expect knowledge generation within the firm to be a collective process, involving both division of labour and coordination. The outcomes of individual research projects need to be combined in order to lead to any marketable output. Thus the dynamics of knowledge generation and utilization can be expected to depend greatly on firm organization. A key concept in this analysis is that of the KB of the firm, defined as the collective knowledge that the firm can use to achieve its productive purposes. The collective nature of this knowledge is due to the fact that the production of new knowledge is intrinsically dependent on the interactions of individuals within organizations. Such interactions are highly organization specific and we cannot expect the same knowledge to be produced by two organizations even if at a previous time they hired people with the same competencies. We can understand that the objective of a good firm is to coordinate the activities of its members so that the overall KB is greater

Theoretical framework for the generation & utilization of knowledge 237 than the sum of the pieces of knowledge held by individual members. In this section the mapping and measurement of relevant properties of the KB will be discussed, as well as the impact of the same properties on firm performance. Furthermore, this type of analysis will be shown to be in principle compatible with the general framework previously described. According to this framework, knowledge can be represented as a network of variables and their connections. Existing data sources include databases on patents and publications. Both patents and publications are traces of knowledge far more aggregate than variables. It is possible to identify within both of them ‘smaller’, or more disaggregated, units in the form of technological classes (in patents) or of themes (in patents and papers). We can then map the KB of the firm as a network whose nodes are technological classes or themes. This representation of the KB can be compatible with the general framework if the units we choose are ‘invariant’ in the conditions in which the behaviour and the performance of the firm are studied. This implies that the definition of a given class or theme must not vary during the period studied. As a consequence the use of patents or of papers can only be considered an approximation to the true mapping of the KB of firms. The approximation is likely to be a good one to the extent that the part of the KB described by patents or papers dominates the overall KB of the firm. A fairly close connection to the general framework previously described can be established by analysing the KB of a firm as a network constituted by some units of knowledge (technological classes or themes) and by their connections. The procedure used to construct an image of the KB of a firm consists of identifying the technological classes, or alternatively the themes, contained in the patents of the firm and to find their interactions by studying the co-utilization of given technological classes/themes in different patents. The KB of the firm is thus represented by a network whose nodes are the technological classes/themes of the firm and whose links are due to the frequencies of co-utilization of technological classes/themes in different patents. The representation of the KB thus obtained is approximate since it does not include all of the possible contributions to it. The KB of a firm, even of the most science based, is not purely scientific, but it contains many components. That this is the case is recognized in the literature, where concepts such as core competencies (Prahalad and Hamel, 1990) and complementary assets (Teece, 1986) describe different parts of the KB. The method described above takes into account only the scientific or technological components of the KB and needs to be complemented by other types of information to provide a complete representation of the KB. The representation that we just described only tells us what ‘pieces’

238

Handbook of knowledge and economics

of knowledge are combined within the KB of the firm. A complete representation of the KB would also include an ‘organizational’ network, describing the subsets of the firm creating each piece of knowledge. The reconstruction of such an organizational network requires information that is available to firms themselves and that they might not be willing to share. Thus, for the moment we focus on the types of knowledge that firms use. In the cases where the scientific and technological components of the KB constitute a good approximation for the overall KB, for example in highly science-based industries, this approach can give us interesting and meaningful results. In these cases it is possible to derive a graphic representation of the scientific and technological components of the KB by means of lexicographic analysis (Saviotti et al., 2003) and to calculate a number of properties of the KB. These two applications will be briefly described here. 10.4.1

Mapping the Knowledge Base (KB) of Firms

The mapping to the KB of the firm is carried out by means of lexicographic analysis (LA), a linguistic engineering technique. In principle this technique can be applied to any text, within that it can recognize not only key words but also themes, that is, sentences that define meaningful subsets of knowledge. The technological classes found in patents constitute a useful approximation to the themes that we can identify, although themes can provide a richer and more easily interpretable source of information on knowledge dynamics. Thus we can consider technological classes or themes as alternative candidates for the nodes of the network of the KB. Such a network can be reconstructed by measuring the frequency of co-occurrence of either technological classes or themes in the patents of the same firm. The existence of co-occurrence of any pair of technological classes or themes identifies a link and the frequency of cooccurrence measures the strength of the link. In this sense, the representation of the KB that will be provided here is compatible with the concept of knowledge as a co-relational structure and with its representation as a network. Figure 10.4. shows the KB of Rhône-Poulenc in the period 1996–98. During the 1990s Rhône-Poulenc underwent a strategic reorientation, moving away from its previously dominant activity (chemistry) and towards the life sciences. In 1996–98 the KB of Rhône-Poulenc still showed a segmented structure, including both chemical classes (the firm’s past) and biological classes (the firm’s intended future). Further analysis of the strategic reconversion of Rhône-Poulenc, leading to its merger with Hoechst to form Aventis, revealed (i) a growing relative weight

239

Figure 10.4

Network representation of the knowledge base of Rhône-Poulenc for the period 1996–98

Note: Each of the labels that constitute the nodes in the figure corresponds to a technological class. Class A61K, indicated by an arrow, separates biotechnology-related classes, on its left, from chemistry-related classes, on its right.

240

Handbook of knowledge and economics

of biological classes with respect to chemical classes and (ii) a growing network density of the KB (Saviotti et al., 2005). This example highlights one of the possible applications of LA to study knowledge creation in firms: changes in KB following a strategic reorientation of firms. There are other possible applications to the study of the following topics: (i) relationship between knowledge dynamics and organisational dynamics; (ii) mergers and acquisitions; and (iii) innovation networks. 10.4.2

The Measurement of KB Properties

The measurement of properties of the KB relies on the extension of methods developed to construct indicators of innovation, starting from databases on patents. The first step towards these measurements is the construction of a matrix of technological co-occurrences. This matrix is constructed starting from a database including all the patents awarded in a given field of technology during a period of time. All the technological classes included in the patents considered are plotted on both axes of the matrix. The frequencies of co-occurrence of any pair of technological classes are then written in the cases situated at the intersection of the two classes. The construction of this matrix relies implicitly on the representation of knowledge as a network. Starting from the frequencies of cooccurrence found in this matrix, we can calculate the values of a number of relevant properties of the KB. For the procedure used for these calculations see Nesta and Saviotti (2005, 2006). The most important of these properties are: coherence/knowledge integration; specialization; differentiation; and similarity of different KBs. The meaning of specialization, of differentiation and of similarity is fairly intuitive. However, the concept of coherence deserves some more detailed comments. That a coherent firm can be more competitive or effective than an incoherent firm has long been suspected. This hypothesis could not be properly tested in the absence of a method to measure coherence. Such a method was devised by Teece et al. (1994), based on the products of firms. It could then be ascertained whether or not coherent firms were more or less frequent than incoherent ones (e.g. conglomerates). In a knowledge-based economy the coherence of the KB of a firm can be expected to be at least as important as the coherence of the outputs of the firm. Furthermore, it is important to understand the meaning of coherence. Teece et al. (1994) were talking about ‘relatedness’, but relatedness can be both similarity and complementarity. Nesta (2001) calculated the coherence and other properties of the KB of biotechnology firms. In this research, coherence was interpreted as predominantly complementarity.

Theoretical framework for the generation & utilization of knowledge 241 10.4.3

The Relationship between the Properties of the KB and Firm Performance

The final link in the analysis of knowledge generation and utilization by firms consists of finding the relationship between knowledge dynamics and firm performance. This can be done by testing econometrically the existence of correlation between a number of independent variables including the properties of the KB, and a dependent variable measuring one of the possible dimensions of firm performance. In two separate papers Nesta and Saviotti (2005, 2006) used either technological performance, defined as the number of patents that a firm can produce during a given period of time, or as the value of Tobin’s Q, which measures stock market performance. In these two papers both properties of the KB turned out to be significant and robust determinants of the performance of firms in biotechnology, although their impact on firms in different sectors dependent on biotechnology varied. This application completes the logical chain leading from knowledge generation to its industrial applications. Of course, the logical chain of reasoning can be considered complete only in principle. Much further work is required to put these methods on a more secure basis and to extend their application to more firms and sectors.

10.5 SUMMARY AND CONCLUSIONS In this chapter a representation of knowledge suitable for the analysis of processes of knowledge creation and utilization was presented. This representation is based on two properties of knowledge: knowledge is (a) a co-relational structure and (b) a retrieval/interpretative structure. This is not a complete representation of knowledge, but one intended to help interpret the collective processes of knowledge creation and utilization involving different types of organizations (firms, public research institutes, universities etc.) and taking place in knowledge-based economies. Although the previous representation of knowledge is not complete, it is compatible with a number of epistemological theories ranging from those of Popper, Kuhn, Lakatos, to more recent ones, such as the structuralist approach to science. According to the first of the above properties, knowledge provides connections between different variables representing the subset of the ExtEnv that is explored in a discipline, speciality and so on. According to the second property, knowledge is useful to retrieve both information and other pieces of knowledge external to an agent but similar to the internal knowledge previously held by him/her. From these

242

Handbook of knowledge and economics

two properties it is possible to deduce that knowledge has a local character, in the sense of providing correlations only among a small number of variables and over limited ranges of their values. The local character of knowledge as formulated here encompasses a number of concepts previously proposed, such as the modified production function proposed by Atkinson and Stiglitz (1969), the local character of search (Nelson and Winter, 1982), absorptive capacity (Cohen and Levinthal, 1989, 1990) and cognitive distance (Nooteboom, 1999, 2000). Furthermore, as a consequence of being a co-relational structure, knowledge can be represented as a network having variables as nodes and their connections as links. A related objective of this chapter is to show that the production of knowledge bears some similarity to the production of material goods and of services. The creation of disciplines can be interpreted as the result of the division of labour within knowledge production. Division of labour can be useful only if the outcomes of individuals and organizations can be coordinated. Furthermore, it is shown that different theories can compete for the explanation of phenomena arising out of a common observable space. Thus well-established economic concepts such as division of labour, coordination, competition, specialization and so on can be used to interpret the production of knowledge, although the use of these concepts in this context may require some adaptation. The characterization of knowledge as a co-relational and a retrieval/ interpretative structure can provide the basis for the analysis of the processes of knowledge creation and utilization in firms and organizations. Such processes cannot be analysed by means of the variables constituting the KB of the firm because the required information is not available. A type of information that is easily available to us is that on patents or publications. Patents are fragments or traces of knowledge far more aggregate than variables, but they can be broken down into smaller units, such as technological classes or themes. By doing this and by applying particular techniques such as lexicographic analysis or co-occurrence matrices, we can (i) construct a graphic representation of the KB of the firm as a network, and (ii) measure properties of the KB of firms, such as coherence/knowledge integration, specialization, differentiation, similarity and so on. The KB network has technological classes as nodes and linkages determined by the co-occurrence of technological classes in different patents. This allows us to follow changes in the composition of the KB occurring as a consequence of changes in firm strategy. Utilizing the co-occurrence matrices of the technological classes contained in patents, we can calculate a number of properties of the KB of the firm, such as its coherence, its degree of differentiation, and the similarity of different KBs,

Theoretical framework for the generation & utilization of knowledge 243 and show that these properties are determinants of firm performance. The representation of the KB presented here is approximate, but it is useful in knowledge-intensive sectors. The final step in the procedure linking knowledge generation to knowledge utilization consists of finding out the relationship between several independent variables, including the relevant properties of the KB, and a dependent variable that measures a possible dimension of firm performance. That such a relationship exists, and that it is significant and robust for biotechnology-based sectors has already been demonstrated. The conceptual framework and the methods described in this chapter can provide us with a representation of the knowledge that can in principle allow us to both measure and model it. These methods are certainly approximations and much further work is required both to improve and to extend them to more firms, industrial sectors, different types of organizations (e.g. research organizations) and organizational forms (e.g. innovation networks). The development of methods to map and measure knowledge based on a sound general conceptual framework is a central task in the construction of an economics of knowledge. The present chapter is an attempt in this direction.

REFERENCES Atkinson, A.B. and Stiglitz, J.E. (1969), ‘A new view of technological change’, Economic Journal, 99, 573–8. Audretsch, D. (1998), ‘Agglomeration and the location of innovative activity’, Oxford Review of Economic Policy, 14 (2), 18–29. Balzer, W., Moulines, U. and Sneed, J. (1987), An Architectonic for Science: The Structuralist Program, Dordrecht: Reidel. Buenstorff, G. (2001), ‘Dynamics of knowledge sharing: from cognitive psychology to economics’, Jena, Max Planck Institute for Research into Economic Systems, Evolutionary Economics Unit. Chalmers, A.F. (1980), What is this Thing called Science?, Milton Keynes: Open University Press. Cohen, M. and Levinthal, D. (1989), ‘Innovation and learning: the two faces of R&D’, Economic Journal, 99, 569–96. Cohen, M. and Levinthal, D. (1990), ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35, 128–52. Cowan, R., David, P. and Foray, D. (2000), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 9, 211–54. Foray, D. (2000), L’Economie de la Connaissance, Paris: Repères. Franck, R. (1999), ‘La pluralité des disciplines, l’unité du savoir et les connaissances ordinaires’, Sociologie et Société, XXXI, 129–42. Freeman, C. and Soete, L. (1997), The Economics of Industrial Innovation, London: Pinter. Georgescu-Roegen (1971), The Entropy Law and the Economic Process, Cambridge, MA: Harvard University Press. Gibbons, M., Limoges, C., Nowotny, H., Schwartzmann, S., Scott, P. and Trow, M. (1994),

244

Handbook of knowledge and economics

The New Production of Knowledge: The Dynamics of Science and Research in Contemporary Societies, London: Sage. Griliches, Z. (1988), R&D and Productivity. The Econometric Evidence, Chicago, IL: University of Chicago Press. Hirshleifer, J. (1988), Price Theory and Applications, Englewood Cliffs, NJ: Prentice-Hall. Kuhn, T.S. (1962), The Structure of Scientific Revolutions, Chicago, IL: University of Chicago Press. Loasby, B. (1999), Knowledge, Institutions and Evolutionary Economics, London: Routledge. Losee, J. (1977), A Historical Introduction to the Philosophy of Science, Oxford: Oxford University Press. Lundvall, B.-Å., Johnson, B. and Lorenz, E. (2002), ‘Why all this fuss about codified and tacit knowledge?’, Industrial and Corporate Change, 11 (2), 245–62. Mokyr, J. (2002), The Gifts of Athena: Historical Origins of the Knowledge Economy, Princeton, NJ: Princeton University Press. Murmann, J.P. (2004), Knowledge and Competitive Advantage, Cambridge: Cambridge University Press. Nelson, R.R. (1994), ‘The co-evolution of technology, industrial structure, and supporting institutions’, Industrial and Corporate Change, 3 (1), 47–63. Nelson, R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Nelson, R.R. and Rosenberg, N. (1993), ‘Technical innovation and national systems’, in R.R. Nelson (ed.), National Systems of Innovation: A Comparative Study, Oxford: Oxford University Press, pp. 3–21. Nesta, L. (2001), ‘Cohérence des bases de connaissances et changement technique: une analyse des firmes de biotechnologies de 1981 à 1997’, PhD thesis, Grenoble: Université Pierre Mendès-France. Nesta, L. and Saviotti, P.P. (2005), ‘Coherence of the knowledge base and the firm’s innovative performance: evidence from the US pharmaceutical industry’, Journal of Industrial Economics, 53, 105–24. Nesta, L. and Saviotti, P.P. (2006), ‘Firm knowledge and market value in biotechnology’, Industrial and Corporate Change, 15, 625–52. Nightingale, P. (1998), ‘A cognitive model of innovation’, Research Policy, 27, 689–709. Nooteboom, B. (1999), Inter-Firm Alliances: Analysis and Design, London: Routledge. Nooteboom, B. (2000), ‘Learning by interaction: absorptive capacity, cognitive distance and governance’, Journal of Management and Governance, 4 (1–2), 69–92. Polanyi, M. (1958), Personal Knowledge: Towards a Post-Critical Philosophy, London: Routledge. Popper, K.R. (1968), The Logic of Scientific Discovery, London: Hutchinson, first published 1934. Popper, K.R. (1972), Objective Knowledge: An Evolutionary Approach, Oxford: Oxford University Press. Prahalad, C. and Hamel, G. (1990), ‘The core competence of the corporation’, Harvard Business Review, 68, 79–91. Saviotti, P.P. (1996), Technological Evolution, Variety and the Economy, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Saviotti, P.P. (1998), ‘On the dynamics of appropriability of tacit and codified knowledge’, Research Policy, 26, 843–56. Saviotti, P.P. (2004), ‘Considerations about knowledge production and strategies’, Journal of Institutional and Theoretical Economics, 160, 100–121. Saviotti, P.P. (2005), ‘Knowledge networks: structure and dynamics’, presented at the Existence workshop: ‘Innovation networks – Developing an integrated approach’, University of Augsburg, 10–14 October. Saviotti, P.P., de Looze, M.A. and Maupertuis, M.A. (2005), ‘Knowledge dynamics and the mergers of firms in the biotechnology based sectors’, Economics of Innovation and New Technology, 14, 103–24.

Theoretical framework for the generation & utilization of knowledge 245 Saviotti, P.P., de Looze, M.A., Maupertuis, M.A. and Nesta, L. (2003), ‘Knowledge dynamics and the mergers of firms in the biotechnology based sectors’, International Journal of Biotechnology, 5, 371–401. Shannon, C.E. and Weaver, W. (1949), The Mathematical Theory of Communication, Urbana, IL: University of Illinois Press. Smith, A. (1970 and following editions), The Wealth of Nations, original edition 1776, Harmondsworth: Penguin Books. Stahl-Rolf, S. (2000), ‘Persistence and change of economic institutions. A social–cognitive approach’, in P.P. Saviotti and B. Nooteboom (eds), Technology and Knowledge: From the Firm to Innovation Systems, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 263–84. Teece, D. (1986), ‘Profiting from technological innovation’, Research Policy, 15, 285–305. Teece, D.J., Rumelt, R., Dosi, G. and Winter, S.G. (1994), ‘Understanding corporate coherence: theory and evidence’, Journal of Economic Behavior and Organization, 22, 1–30. Weber, M. (1968), The Protestant Ethic and the Spirit of Capitalism, London: Unwin University Books, first UK edition 1930.

11 Models of adaptive learning in game theory1 Jacques Durieu and Philippe Solal

11.1 INTRODUCTION It is relatively well known that agents do not always make rational choice. Several experimental studies show that many observed behaviors can be both well described ex post and robustly predicted ex ante by a simple family of learning theories. In the last 20 years, the variety of such learning models that have been used in economics has increased tremendously. Most of these models explain the need for agents to learn because, initially, they lack information. Moreover, agents may have limited ability to make optimal decisions under various constraints. In a framework of strategic interaction among agents, game theory has traditionally assumed that players are perfectly rational. However, this approach is complemented by models with boundedly rational players. In game theory, the assumption of boundedly rational behaviors can be justified by a lack of information on the game structure. This means that players have limited knowledge of the strategic environment, the payoff functions or the rationality of other players. Then, agents try to simplify their decision task. A common assumption in the models discussed in this chapter is that players consider that their environment is stationary. One of the main purposes of these models is to investigate whether agents, who face the same game recurrently over time, learn to play an equilibrium of this game. In such a case, the models provide a foundation of equilibrium based on information collected by players along their interactions. By contrast, it is worth noting that agents do not learn to understand their environment. One difficulty with this approach is that there does not exist a unique way of defining bounded rationality. Thus different models prevail in different situations. This chapter provides an overview of behavioral models developed in the literature on evolutionary game theory. We introduce a classification of these models according to information used by agents to formulate their beliefs about their opponents’ behavior and to evaluate every available action. Two broad classes of learning models are distinguished according to the sophistication of the representation of strategic 246

Models of adaptive learning in game theory 247 environment formed by agents. On the one hand, agents do not elaborate beliefs about their opponents’ behavior. On the other hand, agents form (naive) expectations on the future play of their opponents. For each category of learning models, we present some rules of thumb that are compatible with the representation of the environment and with the available information. Furthermore, we show that, for given classes of games, Nash equilibrium often constitutes a good prediction of behavior of the agents following these rules. It is interesting to note that the well-known model of Bayesian learning is not encompassed within our framework. This model leads to powerful results. For instance, Kalai and Lehrer (1993) demonstrate that if agents begin with beliefs about their opponents’ strategies that are ‘reasonable’, then beliefs updated using Bayes’s rule will generate predictions of future play that are nearly correct. And, if agents optimize subject to their beliefs, then eventually play begins to look like equilibrium behavior. However, this model presumes an enormous amount of sophistication and reasoning power on the part of agents. By contrast, we focus here on models with fewer informational and computational requirements. Indeed, we restrict our attention to the case of adaptive behaviors where agents make naive assumptions about the behavior of their opponents. On the other hand, we also exclude models inspired from biology where agents are preprogrammed. In such models, agents have a null rationality. Evolutionary processes such as the replicator dynamics may approximate processes generated by some behavioral rules but as such are useful only from a technical point of view. Other studies of learning in games that include such types of models are provided by Walliser (1998), Sobel (2000) and Sandholm (2009). The chapter proceeds as follows. In Section 11.2 we propose a classification of learning models. In Section 11.3 we review some learning rules without beliefs. Section 11.4 presents some belief-based models.

11.2 ADAPTIVE LEARNING We consider models in which players are myopic. Such an assumption has two implications. First, each agent assumes that his environment is stationary over time. When agents perceive the strategic dimension of their environment, this means that they believe that their opponents use stationary strategies. In particular, agents think that their own choices do not affect the future play of their opponents. Thus agents do not take into account the possibility that their opponents are similarly engaged in adjusting their own behavior, and agents do not try to influence the future

248

Handbook of knowledge and economics

play of their opponents. Second, at each period, agents use rules of thumb to guide their own choice of action next period. In particular, this means that, at each point of time, each agent does not choose a strategy that maximizes his expected payoff over all future periods. We classify the learning setting on the basis of three characteristics: the representation of environment; the way in which agents evaluate actions; and the decision rule. In the context of game theory, two ways of representing the environment can be identified: either agents are aware of the strategic environment or they do not take into account this dimension. In the first way of representation, agents understand that their payoffs depend on the actions played by their opponents. Moreover, agents are able to form beliefs about the future behavior of their opponents. Even though agents formulate myopic predictions, their beliefs may be more or less sophisticated according to the amount of information used. Two particular ways in which agents construct beliefs should be emphasized. On the one hand, agents build predictions about their opponents’ actions based on information contained in the full history of play. Agents compute probability distributions on the set of their opponents’ actions based on the observed frequency distribution of the actions taken in the past. In particular, agents may assume that their opponents play according to some fixed probability distributions. In order to estimate these distributions, they use information about the actions played by their opponents in the past. On the other hand, agents form predictions based on the actions played in the previous period. At the beginning of every period, agents assume that pure strategies that have been played in the previous period will be chosen in the current period. Trivially, these models with beliefs require that each agent observes actions played in the past by his opponents. However, this type of learning can be used even if agents have no information on their opponents’ payoffs. In the second way of representing the environment, agents do not take into account the strategic nature of the situation. Then, they do not create models of the situation. At least three elements can explain this fact. First, agents may not be aware of the strategic environment. In particular, they may ignore the fact that their own payoffs depend on the choices of their opponents. Second, agents may not be able to observe actions played by their opponents. Third, agents may think that the elaboration of expectations on their opponents’ behavior is too computationally expensive. Models of learning can also be divided according to different ways of evaluating actions. Procedures of evaluation of performance are either forward oriented or backward oriented. In the case of forwardoriented evaluations, agents use criteria of performance based on the expected payoffs of actions. Such criteria imply that agents are aware of

Models of adaptive learning in game theory 249 the strategic environment and construct beliefs about their opponents’ behavior. Given predictions on the future play of their opponents, agents compute the expected payoffs of their actions and then construct an order over these actions. Then agents make a selection among actions based on this order. Forward-oriented evaluations require that each agent knows his own payoff function and is able to compute expected payoffs. A second kind of procedures is concerned with evaluations that are backward oriented. Agents use criteria of performance based on past payoffs. They collect information in the history of interactions in order to construct an order over their actions. There are several ways to make such a classification. One can distinguish between procedures of evaluation based on information concerning the agents’ own payoffs and procedures that require information about other agents. First, each agent may evaluate his available actions on the basis of the payoffs achieved in the past. Note that it is not necessary to assume that agents have information either about their own payoff function or about the payoff functions of their opponents. Each agent has only to observe his own choices and his past realized payoffs. Second, agents may use backward-oriented procedures based on information about other agents. The evaluation of a given action depends on the payoffs obtained by some other agents who choose this action in the previous periods. There are several ways for an agent to evaluate an action, depending on the number of other agents observed. Moreover, evaluations can be constructed either on the basis of the full history of play or on fragmentary information about it. Even though it is not necessary to assume that agents have full knowledge of payoff functions, agents have to observe the payoff associated with the action chosen by some other agents. Furthermore, these procedures make sense only if agents share the same payoff function. However, this information is not necessarily known by agents. In addition, notice that backward-oriented evaluations are applicable in situations where agents do not construct beliefs about their opponents’ behavior. However, agents who form beliefs can also use backward-oriented criteria if they consider that computing expected payoffs is a too difficult (and costly) task. Finally, one can classify learning models according to the decision rule. Choice is either deterministic or probabilistic. When choice is deterministic, a unique action is selected. This means that, given the evaluation of actions, one action is played with probability one. For instance, with a forward-oriented criterion of performance, the action associated with the highest expected payoff (given predictions about the behavior of opponents) may be selected with probability one provided there is no tie. If the criteria of evaluation is backward oriented, the action that has yielded the highest past payoff may be chosen with probability one. When choice is

250

Handbook of knowledge and economics

probabilistic, several actions may be chosen with positive probability. In other words, the rule adopted by an agent leads to a mixed strategy on his set of available actions. Such a rule may be used when agents satisfy rather than optimize, that is, when they try to achieve only a certain aspiration level instead of a maximum. Then each action that is evaluated as satisficing may be chosen with a positive probability. In the next two sections, we review some of the learning models that have been proposed in the literature. First, we focus on models using backward-oriented criteria. Second, we present some models with beliefs and forward-oriented criteria.

11.3 BACKWARD-ORIENTED CRITERIA OF PERFORMANCE In this section, we deal with models in which agents use backward-oriented criteria. In these models, it is not necessary that agents form beliefs about their opponents’ future behavior or that they are even aware of the strategic dimension of their environment. 11.3.1

The Strategic Framework

We begin with a description of the strategic environment. We consider a finite n-player game in strategic form G 5 (N, (Xi) i[N, (ui) i[N) That is, N 5 { 1, . . . , n } is a finite population of players of size n. For each i [ N, Xi denotes the finite set of pure strategies or actions available to i. Let X 5 Pi[NXi be the set of pure strategy profiles. For each i [ N, we write X2i 5 Pj[N \{i} Xi . Let D (Xi) denote the set of distribution probability on the set Xi. And, for each i [ N, let ui: X S R be agent i’s payoff function. Let t 5 1, 2, . . . denote successive time periods. The game G is played once in every period. The payoff received by i [ N at period t when he plays action xti and his opponents choose xt2i is ui (xti, xt2i) . Implicit in this notation is the idea that payoffs are time-independent. When evaluation of actions is based on information about past payoffs, the payoff received by i at period t is denoted uti. As mentioned previously, we divide models with backward-oriented criteria into two broad categories: models in which players’ own past payoffs matter, and models in which past payoffs received by other agents matter.

Models of adaptive learning in game theory 251 11.3.2

Reinforcement

In this subsection, we focus on models in which evaluation of actions is based on players’ own past payoffs. More precisely, we restrict our attention to models in which only realized payoffs matter. Thus we leave aside models based on the performance criterion of no regret, that is, models in which, based on what has happened so far, agents ask whether they could have done better by playing differently.2 Furthermore, we consider models in which the full history of payoffs is taken into account. The basic idea of reinforcement learning goes back to the Law of Effect introduced by Thorndike (1898). Actions that have led to good outcomes in the past are more likely to be repeated in the future. This reveals that, in models of this genre, choice is probabilistic. A well-known version of learning models based on the Law of Effect is due to Bush and Mosteller (1955). This is a central learning model in behavioral psychology. Bush and Mosteller consider a positive reinforcement learning in which the action that has yielded high payoffs in the past will tend to be adopted with a higher probability by agents who face the same game repeatedly. More recently, Börgers and Sarin (1997) have studied the properties of a modified version of Bush and Mosteller’s learning rule. More precisely, at the beginning of every period, each agent updates the probabilities with which he uses the actions at his disposal as follows. First, he observes the action that he has played and the payoff that he has received in the previous period. Then he updates his probabilities. Börgers and Sarin assume that payoffs lie in the open unit interval (0, 1). Let eti (xi) be the Kronecker function of played action defined as follows: eti (xi) 5 e

1 if xi is played at time t, 0 if xi is not played at time t.

Conditional on having received payoff uti at period t, agent i [ N updates his probability of playing action xi [ Xi at period t 1 1 as follows: (xi) 5 eti (xi) uti 1 (1 2 uti) pti (xi) pt11 i

(11.1)

To have a precise understanding of this learning rule, it is interesting to study how the probability associated with each action evolves between two (xi) 2 pti (xi) denote the incremental change in periods. Let dpti (xi) 5 pt11 i the probability with which action xi is played by i between periods t and t 1 1. Straightforward calculations show that for all i [ N, xi [ Xi: dpti (xi) 5 uti (eti (xi) 2 pti (xi))

252

Handbook of knowledge and economics

Table 11.1

0 1

A 2 × 2 constant-sum game 0

1

(0.4,0.2) (0.2,0.4)

(0.2,0.4) (0.4,0.2)

A first point we may note is that as soon as an action is chosen, its probability is increased. Indeed, the incremental change dpti (xi) is always positive when xi is played at period t. And dpti (xi) is always negative when xi is not chosen. Another noticeable feature is that incremental changes between period t and t 1 1 depend on the realized payoff at period t: a higher payoff leads to stronger changes. Börgers and Sarin (1997) study this learning model in finite two-person games. To do this, they use techniques drawn from stochastic approximation theory. In particular, they establish that the replicator dynamics approximates the reinforcement dynamics defined by (11.1) over bounded time intervals. However, the asymptotic properties of the two models are quite distinct. Börgers and Sarin show that for any finite two-person game the reinforcement dynamics defined by (11.1) converges with probability one to a pure strategy profile. This is not the case with the replicator dynamics. For instance, consider the 2 3 2 constant-sum game represented by the payoff matrix shown in Table 11.1. This game has a unique mixed-strategy Nash equilibrium and the replicator dynamics cycles around it. The last result is rather weak. It means that the reinforcement dynamics may converge to a pure strategy profile that is not a Nash equilibrium. However, stronger results can be derived from another variant of the reinforcement learning model. Assume now that the probability of choosing an action at period t 1 1 is proportional to that action’s cumulative payoff up through time t. This model is studied, among others, by Erev and Roth (1998) and Laslier et al. (2001). For any agent i [ N, denote the propensity to play action xi [ Xi at period t by qti (xi) . The propensities evolve as a function of the realized payoff. More precisely, if agent i receives payoff uti at period t his propensities are updated as follows: (xi) 5 qti (xi) 1 eti (xi) uti qt11 i The probability for agent i of playing action xi at period t 1 1 is a monotone increasing function of his propensities. More precisely, the probability that agent i chooses action xi at period t 1 1 is determined as follows:

Models of adaptive learning in game theory 253 (xi) 5 qt11 (xi) / a qt11 (xir) pt11 i i i

(11.2)

xir [Xi

This reveals that payoffs must be positive; otherwise propensities might become negative and then the choice probabilities would be undefined. Once again, we can re-express the process in terms of incremental changes between periods t and t 1 1. We obtain for all i [ N and xi [ Xi,

dpti (xi) 5

uri s a x r [X q (xi r) 1 a s# tui i

i

(eti (xi) 2 pti (xi))

(11.3)

1 i

where q1i (xir) is the initial propensity of action xir, which takes a positive value. Observe that, as in Börgers and Sarin (1997), when an action is chosen, its probability is increased. And incremental changes between period t and t 1 1 still depend on the realized payoff at t. But, due to the fact that the denominator in (11.3) increases with cumulated payoffs, the incremental impact of the payoff in any given period diminishes over time. This means that, as more experience accumulates, learning slows down. More precisely, the step size of learning is decreasing and determined for each agent by his payoff experience. It appears that this feature is crucial to obtain results. Stochastic approximation theory indeed obtains its strongest results when the step size is decreasing. This ensures that random effects are eliminated in the long run. By contrast, with a constant size step, as in Börgers and Sarin (1997), no matter how much time has passed, there is a constant probability that there will a large chance fluctuation away from the deterministic path. This therefore happens eventually with probability one. Beggs (2005) and Hopkins and Posch (2005) show that the long-run behavior of the stochastic process defined by (11.2) is governed by an adjusted form of the replicator dynamics. Furthermore, they establish that, in a large class of 2 3 2 games, the reinforcement dynamics defined by (11.2) converges with probability one to a Nash equilibrium. 11.3.3

Imitation

In the present subsection, evaluation of actions is based (at least partially) on past payoffs received by other agents. The basic idea is to copy or imitate successful actions played by other agents. We do not take into consideration models in which imitation is determined by other criteria such as the popularity of an action. An interesting feature of imitation rules is that when one action is used by all agents, the probability of choosing any other action is zero.

254

Handbook of knowledge and economics

Several variants of model of imitation based on past payoffs have been suggested in the literature. One well-known imitation rule is ‘imitate-thebest’. Such a rule prescribes imitating the action that achieved the highest payoff in the last period. To provide a description of this rule, consider a symmetric finite n-person game. All agents have the same set of available actions and the same payoff function. Let St denote the set of actions associated with the highest payoff in period t: St 5 { xtj, j [ N: uj (xtj, xt2j) 5 max uk (xt) } k[N

At period t 1 1, agent i chooses xi [ Xi with probability (xi) 5 e pt11 i

1/ 0 S t 0 if xi [ S t, 0 otherwise

It appears that this decision rule is deterministic provided there is no tie between strategies, that is, St is a singleton set. In a context of asymmetric finite n-person games, this version of imitation can be modified as follows. Consider n populations of agents: to each player role i [ N corresponds one population of agents who all have the same set of available actions and the same payoff function. In every period, agents imitate the action with the highest payoff in their population at the previous period. However, such a learning rule may lead to inefficient results, even in simple contexts. To see this, consider the two following examples. Example 1 This example is due to Schlag (1998a). Consider a population of agents who are engaged in a game against Nature. This can be interpretated as a situation where each agent of the population faces an opponent who plays according to a given stationary mixed strategy. At each period, all agents play the game. Assume that the payoffs of each agent in the population are given by the matrix in Table 11.2, where the set of actions of agents is { 0, 1 } and the set of actions of Nature is { 2, 3 } . Moreover, assume that Nature plays action 2 with probability 2/3 and action 3 with probability 1/3. This implies that action 0 yields a higher expected payoff than action 1 Table 11.2

0 1

A game against Nature 2

3

4.5 5

4.5 2.5

Models of adaptive learning in game theory 255 Table 11.3

0 1

A 2 × 2 game with a dominant strategy 0

1

(4,4) (2,1)

(1,2) (0,0)

(approximately 4.5 . 4.16). However, assume that an agent in the population chooses action 1. With probability 2/3, this agent receives the highest payoff. In such a case, all agents imitate action 1, that is, the action associated with the worst average payoff in the long run. Example 2 Consider two agents who repeatedly play a symmetric two-person game given by Table 11.3. This game has a strictly dominant strategy, 0. Let us introduce a small degree of noise in the choice of action. For instance, suppose that agents experiment with a small probability in each period. When they experiment, agents play an action at random. It is possible to verify that it is ‘easier’ (with ‘easier’ interpreted as requiring fewer experiments) to move from equilibrium (0,0) to equilibrium (1,1) than to move from equilibrium (1,1) to equilibrium (0,0). Starting from (0,0), if one agent experiments and plays action 1, then he receives the highest payoff (2 > 1). Consequently, both agents play action 1 at the following period. Conversely, starting from (1,1), if one agent experiments and plays action 0, then he receives the lowest payoff (1 < 2). Consequently, both agents play action 1 in the following period. A transition from (1,1) to (0,0) requires that both agents experiment simultaneously and play action 0. From this it can be inferred that, when the probability that  an  agent experiments tends to zero, the fraction of time the inefficient equilibrium (1,1) is visited by the process tends to one in the long run.3 One can specify imitation rules in such a way that the above difficulties can be ruled out. In particular, in stationary environments, Schlag (1998b, 1999) analyzes which imitation rules an agent should choose when he has the opportunity to observe a limited number of other agents in the same player position. For instance, Schlag (1998b) considers a situation where in every period each agent draws at random one other agent and observes his previous action and his realized payoff. He establishes that an agent should never imitate an agent whose payoff realization was worse than his own. Furthermore, an agent should imitate agents whose payoff realization was better than his own. Schlag studies an imitation rule that satisfies the above requirements: the ‘proportional imitation rule’. This rule is defined as follows. The probability of imitation is proportional to the

256

Handbook of knowledge and economics

difference between sampled and own realized payoff. Formally, consider agent i who plays action xti and receives payoff uti. Assume that payoffs lie in the interval [ a, b ] . Conditional on having sampled agent j, agent i imitates agent j with probability (utj 2 uti)

t11 i

p

t t (x ) 5 • (b 2 a) if uj . ui , 0 if utj # uti and xtj 2 xti t j

To have a complete description of the proportional imitation rule, note (xti) 5 1 if xtj 5 xti. And the probability of choosing an action that that pt11 i is not observed (i.e. not played by i or j) is zero. Clearly, the proportional imitation rule is stochastic. An application of the proportional imitation rule to Example 1 yields the following result. Consider first an agent i who uses action 0 and draws an agent j choosing action 1. The probability with which i imitates j is (1 / 2.5) 3 (5 2 4.5) 3 (2 / 3) 5 2 / 15. Consider now an agent j who uses action 1 and draws an agent i choosing action 0. The probability with which j imitates i is (1 / 2.5) 3 (4.5 2 2.5) 3 (1 / 3) 5 4 / 15. Consequently, agents eventually choose action 0, even if an agent playing action 1 receives the highest payoff. In other words, the proportional imitation rule allows the slowing down of the imitation of action 1. In an interactive environment, the introduction of bounded memory may have some interesting effects. Alós-Ferrer (2008) studies how bounded memory affects the ‘imitate-the-best’ rule in symmetric finite n-person games. Assume that agents have finite memory, recalling the actions chosen and the payoffs received by all agents for exactly m $ 2 periods. Moreover, assume that agents mimic the actions that yield the highest payoffs remembered. Let M t denote the set of actions with the highest payoff from period t 2 m 1 1 to t, that is: Mt 5 { xtj , j [ N, t [ { t 2 m 1 1, . . . , t } : uj (xtj , xt2j) 5 max { uk (xt2m11) , . . . , uk (xt) , k [ N } } At period t 1 1, agent i chooses xi [ Xi with probability (xi) 5 e pt11 i

1/ 0 Mt 0 if xi [ Mt, 0 otherwise

(11.4)

It appears that this decision rule is deterministic provided there is no tie between strategies, that is, M t is a singleton set. Alós-Ferrer (2008) states

Models of adaptive learning in game theory 257 that bounded memory favors high-payoff outcomes. To see why, consider the game described in Example 2. Suppose now that m $ 3. Furthermore, let us introduce a small degree of noise in the process defined by (11.4). Assume that agents are allowed to experiment and to choose one action at random. It is possible to check that it is ‘easier’ to move from equilibrium (1, 1) to equilibrium (0, 0) than to move from equilibrium (0, 0) to equilibrium (1, 1). Indeed, starting from (0, 0), if one agent experiments and plays action 1, then he receives payoff 2. However, both agents remember that action 0 has achieved payoff 4. So they play action 0. And the same result holds if one agent experiments twice in succession. Consequently, a transition from (0, 0) to (1, 1) requires more than two experiments. Conversely, starting from (1, 1), if both agents experiment simultaneously and play action 0, then they receive payoff 4. Consequently, both agents play action 0 at the following period. Next, since agents recall that action 0 is associated with the highest payoff, action 0 is imitated at the following periods. Thus two experiments are sufficient to transit from (1,1) to (0,0). This means that, when the probability that an agent experiments tends to zero, the process tends to favor the efficient equilibrium over the long run. The fraction of time the inefficient equilibrium (1,1) is visited by the process tends to zero in the long run. A key point to observe is that memory slows down the speed of learning when the inefficient action is adopted. In addition, Alós-Ferrer (2008) claims that for symmetric 2 × 2 coordination games, the imitation process defined by (11.4) converges with probability one to a state where all agents choose the same strategy for m periods. This result depends crucially on the assumption of symmetry of the game. Since agents have the same set of actions and observe the same payoff realizations, they tend to adopt similar strategies in each period. And the observed action that is more successful against itself is imitated by everyone. Josephson and Matros (2004) establish a similar result for the class of finite n-person games. They consider a model with n populations. Each player role in the game is associated with a population. They assume that in every period, one agent in each population is drawn at random to play the game. The agent drawn to play in role i chooses an action according to an imitation rule. More precisely, all agents have a finite memory of length m. Each agent remembers the most recent m elements in the history of his population’s past strategies and payoff realizations. And each agent imitates the actions that yield the best average payoff in a sample of length s drawn at random in his memory. It appears that this decision rule is stochastic since samples are drawn at random. The sampling procedure enables us to establish that the stochastic process defined on the set of truncated histories of length m converges to a

258

Handbook of knowledge and economics

state in which only one action is played in each population in the most recent m periods. Such a state is called ‘monomorphic’ by Josephson and Matros. The intuition behind this result is that the sampling procedure may enable each agent to observe the same subset of observations in several periods. As a consequence, during these periods, each agent chooses the same action. If an agent draws these periods, then he chooses the same action again. In this way, only one action may be played in the most recent m periods. Thus the sampling procedure is helpful to obtain a convergence to monomorphic states in finite games, provided the length of sample is not too large. In other words, the imitation process with bounded memory and sampling converges to monomorphic states if information is sufficiently incomplete. Furthermore, if agents have an opportunity to experiment, Josephson and Matros show that, in a certain class of games, efficient outcomes are played significantly more frequently in the long run provided the probability that agents experiment tends to zero. We shall see in the sequel that the procedure of sampling also matters in models with beliefs and with forward-oriented procedures of evaluation.

11.4 FORWARD-ORIENTED CRITERIA OF PERFORMANCE In this section, we consider learning models with beliefs and with forward-oriented evaluations of actions. We restrict our attention to learning models in which agents choose actions to maximize expected payoffs given their prediction. We distinguish two kinds of models according to information used by agents to predict what their opponents are going to do in the future. First, beliefs are based on (fragmentary) information about the history of play. This means that we endow agents with (un)bounded memory. Second, agents keep track of the last play by other agents. In other words, previous periods are immediately forgotten and only the last profile matters. In addition, we consider a population of agents positioned on a local interaction structure. More precisely, each agent interacts with a subset of the population, his neighbors. In such a case, each agent forms beliefs only about what his neighbors will do. Such beliefs are based solely on information about his neighbors’ choices in the last period. One implication of such a spatial structure is that agents have a very naive representation of their interactive environment. Each agent does not understand that his neighbors’ future behavior depends on actions chosen by some agents who may not belong to his own neighborhood.

Models of adaptive learning in game theory 259 11.4.1

Fictitious Play Process

One strand of the literature on learning in game theory focuses on procedures that are based on Fictitious Play procedures. Originally, Fictitious Play models are two-person games (Brown, 1951; Robinson, 1951). These procedures are belief-based models in that past plays are mapped to beliefs about their opponents’ future play. In a Fictitious Play process, each agent knows his strategic opportunities but he has no information about those of his opponents. Furthermore, agents do not necessarily realize that their own choice affects their opponents’ future play: they behave as if the environment is stationary. For instance, each agent a priori believes that each of his opponents is using a stationary mixed strategy that he tries to learn. Each agent keeps track of the empirical distribution of the opponents’ past strategy choices, updates his beliefs about his opponents’ strategies, and responds optimally given these beliefs. Formally, let qt2i [ Pj2iD (Xj) denote the frequency distribution computed by agent i at the end of period t and based on the history of his opponents’ choices up through time t. Let Rti denote the set of actions that are best responses against qt2i, that is: Rti 5 e xi: a qt2i (x2i) ui (xi, x2i) 5 max a qt2i (x2i) ui (xir, x2i) f xr [ X x2i [X2i

i

i

x2i [ X2i

The probability with which agent i chooses xi [ Xi at period t 1 1 is: (xi) 5 e pt11 i

1/ 0 Rti 0 if xi [ Rti, 0 otherwise

It appears that this decision rule is deterministic provided there is no tie between strategies, that is, Rti is a singleton set. A game has the Fictitious Play property if every limit point of every sequence of beliefs generated by a Fictitious Play procedure is a Nash equilibrium of the game. Procedures of this kind have been analyzed extensively. Robinson (1951) shows that every zero-sum two-person game has the Fictitious Play property. Miyasawa (1961) generalizes this result to the class of two-person games with two strategies. Monderer and Shapley (1996) show that every finite n-person game with identical payoff functions has the Fictitious Play property. On the negative side, Shapley (1964) has shown that a generalization of the Fictitous Play property to the class of finite games is impossible. More recently, Fictitious Play has again attracted interest because the study of convergence in beliefs is not necessarily consistent with the idea that agents learn to play an equilibrium. As Young (1993), Fudenberg and Kreps (1993), and Jordan (1993) all note,

260

Handbook of knowledge and economics

convergence in beliefs to a mixed Nash equilibrium may only entail that play passes through a deterministic cycle of miscoordinated pure strategies that belong to the support of the mixed Nash equilibrium. Consequently, Sela and Herreiner (1999), but also Young (1993), look for convergence in (pure) strategies rather than looking for convergence in beliefs. To be precise, Young (1993) introduces a Fictitious Play process with bounded memory and sampling. Each agent assumes that every other agent is choosing a strategy according to a fixed probability distribution and that these distributions are independent among agents. At the beginning of any period, each agent inspects s plays drawn without replacement from the most recent m periods. And each agent estimates the probability distributions of his opponents by computing the empirical distributions of his opponents’ past strategies in the sample. In determining his strategy to that period, each agent chooses a (pure) best response to the empirical distributions of his opponents’ past strategies in the sample. This Fictitious Play process with bounded memory and sampling defines a finite Markov chain on the set of truncated histories of length m. Young looks for convergence in strategies. He defines the class of weakly acyclic finite games. A finite game is weakly acyclic if each strategy profile is connected to a strict Nash equilibrium through an asynchronous best-response path. Consider an arbitrary weakly acyclic game. Young shows that if s/m is sufficiently small, the Fictitious Play process with bounded memory and sampling converges with probability one to a strict Nash equilibrium from any initial state (or initial truncated history). In other words, when factual information that agents obtain through sampling is sufficiently incomplete, the population of agents learns to play a strict Nash equilibrium. As in Josephson and Matros (2004), it is crucial that agents sample in their memory. With such a sampling procedure, there is a positive probability that each agent observes a given subset of profiles during several successive periods. Therefore there is a positive probability of obtaining specific sequences of realized profiles. In particular, it is possible to construct sequences of states (or truncated histories) that are associated with the asynchronous best-response paths leading to strict Nash equilibria. By contrast, this result does not hold if agents do not sample in the truncated history of play. Thus the mere introduction of bounded memory is not sufficient to rule out cycles. The analysis conducted by Young (1993) demonstrates that the combination of incomplete information and myopic optimization leads to convergence to some strict Nash equilibrium from any initial truncated history. This result, however, does not address the issue of equilibrium selection. For, in general, the game admits multiple strict Nash equilibria and the long-run behavior of the Fictitious Play process depends on the

Models of adaptive learning in game theory 261 initial truncated history. Dependence on initial conditions, and more generally path dependence, plays a critical role in evolutionary game theory. It appears that a little bit of persistent noise can break up this path dependence and overcome lock-ins. With such a noise, the perturbed Fictitious Play process is an irreducible Markov chain. Irreducibility means that the process passes in finitely many steps from any given truncated history to any other truncated history, with positive probability. Almost certainly such a perturbed process will not remain in an inferior truncated history forever, nor it will stay in any other truncated history. Rather it tends to visit every truncated history from time to time. But some truncated histories may be visited much more frequently than others. While the Fictitious Play process no longer gets locked into a particular truncated history in a deterministic sense, it may be hooked to certain truncated histories in a statistical sense. The question now is which are these ‘stochastically stable states’. To address such a problem, suppose, in addition, that in each time period there is a small independent probability e . 0 that agents fail to implement a best response to their beliefs and choose a strategy randomly from their strategy set. By allowing deviations from a myopic best response, perturbations introduce another element of bounded rationality. Such a perturbation admits several interpretations: mistake, active experimentation, population renewal. With these perturbations as part of the process, the full Fictitious Play process with bounded memory and sampling is an irreducible and aperiodic Markov chain. Consequently, for each e . 0, this process has a unique stationary distribution me defined over the set of truncated histories of length m. Assuming that e is close to zero, a state is stochastically stable if it belongs to the support of the limit stationary distribution lim e S 0me 5 m*. Young (1993, 1998, chs. 4 and 7) and Maruta (1997) determine the stochastically stable truncated histories of the Fictitious Play process with bounded memory and sampling for various finite games. In particular, if the game is a 2 3 2 symmetric coordination game, the only stochastically stable truncated history is the repetition of the risk-dominant equilibrium. This means that the riskdominant equilibrium is the strategy profile that the process visits most of the time – but not necessarily all the time – when very little, but some noise remains. Using a concept that extends the risk-dominant equilibrium to product sets of strategies, Durieu et al. (2011b) obtain a similar result for the class of finite games. 11.4.2

Local Interaction

Many important interactions in an economic, social, political or computational context are local in the sense that they are based on local rather

262

Handbook of knowledge and economics

than global relationships. A local interaction structure formally describes who interacts with whom. It specifies for each agent a set of neighbors, the set of agents with whom the agent interacts. The agent chooses from a set of available actions and the agent’s choice depends or, more concisely, responds to his neighbors’ choices. Note that we leave aside models in which the local interaction structure is endogenous, that is, depends on individual choices. The interesting feature of such models is that a local change of an agent’s behavior will not affect all agents simultaneously but rather diffuse across the local interaction structure. Some works in evolutionary game theory analyze contagion phenomena in a society with a local interaction structure. Contagion is said to occur if one strategy can spread by a contact effect from a particular group of agents to the entire population. Local interactions between agents are either pairwise or multilateral. In the first case, the interaction structure is modeled as an undirected graph whose vertices are the agents. Two agents are neighbors if they are joined by an edge, and direct interaction is possible only between neighbors. Namely, given a population of agents, the edges of the graph define the opportunities for direct interaction. Each agent chooses a strategy from a fixed set of strategies and is paired with his neighbors to play a two-person game. The total payoff to an agent is the weighted sum of the payoffs from playing with each of his neighbors. The weights reflect the importance an agent attaches to the strategic interactions he is involved in. Formally, let Vi be the set of agent i’s neighbors. For each neighbor j [ Vi, agent i associates a weight aij. From each bilateral encounter, say with agent j [ Vi, agent i earns a payoff pij (xi, xj) if i plays action xi [ Xi and j chooses action xj [ Xj. Thus, for a profile x [ X , the total payoff to agent i is ui (x) 5 a aijpij (xi, xj) j[Vi

In the second case, the interaction structure is modeled as a hypergraph whose vertices are the agents. An agent participates in a collection of local games associated with the groups or hyper-edges he belongs to and chooses a strategy that affects his local payoff in each of these constituent games. The interpretation of the interaction hypergraph is that interaction is possible only within groups of agents that define the hyper-edges of the hypergraph. With this interpretation, two agents are neighbors if they belong to the same hyper-edge. Note that a pair of neighbors can be involved in several local games. The agent’s payoff is the weighted sum of the payoffs from interacting with the various groups of agents he belongs to. Formally,

Models of adaptive learning in game theory 263 let xi denote the family of groups agent i belongs to. For each group E [ xi, agent i associates a weight aiE . And, for each group E [ xi, agent i receives a payoff piE (xE) when the partial profile xE [ Pj[EXj is played. Thus, for a profile x [ X , the total payoff to agent i is ui (x) 5 a aiEpiE (xE) E [xi

Whenever the hypergraph contains only one group of agents, the interaction is global. Otherwise the interaction is local. In the special case where each group of agents is of size two, the game is a game played on an undirected graph. While the direct interaction between agents is confined to a collection of subsets of the population, all agents are indirectly connected through the interaction structure. In most evolutionary models, interaction structure is supposed to be pairwise (Ellison, 1993, 2000; Anderlini and Ianni, 1996; Blume, 1993; Berninghaus and Schwalbe, 1996a, b; Young, 1998; Durieu and Solal, 2003; Durieu et al., 2011a; Berninghaus et al., 2006; Weidenholzer, 2010). The evolution process is either deterministic or stochastic. In a deterministic process, the behavior of each agent is described by a transition function that maps past choices of his neighbors to his strategy set. This means that the only information that determines an agent’s next strategy is his neighbors’ past choices. This is another expression of myopia of the agent’s behavior. Agents are not supposed to take into account or to be informed about the strategy choices of the agents outside their neighborhood although the behavior of some of them has a direct influence on their neighbors’ behavior. In the simplest situation, each agent has two strategies and plays a best response to the current choice of his neighbors, that is, a myopic best response. Ellison (1993), Berninghaus and Schwalbe (1996a) and Durieu et al. (2007) study such an evolutionary process. The interaction structure is a regular graph, each agent plays the same symmetric 2 3 2 game with each of his neighbors and updates his strategy at each discrete period of time according to the myopic best-response procedure. A classical result states that the sequence of strategy profiles converges either to a fixed point or to a two-period limit cycle (Goles and Olivos, 1980). From this point of view, a formal analogy can be drawn between a myopic best-response process with local interaction effects and a Fictitious Play process with bounded memory since both processes admit limit cycles. Contagion occurs whenever the sequence of strategy profiles reaches a fixed point with the property that each agent chooses the same strategy. The analysis of Durieu et al. (2007) indicates that there is no fast way to check when contagion occurs, whereas there exist several fast routines (easy-to-check necessary condition) to rule out contagion.

264

Handbook of knowledge and economics

Clearly, contagion cannot occur if the constituent 2 3 2 game is an anticoordination game. Obviously, contagion occurs whenever one of both strategies is strictly dominant. In the case of coordination games, the outcome depends on several parameters, such as the interaction structure and the payoffs of the constituent game. In a stochastic process, some noise is introduced into the process. There are several ways to proceed. Here we shall continue to draw the formal analogy between a Fictitious Play process with bounded memory and sampling and a myopic best-response process with an interaction structure. As mentioned previously, Young (1993) shows that the addition of a sampling procedure is a sufficient condition to rule out limit cycles in a Fictitious Play process with bounded memory. Durieu and Solal (2003) apply the sampling procedure to a model with a local interaction structure. In particular, they determine to what extent a spatial sampling procedure can be a sufficient condition to rule out two-period limit cycles. They assume that agents are spaced along a ring. For any agent i [ N, set 0 Vi 0 5 2l, where l is an integer. Precisely, each agent has l neighbors on his left and l neighbors on his right. At each time period, each agent is paired with each of his neighbors to play a symmetric 2 3 2 coordination game. At the beginning of each period, each agent draws without replacement r , 2l agents among his neighbors, observes the strategies played in this sampling in the last period and chooses a best response against this observed local profile. In this way, we can distinguish the interaction neighborhood of an agent and his information neighborhood. Note that whereas each agent always has the same interaction neighborhood, he may have different information neighborhoods across periods. The authors obtain three main results for 2 3 2 coordination games. First, if each agent draws a number of neighbors at least equal to one and at most equal to half the size of the interaction neighborhood, the process converges to an absorbing state – associated with a strict Nash equilibrium of the constituent game – whatever the initial condition may be. As in Young (1993), when information that agents obtain through sampling is sufficiently incomplete, the population of agents learns to play a strict Nash equilibrium. Second, when occasionally agents choose non-optimal responses, the strategy corresponding to the risk-dominant Nash equilibrium is the only stochastically stable strategy of the game. This latter result is similar to that in Ellison (1993) but with a weaker assumption on available information. Third, the speed of convergence of this process is better than in Ellison (1993). Consequently, the addition of a sampling procedure in a myopic best-response process with a local interaction structure has similar implications to that in Young (1993) for a myopic best-response process with effects of memory.

Models of adaptive learning in game theory 265

11.5 CONCLUSION Economists have intensively studied learning in games over recent years.4 This survey puts some of the work in perspective. The adaptive learning models described above require little information or computational capacity. Nevertheless, the chapter shows that lack of information and bounded rationality can be overcome by repeating physical interactions between agents.

NOTES 1. Support from the French Ministry for Youth, Research and Education through project SCSHS-2004-04 is gratefully acknowledged. We also wish to thank A. Festré, Laurent Mathevet, Nicolas Querou, Jean-Marc Tallon and Bernard Walliser for helpful comments. 2. For a review of this type of approach, see Young (2004). 3. For a more detailed presentation of this method, see the next section. 4. See for example Fudenberg and Levine (1998), Young (2004) and Sandholm (2010).

REFERENCES Alós-Ferrer, C. (2008), ‘Learning, memory, and inertia’, Economics Letters, 101, 134–6. Anderlini, L. and Ianni, A. (1996), ‘Path dependence and learning from neighboors’, Games and Economic Behavior, 13, 141–77. Beggs, A.W. (2005), ‘On the convergence of reinforcement learning’, Journal of Economic Theory, 122, 1–36. Berninghaus, S.K. and Schwalbe, U. (1996a), ‘Conventions, local interaction, and automata networks’, Journal of Evolutionary Economics, 6, 297–312. Berninghaus, S.K. and Schwalbe, U. (1996b), ‘Evolution, interaction, and Nash equilibria’, Journal of Economic Behavior and Organization, 29, 57–85. Berninghaus, S.K., Haller, H. and Outkin, A. (2006), ‘Neural networks and contagion’, Revue d’Economie Industrielle, 114, 205–24. Blume, L. (1993), ‘The statistical mechanics of best-response strategy revision’, Games and Economic Behavior, 11, 111–45. Börgers, T. and Sarin, R. (1997), ‘Learning through reinforcement and replicator dynamics’, Journal of Economic Theory, 77, 1–14. Brown, G. (1951), ‘Iterative solutions of games by fictitious play’, in T.C. Koopmans (ed.), Activity Analysis of Production Allocation, New York: John Wiley and Sons, pp. 374–6. Bush, R. and Mosteller, F. (1955), Stochastic Models for Learning, New York: John Wiley and Sons. Durieu, J. and Solal, P. (2003), ‘Adaptive play with spatial sampling’, Games and Economic Behavior, 43, 189–95. Durieu, J., Haller, H. and Solal, P. (2007), ‘Contagion and dominating sets’, in R. Topol and B. Walliser (eds), Cognitive Economics, Oxford: Elsevier, pp. 135–52. Durieu, J., Haller, H. and Solal, P. (2011a), ‘Nonspecific networking’, Games, 2, 87–113. Durieu, J., Solal, P. and Tercieux, O. (2011b), ‘Adaptive learning and p-best response set’, International Journal of Game Theory, 40, 735–47. Ellison, G. (1993), ‘Learning, local interaction, and coordination’, Econometrica, 61, 1047–71.

266

Handbook of knowledge and economics

Ellison, G. (2000), ‘Basins of attraction, long run stochastic stability, and the speed of stepby-step evolution’, Review of Economic Studies, 67, 17–45. Erev, I. and Roth, A. (1998), ‘Predicting how people play games: reinforcement learning in experimental game with a unique, mixed strategy equilibria’, American Economic Review, 88, 848–81. Fudenberg, D. and Kreps, D. (1993), ‘Learning mixed equilibria’, Games and Economic Behavior, 5, 305–67. Fudenberg, D. and Levine, D. (1998), The Theory of Learning in Games, Cambridge, MA: MIT Press. Goles, E. and Olivos, J. (1980), ‘Periodic behaviour of generalized threshold functions’, Discrete Mathematics, 30, 187–9. Hopkins, E. and Posch, M. (2005), ‘Attainability of boundary points under reinforcement learning’, Games and Economic Behavior, 53, 110–25. Jordan, J. (1993), ‘Three problems in learning mixed-strategy Nash equilibria’, Games and Economic Behavior, 5, 368–86. Josephson, J. and Matros, A. (2004), ‘Stochastic imitation in finite games’, Games and Economic Behavior, 49, 244–59. Kalai, E. and Lehrer, E. (1993), ‘Rational learning leads to Nash equilibrium’, Econometrica, 61, 1019–45. Laslier, J.F., Topol, R. and Walliser, B. (2001), ‘A behavioral learning process in games’, Games and Economic Behavior, 37, 340–66. Maruta, T. (1997), ‘On the relationship between risk-dominance and stochastic stability’, Games and Economic Behavior, 19, 221–34. Miyasawa, K. (1961), ‘On the convergence of the learning in a 2 3 2 non-zero-sum two person game’, Research Memorandum, 33, Economic Research Program, Princeton University, NJ. Monderer, D. and Shapley, L. (1996), ‘Fictitious Play property for games with identical interests’, Journal of Economic Theory, 68, 258–65. Robinson, J. (1951), ‘An iterative method of solving a game’, Annals of Mathematics, 54, 296–301. Sandholm, W.H. (2009), ‘Evolutionary game theory’, in R.A. Meyers (ed.), The Encyclopedia of Complexity and Systems Science, Oxford: Springer, pp. 3176–205. Sandholm, W.H. (2010), Population Games and Evolutionary Dynamics, Cambridge, MA: MIT Press. Schlag, K. (1998a), ‘Justifying imitation’, mimeo, University of Bonn. Schlag, K. (1998b), ‘Why imitate, and if so, how?’, Journal of Economic Theory, 78, 130–56. Schlag, K. (1999), ‘Which one should I imitate?’, Journal of Mathematical Economics, 31, 493–522. Sela, A. and Herreiner, D. (1999), ‘Fictitious Play in coordination games’, International Journal of Game Theory, 28, 189–98. Shapley, L. (1964), ‘Some topics in two-person games’, in M. Dresher, L. Shapley and A. Tucker (eds), Advances in Game Theory, Princeton, NJ: Princeton University Press, pp. 1–28. Sobel, J. (2000), ‘Economists’ models of learning’, Journal of Economic Theory, 94, 241–61. Thorndike, E.L. (1898), ‘Animal intelligence: an experimental study of the associative processes in animals’, Psychological Review, 8, 1–109. Walliser, B. (1998), ‘A spectrum of equilibration processes in game theory’, Journal of Evolutionary Economics, 8, 67–87. Weidenholzer, S. (2010), ‘Coordination games and local interactions: a survey of the gametheoretic literature’, Games, 1, 551–85. Young, H.P. (1993), ‘The evolution of conventions’, Econometrica, 61, 57–84. Young, H.P. (1998), Individual Strategy and Social Structure. An Evolutionary Theory of Institutions, Princeton, NJ: Princeton University Press. Young, H.P. (2004), Strategic Learning and its Limits, Oxford: Oxford University Press.

12 The fragility of experiential knowledge Dominique Foray

12.1 INTRODUCTION Every summer, forest fires make the headlines. With the arrival of the hot season, not only the Mediterranean regions but also central Europe once more become the scene of these anticipated disasters that wipe entire areas of forest and woodland off the map. When these blazes were at their height recently, some good news was announced, however. And, naturally, it came from the scientific domain. Researchers have developed tools for the numerical simulation of wind flow, allowing the analysis of the effects of variations in wind speed upon contact with a forest fire. According to these researchers, this tool can be used for the organization of the prevention and fight against forest fires. Ideally fire fighters would know, in real time, how a fire is likely to progress depending on the wind and would adapt their tactics accordingly. So this is a fantastic scientific breakthrough that provides very useful knowledge in this period of heat waves – knowledge that is all the more useful since certain other knowledge has deteriorated. And it is in fact this latter knowledge that enabled the problem to be avoided!

12.2 EXPERIENTIAL KNOWLEDGE Know-how or knowledge is what gives human beings the ability to act, on a practical or intellectual level. This ability to act is exercised in the domains of production (I know how to garden, I know how to solve a problem), consumption (thanks to my knowledge of music, I can appreciate opera) and also anticipation (my knowledge of the mountains leads me to postpone the hike I wanted to go on). Experiential knowledge springs from the experience of individuals and organizations. It is not anti-scientific; it has simply not undergone the tests that give a piece of knowledge scientific status. It is nonetheless wideranging, sound, rational and effective in a particular circumstance or life event. It confers on those possessing it capacities for action that allow them to resolve not only practical but also intellectual problems. It is no doubt less general than other knowledge since the experiences that generated it are 267

268

Handbook of knowledge and economics

local and specific. Here we are talking about experiential knowledge rather than traditional knowledge, as although traditional knowledge represents a significant part of experiential knowledge, it does not account for all of it. By definition, experiential knowledge not only belongs to the past but is also the fruit of present experiences to be utilized in the future. It seems to me that a useful distinction must be made immediately between on the one hand private experiential knowledge – for example traditional know-how, manual skills and recipes, kept secret or shared by a small number of individuals – and on the other, collective experiential knowledge possessed by a community, tribe or population, and mainly concerning methods and techniques used in daily life. For the first category, I take the example of the knowledge acquired by the famous Italian stringed-instrument makers or certain master glassmakers. In the other category, I use as an example such factual knowledge as that concerning the avalanche zones in the vicinity of a particular alpine valley with which each person in the village is familiar. Other examples involve some kind of collective experiential know-how – that concerning the virtues of a certain irrigation system or farming method in a dry area or that of regarding the art and techniques of building floating gardens – very useful knowledge where rural areas are affected by significant floods. The central argument of this chapter is that the memorization, management and optimization of this knowledge pose delicate and difficult problems since it is by its very nature not very visible, often tacit and less generally applicable than scientific know-how. Before proceeding to the analysis of the fundamental problems raised by the very nature of experiential knowledge, we need to recall the main elements of the economics of reproduction of knowledge to emphasize the general difficulties of knowledge reproduction and transmission, which are created by its tacit and local character.

12.3 THE ECONOMICS OF KNOWLEDGE REPRODUCTION: A PRIMER An essential aspect of knowledge that makes its reproduction difficult is pointed out by Polanyi (1967), who introduced us to the concept of tacit knowledge. Tacit knowledge cannot be expressed outside the action of the person who has it. In general, we are not even aware that we have such knowledge, or else we simply disregard it: ‘We can know more than we can tell’ (Polanyi, 1967, p. 4). We can use the example of the rugby player who tries to describe all the gestures and know-how required to score a goal. At the end of a long description, the player concludes: ‘If you tried to

The fragility of experiential knowledge 269 write down exactly, with absolute certainty, everything you do when you kick a ball between two posts, it would be impossible, you’d still be here in a thousand years. But you just need to have done it once and your body and mind have the exact formula, ready to be repeated’ (interview with J. Webb, British journalist, quoted in Foray, 2006, p. 71). It is only when the player is prompted to describe in detail what he does that he becomes aware of all the gestures he made and the intentions he had ‘without thinking’. For this very reason, tacit knowledge is a good that is difficult to make explicit for transfer and reproduction. The reproduction of knowledge primarily involves the composition, delivery and use of a script, that is, a set of rules similar to those given to an actor who is asked to improvise on a particular theme. Three main forms of elaboration and transmission of scripts can be distinguished (see Figure 12.1). Form (a) consists in demonstration, which takes place primarily in the context of relations between master and apprentice or teacher and learner. The teacher lays down a set of rules that he or she transmits to the learner through gestures and speech. Form (b) is that of codification, in which the script is detached from the person in possession of the knowledge, with a view to inscribing it in a medium. This form may require successive modelling phases and the mobilization of languages other than natural language. In form (b) the script may be imperfect (e.g. the operating manual for a machine) but it has the virtues of a public good (it is a non-rival good that can be copied and distributed at a very low cost). Both forms (a) and (b) imply the elaboration and presentation of the script, a phase in the modelling of tacit knowledge. It is a difficult and costly process. Take, for example, a tennis teacher who wants to transmit his knowledge. Whether he wants to write a book or provide teaching on the court, he has to create a model that breaks down the gesture into micro-movements. Codification (form (b)) would probably require additional modelling phases, although not necessarily. For instance, codification of a cooking recipe would involve knowledge modelling very similar to that required for its demonstration. Form (c) consists in an audiovisual recording of the action. The recording of voices and images provides a means for facsimile reproduction, which allows the memorization and analysis of knowledge mobilized during that action. In this case, the script is not really created, but the subject matter is there, faithfully memorized, available to be worked on in constructing the script (one can, for example, show a scene in slow motion or enlarge a photo to study a particular mechanism better).

270

Handbook of knowledge and economics

Tacit knowledge

(c) Filming and recording

Composing a script (modelling phase) successive modelling

(a) Defining ‘gestures and speech’

Figure 12.1

(b) Codifying

Three forms of reproduction of tacit knowledge

Such a sequence (tacit knowledge, elaboration of a script, reproduction) involves the key moment of the dynamics of knowledge that is the composition of a script. Even if the knowledge is then reproduced through demonstration and if no written description exists, a script has been elaborated and this dramatically changes the nature of the knowledge. It can now be expressed, transmitted and reproduced. Socialization of knowledge starts with the creation of the script (whatever form this takes). As already said, scripts can be codified – that is to say, they can be expressed in a particular language and recorded on a particular medium.

The fragility of experiential knowledge 271 As such, they are detached from the individual, and the memory and communication capacity created is made independent of human beings. Although it involves high fixed costs, codification also enables agents to perform a number of operations at a very low marginal cost (Cowan et al., 2000). It reduces the costs and improves the reliability of storage and memorization. As long as the medium remains legible and the code has not been forgotten, codified scripts can, theoretically, be stored and retrieved indefinitely. Other aspects of transmission – such as transport, transferral, reproduction and even access and search – are functions whose costs always decrease with codification. Because codified script is easy to reproduce, the number of copies can be multiplied. This makes it easier to retrieve and transport. Considering our focus – the fragility of experiential knowledge – the highlighted function of codification, which is of creating memory, communication and learning capabilities, is crucial. When codifying became common, as Goody (1977, p. 37) writes, ‘no longer did the problem of memory storage dominate man’s intellectual life’. Codification generates new opportunities for knowledge reproduction. For example, a written recipe is a ‘learning programme’ enabling people who are not in direct contact with those who possess the knowledge to reproduce it at a ‘lower’ cost. Goody (ibid., p. 143) writes: ‘The written recipe serves in part to fill the gap created by the absence of Granny, Nanna or Mémé (who has been left behind in the village, or in the town before last)’. ‘In part’ is the important term here. Naturally, codification mutilates knowledge. Getting the written recipe does not totally eliminate the learning costs. What is expressed and recorded is not complete knowledge; it is a learning programme that helps to reproduce knowledge. When a young technician receives a user’s manual, he or she is not directly given knowledge on ‘how to run the machine’. That said, the manual is helpful and will serve to reduce the costs of knowledge reproduction. In many cases, when technicians have ‘learned to learn’ and are dealing with a more or less standard machine, knowledge reproduction becomes almost instantaneous and assumes characteristics close to those of information reproduction. In more complex cases, however, the codified knowledge, while certainly useful, will provide only partial assistance. Knowledge reproduction will then occur through training, practice and simulation techniques (e.g. for aircraft pilots or surgeons). The other aspect of this function of codification concerns the locus of power in social institutions. Once again, Goody (1977) offers acute observations. Codification depersonalizes knowledge. The written recipe acquires independence from those who teach it. It becomes more general and universal. It reduces the relation of subordination between master and

272

Handbook of knowledge and economics

apprentice: the latter can learn when he or she has decided to do so and does not need to wait until the master is willing to teach. The important aspect of the function of codification is economic. Once a recipe has been written, it can be disseminated at a very low cost or even virtually free of cost, owing to new information technologies. This means that although the production cost of the first copy (basically, the codification cost) may be very high, the cost of all subsequent copies will rapidly decrease so that the codified knowledge can be reproduced and disseminated ad infinitum. It is clearly the codification of knowledge that changes the conditions of its circulation and that constitutes the condition on which advances in information technology can serve to improve that circulation until it is almost perfect. It should be noted that in our contemporary context, the recording of voices and images provides a means of ‘facsimile’ reproduction (referred to as form (c) in Figure 12.1). As such, facsimile recording involves no ‘higher-level’ codification of the structure or meaning of the recording. The important new ICT-based features that permit ‘illustration’ of these recordings, their deeper, second-order inscription, suggest new possibilities for the transmission of and distant access to all kinds of knowledge, far beyond the traditional forms of codified knowledge and written instructions. There is, thus, a sort of convergence (of course, far ahead of us) between various kinds of knowledge in terms of marginal cost of storage and transfer. In this sense, the traditional forms of codified knowledge are losing their singularity as a category of knowledge that is more appropriate than others for achieving the operations of storage and transfer at low marginal costs.

12.4 DISINVENTION AND DEACTIVATION OF EXPERIENTIAL KNOWLEDGE Experiential knowledge may disappear for several reasons, two of the main ones being: ●

It is local (it originates from specific experiences and applies to very particular contexts) and therefore fragile; few people have this knowledge, it is not codified to any great extent and can thus only be transmitted with difficulty; it disappears with the disappearance of those who possessed it and activated it. Private experiential knowledge often dies out in this way; the domain of artistic trades provides an abundance of examples.

The fragility of experiential knowledge 273 ●

It is disturbing and disruptive because its application would obstruct private or public interests. There was much talk, after the tsunami in 2004, of the role of mangroves and coral reefs as effective protective barriers against cyclones and large-scale storms. And yet this experiential knowledge did not carry much weight in the face of the economic interests represented by industrial prawn farming. Old avalanche maps have often remained in a drawer at the town hall when promoters came to propose the extension of building projects that overlapped into risk zones in the Alps.

The deterioration of experiential knowledge may not have any repercussions because its field of application was indeed very limited, or scientific enterprise takes an intellectual interest in reconstructing and reproducing it. This is the well-known case of the Stradivarius, whose technical mystery is gradually being revealed. The disclosing of the best-kept secrets is a very stimulating scientific game! In other domains, however, this deterioration will greatly jeopardize the community, henceforth deprived of certain experiential knowledge, or at any rate disrupt its functioning on a long-term basis. Urban, health, ecological or food crises are to a great extent created by the forgetting or depreciation of experiential knowledge concerning these different domains. The fragility of experiential knowledge is not necessarily expressed by its disappearance or disinvention, as is the case with certain valuable know-how that has vanished. It is also apparent in its deactivation: the knowledge is not lost in the literal sense but it no longer has the capacity to generate the actions that should result from it. The fact that physical activity is a good antidote to all sorts of illnesses and physiological problems is knowledge that has not been lost; this fact is known and affirmed time and again. But this knowledge has been ‘deactivated’, especially in certain areas and for certain types of social groups. This means that even if we acknowledge its pertinence, certain socioeconomic, cultural and societal changes mean that it is no longer possible (or has become very difficult) to apply the ensuing principles. The disappearance or deactivation of this knowledge may itself be hastened by belief in the omnipotence of scientific know-how that should enable the problems that will certainly arise to be corrected after the event. But the latter can under no circumstances be a substitute for the former. For example, it is very interesting to produce scientific knowledge to construct numerical models for predicting the speed and direction of the propagation of forest fires so that they can be more easily brought under control. However, this scientific knowledge does not replace the experiential

274

Handbook of knowledge and economics

knowledge accumulated over centuries that quite simply allowed forest fires to be avoided in the first place – knowledge mainly relating to the art of the planting and care of trees. And yet this experiential knowledge has deteriorated or been forgotten, although it is by no means obsolete.

12.5 ANALYTICAL FRAMEWORK: EXTERNALITIES, MARKET FAILURES AND INSTITUTIONS The examples provided above clearly illustrate the contrast between the strength and vigour of the processes of creating, codifying and circulating scientific know-how and the fragile nature of experiential knowledge. This contrast is characteristic of numerous domains: health, environment, food security, regional planning and development, and natural risk management. It is a dangerous illusion to believe that a society could function solely on the basis of scientific knowledge – the sort of society that would have all the possible ‘vaccines’ at its disposal to rectify problems and could therefore do without the experiential knowledge that is generally speaking applied beforehand to prevent these problems from ever occurring. The objective of the economics of knowledge is thus certainly not a society in which all the vaccines would be available, but a society in which the balance of the allocation of resources between scientific knowledge and experiential knowledge is properly protected. In contrasting the strength and vigour of the creation and management processes of scientific knowledge with the fragility of experiential knowledge, we must ask ourselves to what extent the difficulties of a market system to correctly allocate resources in the knowledge production and management domain are more serious or less well corrected in the area of experiential knowledge than in that of scientific knowledge. Economists tend to identify three generic causes of market failure (see for example, Swann, 2003). The first is that externalities drive a wedge between private and social returns from a particular private investment. If externalities are positive, some socially desirable investments will not appear privately profitable, so the market does not support enough activity. The second is where economic activities are subject to increasing returns. The third is that of asymmetric information. The usual working hypothesis is that the most important source of market failure that arises in the context of the production and management of knowledge is the existence of positive externalities from such activities. We can go further by distinguishing different types of externality that may give rise to these market failures.

The fragility of experiential knowledge 275 ●

●

Public-good externalities (Samuelson, 1954): a public good can be viewed as an extreme form of externality. It is defined according to two properties. First, it is difficult to exclude anyone from the benefits of a public good. Second, the marginal cost of enjoying the good is zero (consumption is non-rivalrous). Knowledge has both of these properties (it is difficult to exclude others from the benefits of the knowledge, and the marginal cost of an additional person making use of an idea is zero) and, as with all public goods, private markets are likely to provide an undersupply of knowledge.1 Ownership externalities (Bator, 1958): it is often difficult to attribute to a resource its real social value (the shadow value), which generates an inability to allocate resources correctly. This therefore concerns the difficulty of measuring the social value of a resource and thus attributing to it a series of results to which its contribution is difficult to observe. Certain ‘goods’ with determinate positive shadow values are not attributed, simply because ‘keeping account’ of who produces and who gets what may be impossible, clumsy or costly in terms of resources. Clearly, this is the case with knowledge. Evidence about the positive (direct and indirect) effects of the production of knowledge in the society and the economy is difficult to build and it is also difficult to try to measure returns on individual research projects.

The difference between these two types of externalities is that, in the latter case, the difficulty only concerns keeping account and might be eliminated if correct measurements were made and attribution devices set up. This is just a failure of enforcement (Bator, 1958), whereas the difficulty revealed in the case of the first externality (public good) cannot be eliminated – this is a failure of existence (ibid.). The public-good nature of the resource implies that the price that would encourage private agents to produce the optimal quantity of this good would inevitably be inefficient as regards the allocation of this resource: because the marginal cost of an additional person making use of the knowledge is zero, the maximization of allocative efficiency requiring prices equal to marginal cost will make the activity unprofitable. ●

Tyranny of small decisions (Kahn, 1966): a last type of externality seems important in the case of knowledge production and management. It expresses the fact that the market economy makes its major allocation decision on the basis of a host of ‘smaller decisions’ (smaller in size and time dimension). The tyranny of small decisions suggests that the total effect of small decisions may not be optimal

276

Handbook of knowledge and economics because the decisive determinations are individually too small – in terms of size, scope and time perspective. If 100 consumers choose option x, and this causes the market to make decision X (where X = 100 x), it is not necessarily true that those same consumers would have voted for X if that large decision had ever been presented for their explicit consideration. According to Weisbrod (1964), there is an externality and hence a market failure when: (a) the option is not (always) exercised; (b) revenues from actual purchasers are insufficient to cover the costs of continued operation; and (c) expansion or restarting of production at the time when occasional purchasers wish to make a purchase is difficult or impossible. The external benefit here is the mere availability of the service (or the knowledge) to non-users, the continued ability to satisfy as yet unexerted option demand. The deterioration or even disappearance of a great deal of knowledge is often caused by this externality.

These different types of externality apply to both the scientific knowledge and experiential knowledge domains. However, science has developed institutions allowing them to be corrected or their effects to be attenuated, and it would therefore be the presence of these institutions in one case and their absence in the other that could explain the contrast between the vigour and power of the creation and circulation processes of scientific knowledge and the fragility of experiential knowledge. An additional step is thus necessary to explain this contrast. This consists of identifying and assessing the institutional solutions that science benefits from but that are not applied in the other case. First, the institutional solutions allowing the problem of public-good externalities to be attenuated are well known. Pigou was the first to identify the three mechanisms for providing public goods: subsidies, directed governmental production and regulated monopoly. These three mechanisms have a clear application in the domain of scientific knowledge production and R&D. The first mechanism consists of the government engaging itself directly in the production of knowledge; the second mechanism is one where production is undertaken by private agents, who in turn are subsidized for their effort by the public purse. The third mechanism is to establish a competitive market mechanism for some type of knowledge to which private ownership can be legally assigned and whose ownership can be enforced (Dasgupta, 1988; David, 1993). Second, the scientific institution has given rise to the development of institutional mechanisms to evaluate and even measure the intrinsic value of new scientific knowledge. According to the historical analysis of Paul David (2007), the competition for the ‘best’ scientists between potential

The fragility of experiential knowledge 277 patrons required open science as a solution to the asymmetric information problem that the patrons faced, namely to identify the truly leading scientists of their generation. Only within communities in which full disclosure was exercised could the scientific findings be evaluated and discussed and credible reputations be established that would allow wealthy patrons to identify truly distinguished scientists from fraudulent ones. From these historical origins the institutionalization of effective mechanisms for the systematic evaluation of scientific knowledge took place as a valuable solution to the Bator ownership externalities. Finally, as in the case of many other types of public good, decisions regarding the allocation of resources to scientific research are delegated to administrative entities operating at the appropriate levels – usually at the national level but also in certain domains at the supranational (case of the CERN) or global (case of the IPCC) levels, which allows the ‘tyranny of small decisions’ effects to be avoided. Developing and protecting appropriate levels of decision makers in science means that the fundamental allocation choices do not occur by surprise ex post – as the result of a series of individual micro-decisions – but are made ex ante. And yet none of these mechanisms seems to exist in the experiential knowledge domain. Or rather, you have to search thoroughly for a long time to discover more or less hidden mechanisms likely to offer solutions to the problems of externalities – and that is the objective of the rest of the chapter.

12.6 DISCUSSION If experiential knowledge and scientific knowledge are not substitutable, as is claimed above, then the disinvention problem may be socially costly if the experiential knowledge considered was valuable in certain contexts and circumstances and these circumstances are likely to recur. It is therefore useful to identify some potential solutions and institutions that can be relied upon to maintain, reproduce and exploit experiential knowledge – that is institutions that can sustain an efficient ‘infrastructure’ to reproduce this particular type of knowledge. In view of these problems, the museum solution rapidly springs to mind. Societies have built and sustained institutions – such as libraries, archives and museums – to collect, organize and provide access to knowledgebearing objects for more than two millennia (Hedstrom and King, 2006). It is, therefore, legitimate to think of a potential role for this so-called ‘epistemic infrastructure’ when issues of knowledge loss and of disinvention need to be addressed. This calls to mind the UNESCO project aimed

278

Handbook of knowledge and economics

at setting up a world bank in Florence to make an inventory of, safeguard and promote traditional know-how.2 This is certainly a laudable objective; however, as already stated, experiential knowledge involves more than a mere catalogue of traditional techniques. The main question is therefore less that of the creation of libraries or academies than that of the capacity of living communities to adapt and utilize their experiential knowledge within the framework of their current socioeconomic activities – in other words, to attribute a certain economic value to this knowledge. Museums are no doubt necessary but under no circumstances sufficient to obtain the appropriate balance between scientific know-how and experiential knowledge. Of course the codification of experiential knowledge is an important tool. As has been demonstrated above, the composition of a script and its codification provide societies with stronger capabilities for memory, communication and learning. However, as for any economic operation, agents respond to incentives: costs and benefits will explain the decision to codify, at least in the case of ‘codifiable but not yet codified knowledge’ (Cowan et al., 2000).3 This is where price considerations come in as well as the expected private and public value of the codified form of the experiential knowledge. Viewed in this perspective, the demand for codification is influenced by a set of factors, including institutional arrangements affecting the structure of incentives for codification activities. They also concern the state of technology, which determines codification costs. This position on the endogeneous nature of boundaries between tacit and codified knowledge and the importance of economic determinants is in fact very similar to that of Nelson and Winter (1982). Thus knowledge codification, like all other knowledge memorization and management processes, is a consequence of economic dynamics rather than its cause. So the main issues to be addressed concern not so much the mobilizing of the epistemic infrastructure or proceeding to massive codification, but for the experiential knowledge to regain its vigour and strength through overcoming some of the most significant market failures that create inefficiencies in the way experiential knowledge is produced, managed, distributed and used. The main issue is for any piece of experiential knowledge to regain its former status: instruments and tools give the individual and the community the capacity for effective action in the current socioeconomic contexts. From this perspective, we can identify three principal logics regarding the utilization and valorization of experiential knowledge. There is no oneto-one correspondence between each of these processes and a particular type of externality previously mentioned. Each of these processes must rather be examined as creating organizational forms appropriate for the

The fragility of experiential knowledge 279 ‘treatment’ of the different types of externality considered that may contribute to making experiential knowledge more fragile. The first logic is that of industrial coordination, a process whereby a firm observes that certain of its products or services are closely dependent on the survival of a certain type of experiential knowledge possessed by a very small number of individuals, and invents an industrial organization mechanism to maintain this knowledge. The jewellery, fashion, shoe, watch or fine leather goods industries offer good examples of such situations: there is a collection of traditional and high-tech knowledge and know-how deeply embedded in local industrial materials that forms the foundations of the sector’s innovation and competitiveness. However, most of such knowledge is particularly fragile. Rare professions, guardians of technical knowledge and precious know-how, they are threatened because they often depend on tiny markets. The small scale of the markets complicates all the problems regarding vocational training and human capital, financing of possible modernization projects and so on. And yet, for some industries, these artistic trades often play the role of weakest link: their undermining or decline creates a risk for the industry as a whole. Companies in these industries therefore have a strategic interest in undertaking actions aimed at consolidating certain activities over the long term. By maintaining dense subcontractor networks oriented towards and paid for high-quality goods and services, these industries contribute to the perpetuation of memory capacity and transmission of traditional knowledge and know-how. They enable this knowledge to be preserved in both time (memorization and transmission) and space (territorialization of this knowledge, which remains very difficult to relocate in ‘cheaper’ places). This constitutes an essential contribution to the issue of maintaining fragile knowledge in so far as the disinvention, forgetting and loss of knowledge are real phenomena and costly in the long run, linked particularly with the tacit nature of this knowledge and know-how and the fragility of its ecosystem. But the question of the maintenance, preservation and transmission of practical knowledge and know-how relating to certain artistic trades is a difficult one and remains mysterious: what kind of organizational models can sustain such preservation? The externality at stake here is the third form (see above). There is an externality because (a) the option (of contracting for the service of a particular artistic trade or craftsman) is not (always) exercised; (b) revenues from actual users are insufficient to cover the costs of continued operation; and (c) expansion or restarting of production at the time when occasional users wish to make a purchase is difficult or impossible. The external benefit here is the mere availability of the knowledge to non-users, the continued ability to satisfy as yet unexerted option demand (Weisbrod, 1964).

280

Handbook of knowledge and economics

Economists observe that what is at stake in such cases is the provision of a public good specific to an industry. They argue that in this situation no economic case can be made for a public finance solution: while the experiential knowledge really is a public good, its industrial specificity implies that there are no strong arguments for the state to assume responsibility for its production. It is therefore the task of the industry concerned to develop the institutions that will enable these public goods to be maintained. In such circumstances, a typically ‘Coasean’ response would involve the implementation of a club-good solution. A club-good response to the externality argument is to ask why the diverse beneficiaries from the knowledge cannot form a club to fund it. If a sufficient number join together for their joint benefits to exceed cost, then the knowledge will be funded – even if the club does not capture all the benefits. In short, the club solution is to internalize the externalities (Weder and Grubel, 1993). The second logic is that of innovation. We can be brief here, since this is the most obvious case of the revitalization of experiential knowledge. Innovation here refers to the sense of the transposition and adaptation of an existing experiential knowledge to new contexts; there are plenty of cases of such innovations that allow some kind of experiential knowledge not only to survive but to enjoy new roles and functions in specific communities and contexts – to be revitalized. For example, if studied and properly disseminated, there is a wealth of experiential knowledge in management of water resources (produced by people who have learnt to operate limited water resources in a sustainable manner) that can reduce common errors in policy, technical and managerial practices while giving valuable hints to those seeking innovative approaches to the problem of dealing with water scarcity. However, systems of innovation are not taken for granted. The defining characteristics of a system of innovation require that its components – including in our case the experiential knowledge holder – are connected for different invention and innovation purposes. When the components of the knowledge ecology are not connected, there is an ecology but not a system (David and Metcalfe, 2008). Therefore systems of innovation emerge or not as the elements of the ecology interact to further the innovation process. The difficulties and obstacles to innovation are therefore largely to do with barriers and incentives to collaborate in the solution of problems. Innovation policy seen from this perspective has a responsibility to frame the institutional architecture and the structures of regulatory constraints and rewards available to present and future researchers, entrepreneurs, managers and those attempting to master the experiential knowledge in a way that allows sufficient flexibility and mobility to stimulate and reinforce the connections that transforms the ecology into adaptive innovation systems.

The fragility of experiential knowledge 281 The third logic is about integration, a process whereby a scientific institution recognizes the potential value of experiential knowledge as a complement to the scientific knowledge that it produces and implements mechanisms to identify, collect, codify and use this knowledge. Some cases of integration or combination between scientific knowledge and experiential knowledge have already been documented in the literature. The most famous example is no doubt the case of the parents of sick children who collaborate with medical research by regularly and systematically collecting data thanks to their privileged point of view (Rabeharisoa and Callon, 2002). There are also several cases in the natural sciences domain where research succeeds in mobilizing the experiential knowledge of hundreds of volunteers to collect and assemble data concerning a particular animal or plant species. Finally, there is an increasing effort to gather traditional knowledge in order to compare it and integrate it with scientific knowledge in certain areas of environmental management. In short, we can clearly distinguish two logics for the mobilization of experiential knowledge within the framework of a scientific approach. On the one hand, the scientific institution realizes that amateurs and laypersons who are ‘in contact’ with a particular environment or phenomenon form an extraordinary set of distributed capacities for data collection. It is therefore up to the scientific institution to organize this collection and then integrate the data while devising an organization to facilitate the system’s continuity. A superior logic is undoubtedly to acknowledge that persons in contact are not only useful as collectors but have developed experiential knowledge, expertise that is admittedly local and non-scientific but rigorous and rational, enabling them to formulate hypotheses and strategies, test them and thus broaden the variety of possible options, for example in terms of treatment of the considered subject (whether an ecosystem or a sick child is involved). This second logic is far more demanding regarding the involvement of both the scientific institution and the amateurs and laypersons that possess this pertinent experiential knowledge. For example, the Institute for Snow and Avalanche Research (SLF) in Davos (Switzerland) is operating in an absolutely exceptional domain for studying integration mechanisms between scientific knowledge and experiential knowledge. Indeed, first, the scientists themselves consider that their knowledge is not sufficient: With respect to promoting avalanche safety, snow and avalanche research has primarily focussed on developing a better understanding of the physical processes that lead to avalanches. While this knowledge has provided tremendous insights for improving our understanding on why avalanches occur and how they are triggered, there is still significant uncertainty associated with this knowledge. On the other hand, mountain guides have been making decisions about

282

Handbook of knowledge and economics

when and where to travel in avalanche terrain for centuries and have therefore tremendous practical experience in dealing with this uncertainty. (Pascal Haegeli, professor at Simon Fraser University, personal communication)

And second, those who possess experiential knowledge are more often than not motivated and very enthusiastic to collaborate with scientists and share their knowledge (ibid.). The Davos Institute has long been implementing the first logic for mobilizing experiential knowledge. It has set up a decentralized data collection system concerning snow conditions and avalanche risks that is composed of volunteers living in mountainous areas. The progression to the second logic is more complicated, but does really occur: ‘Sometimes, I have started research projects because practitioners have repeatedly asked me, or they put forward hypotheses I wanted to challenge’ (extract from an interview with J. Schweizer, SLF Director). The issues of industrial coordination, innovation and integration place experiential knowledge in a very different context from that of conserving the legacy of know-how extracted from the past. It is a matter of finding the right economic institutions to generate the motivations and incentives that will allow certain economic agents, certain communities, certain industries and professions to allocate resources to the management, distribution and exploitation of such experiential knowledge. It also means devising the institutional mechanisms necessary for the relevant experiential knowledge, possessed by non-experts, to be better represented in political decision-making processes. This particularly concerns experiential knowledge that is relevant regarding a particular political or socioeconomic decision. The acknowledgement and integration of experiential knowledge in political decision-making processes are no doubt an essential factor in creating a better link between the broadest popular participation – the very foundation of democracy – and expert skills – a question already posed by John Stuart Mill in his famous work Considerations on Representative Government (2004).

12.7 CONCLUSION Our starting point was the identification of the different types of externality that create problems regarding resource allocation to knowledge creation and management and the different institutional solutions adopted in the scientific research and R&D domain to correct these problems. Then we discussed the different solutions that have been discovered and analysed in the experiential knowledge domain.

The fragility of experiential knowledge 283 Whereas the notion of experiential knowledge is an important one, allowing numerous problems and anomalies in the economics and management of knowledge to be explained (fragility, forgetting, disinvention, social loss), this notion has been very little analysed from the economic and management sciences point of view. It is on the other hand well understood by other disciplines, especially the environmental sciences, which are reflecting on the importance and difficulties of combining scientific knowledge and experiential knowledge to manage a particular ecological resource or support a particular ecosystem (Donovan and Puri, 2004; Rist et al., 2010). However, lacking the tools of modern microeconomics and the management sciences applied to the problem of knowledge management and optimization, these disciplines cannot get very far with their investigations regarding the conditions, methods and procedures of experiential knowledge management. This is why an economic approach is important, with its objective of developing the economics and management of fragile knowledge, whose contribution to these other disciplines could prove valuable.

NOTES 1. A whole series of phenomena exists that mitigate the public-good nature of knowledge while not altering the economic logic of the argument. In particular, although it is correct to recognize that developing human capability to make use of knowledge involves processes that entail fixed costs, the existence of the latter does not vitiate the proposition that reuse of the knowledge will neither deplete it nor impose significant further marginal costs. 2. See www.tkwb.org. 3. The ‘codifiability’ of knowledge depends on the existence of appropriate languages, printing technologies and modelling capabilities for the knowledge under consideration.

REFERENCES Bator, F. (1958), ‘The anatomy of market failures’, Quaterly Journal of Economics, LXXII, 351–79. Cowan, R., David, P. and Foray, D. (2000), ‘The explicit economics of knowledge codification and tacitness’, Industry and Corporate Change, 9 (2), 211–53. Dasgupta, P. (1988), ‘The welfare economics of knowledge production’, Oxford Review of Economic Policy, 4 (4). David, P.A. (1993), ‘Knowledge, property and the system dynamics of technological change’, World Bank Annual Conference on Development Economics, pp. 215–48. David, P.A. (2007), ‘The historical origins of “open science”: an essay on patronage, reputation and common agency contracting in the scientific revolution’, Stanford University and University of Oxford. David, P.A. and Metcalfe, S. (2008), ‘Only connect: academic–business research collaboration and the formation of ecologies of innovation’, SIEPR Discussion Paper 0733, Stanford Institute for Economic Policy Research, Stanford University.

284

Handbook of knowledge and economics

Donovan, D. and Puri, R. (2004), ‘Learning from traditional knowledge of non-timber forest products’, Ecology and Society, 9 (3), http://www.ecologyandsociety.org/vol9/iss3/art3. Foray, D. (2006), The Economics of Knowledge, Boston, MA: MIT Press. Goody, J. (1977), The Domestication of the Savage Mind, Cambridge, MA: Cambridge University Press. Hedstrom, H. and King, J. (2006), ‘On the L.A.M.: libraries, archives and museums as epistemic infrastructure’, in B. Kahin and D. Foray (eds), Advancing Knowledge and the Knowledge Economy, Boston, MA: MIT Press, pp. 113–34. Kahn, A. (1966), ‘The tyranny of small decisions: market failures, imperfections and the limits of economics’, Kyklos, 19, 23–47. Mill, J.S. (2004), Considerations on Representative Government, Pennsylvania State University,  Electronic Classis Series, www2.hn.psu.edu/faculty/jmanis/jsmill/considera tions.pdf. Nelson, R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Belknap Press of Harvard University Press. Polanyi, M. (1967), The Tacit Dimension, Garden City, NY: Doubleday Anchor. Rabeharisoa, V. and Callon, M. (2002), ‘L’engagement des associations de malade dans la recherche’, Revue Internationale des Sciences Sociales, no. 171, 65–73. Rist, L., Shaanker, U., Milner Gulland, E. and Gazoul, J. (2010), ‘Combining traditional knowledge and scientific data in forest management’, United Nations University, IAS. Samuelson, P. (1954), ‘The pure theory of public expenditure’, Review of Economics and Statistics, 36. Swann, P. (2003), ‘Funding basic research: when is public finance preferable to attainable club-goods solutions?’, in A. Guena, A.J. Salter and W.E. Steinmueller (eds), Science and Innovation: Rethinking the Rationales for Funding and Governance, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 335–74. Weder, R. and Grubel, H. (1993), ‘The new growth theory and Coasean economics institutions to capture externalities’, Weltwirtschaftliches Archiv, 129 (3). Weisbrod, B. (1964), ‘Collective-consumption services of individual consumption goods’, Quaterly Journal of Economics, LXXVIII, 475–86.

13 One knowledge base or many knowledge pools? Bengt-Åke Lundvall

13.1 INTRODUCTION In Lundvall (1992) I started the analysis of innovation systems from a characterization of the current state of the economy as one where ‘knowledge is the most important resource and learning the most important process’. But this declaration was not given much analytical backing in the book. I did not give much insight into how knowledge and learning relate to innovation and to economic performance. In this chapter I present a conceptual framework and make distinctions between private/ public, local/global, individual/collective and tacit/codified knowledge. The purpose is both ‘academic’ and practical. My analysis demonstrates the limits of a narrowly economic perspective on knowledge but I also show that these distinctions have important implications both for innovation policy and for management of innovation. The chapter introduces a conceptual framework for analysing knowledge in relation to economic development. It does so through a critical analysis of the perception that the economy has ‘a knowledge base’. Over the last decade it has become commonplace among policy makers to refer to the current period as characterized by a knowledge-based economy and increasingly it is emphasized that the most promising strategy for economic growth is to strengthen the knowledge base of the economy (Abramowitz and David, 1996; Foray and Lundvall, 1996; OECD, 2000). This discourse raises a number of unresolved analytical issues. What constitutes the knowledge base? At what level can we locate and define a knowledge base? I shall show that the idea of ‘one knowledge base’ is misleading and that the kind of knowledge that matters for the economy should rather be regarded as many separate ‘pools’, each with limited access. Using the standard terminology of economics, most knowledge is neither a strictly private nor a strictly public good (Arrow, 1994). Rather than regarding such knowledge pools as individually owned assets, we should see them as constituting ‘community resources’ not easily transformed into private property. This perspective gives new inspiration for innovation policy both in the more and in the less developed part of the world. In developing countries 285

286

Handbook of knowledge and economics

there is a need to build absorptive capacity in order to get access to the knowledge pools in the rich part of the world and there is a general need to reconsider the rules of the game regarding intellectual property rights. In rich countries, finding ways of connecting specific separate pools of knowledge may be seen as a key to stimulate radical innovation through exploiting knowledge diversity.

13.2 THE ECONOMICS OF KNOWLEDGE Knowledge and information appear in standard economic models in two different contexts. One relates to decision-making. The other context is one where knowledge appears as technology that is transformed into techniques used by firms to produce scarce tangible products. It is of interest to consider how knowledge is treated in these two contexts in the old and the new neoclassical economics. Both the old and the new neoclassical economics stick to the assumption that agents are rational – and sometimes hyperrational (i.e. assuming rational expectations). Here information is free and costless. The old neoclassical growth theory consistently regards technological change as exogenous. New technologies appear for reasons not spelled out in the theory, and access to the technical knowledge is assumed to be unlimited. The image often used is a ‘book of blueprints’ free for use. The new growth theory and the new trade theory are more ambitious in their attempts to explain technological change. In new growth theory, the output of the R&D sector is viewed either as a blueprint for a new production process more efficient than the previous one; it is assumed that it can be protected by private property instruments such as patents; or as a new semi-manufactured goods not easily copied by competitors (Verspagen, 1992, pp. 29–30). But the assumption of rational agents is not fundamentally revised. Firms are assumed to optimize their investment in new products and processes. The old neoclassical economics is consistent in treating knowledge as exogenous but it has nothing to offer in terms of explaining knowledge production and use. The analytical framework is valid for a stationary state corresponding to Schumpeter’s ‘circular flow’. In such a state all agents may have established access to all information they need and there is neither new information nor new technology emerging that needs to be explained. The new neoclassical theory is not consistent in this respect. It operates on the basis of two types of knowledge with opposite characteristics. Some information – the information needed to make a decision – is not at all scarce; other elements of knowledge can be obtained only by investing or

One knowledge base or many knowledge pools? 287 buying. The only solution to this inconsistency would be to assume that the information utilized by decision makers and the technological knowledge emanate from two different universes – one static and one dynamic. The other major problem for standard economics when analysing knowledge is the dictum of methodological individualism. It is assumed that the agent operating as decision maker and as owner of knowledge is an individual. Either knowledge is the property of an individual or it is accessible by all individual agents. The fact that fundamental elements of knowledge are shared without being public cannot be captured without abandoning methodological individualism. Languages, common codes, trust relationships, shared routines and standards shared within a community cannot be reduced to assets and neither can they be transacted in the market. But since access to them is restricted to a community, neither are they public goods. The basic starting point that everything, including knowledge, is an asset and potentially ‘property’ that can be transacted in markets makes standard economics less well suited for analysing knowledge. Evolutionary economics does not suffer from such inconsistencies. Bounded rationality and innovation are core elements in the theory, and so are shared routines (Nelson and Winter, 1982). Evolutionary economics is therefore much better suited to analyse an economy where knowledge is a key resource. 13.2.1

Rational Choice

The very foundation of standard economics is the analysis of rational choice made by individual agents. Thus, how much and what kind of information agents have about the world in which they operate and their ability to process the information are crucial issues that draw the lines between major economic schools. While neoclassical economics sticks to the assumption that agents are rational – and sometimes hyperrational (i.e. assuming rational expectations) – Austrian economists (Hayek, 1937), Keynesians (Keynes, 1936) and organizational economists such as Herbert Simon (1957) assume bounded rationality – a combination of a complex and uncertain environment and agents with limited capacity to process information. It is obvious that agents do make choices between well-defined alternatives from time to time and that in some contexts these choices involve a calculation of costs and benefits in order to find the alternative that is most attractive in economic terms. But agents not only make choices, and they do not make choices all the time. Actually, most of the time agents follow routines or they do things without considering their own costs and

288

Handbook of knowledge and economics

benefits in economic terms. Many ‘consumption activities’ take place in interaction with others and the ‘utility’ of these activities is highly dependent on what pleasure others get from the activity – this is true for sports, love-making and most cultural and social activities (Mead, 1934). As societies grow rich, these types of activities may become even more important. This kind of argument is however lost on neoclassical economists. To give up the basic assumption about ‘rational behaviour’ would imply giving up or rethinking even their most fashionable tools, including game theory. Therefore I shall only state here that in a world where agents are involved in innovation and where innovation is important for economic performance – and this seems to be the case in the world we live in – the idea of explaining economic dynamics by models assuming that agents know all possible outcomes is not reasonable. Innovation is by definition a process where the alternative outcomes cannot be defined in advance – if they could be defined in advance we would not regard it as innovation. Actually we might say that not only do we operate under fundamental uncertainty in the modern economy – we operate under radical fundamental uncertainty (the antithesis of rational expectation hypothesis): the only thing we know for certain is that we should constantly expect surprising outcomes of our decisions. We know that new technologies, new patterns of consumer behaviour and new forms of organization that we cannot define in advance will emerge. Whatever the usefulness of neoclassical economics, it is not its capacity to explain what is going on in an economy where knowledge is the most important resource and learning the most important process. 13.2.2

Is Knowledge a Public or a Private Good? Or is it a Community Resource?

The other context where knowledge appears in economics is as an asset that can be owned, exchanged and reproduced. It may also appear as input or output in an economic process. In economic theory, the properties that give a good the attribute of ‘public’ are the following: ● ●

the benefits can be enjoyed by many users concurrently as well as sequentially without being diminished – the good is non-rival; it is costly for the provider to exclude unauthorized users – the good is partially non-excludable.

One reason for the interest in the public-good issue is that it is crucial for defining the role of government in knowledge production. If knowledge

One knowledge base or many knowledge pools? 289 were a public good, freely accessible to anyone, there would be no (economic) incentive for private agents to invest in its production. More generally, if it is less costly to imitate than to produce new knowledge, the social rate of return would be higher than the private rate of return and, again, private agents would under-invest in the production of knowledge. Nelson’s (1959) and Arrow’s (1962b) classical contributions demonstrated that, in such situations, there is a role for government either to subsidize or to take charge directly of the production of knowledge. Public funding of schools and universities, as well as of generic technologies, is rational according to this analysis. It also brings to the fore the legal protection of property rights to knowledge, for instance by patent systems. The analysis of knowledge as a private or public good may be contrasted with another perspective with roots further back in economic theory. Marshall (1923) made the observation that firms belonging to the same sector tended to be located together in ‘industrial districts’. He also found that such groupings of firms often remained competitive for very long periods. He said that ‘the secrets of industry are in the air’ and specified by pointing to skills in the local labour force and in local specialized institutions inherited from one generation to the next. These two perspectives are opposed not only because the first points to the need to protect knowledge while the second points to the need to support the diffusion of knowledge. The industrial district perspective represents a radical break with the neoclassical analysis since here knowledge is neither private nor public in the neoclassical sense. Since there is no simple way to enforce private ownership of the regional knowledge – to privatize the benefits from the knowledge commons – it is more correct to refer to knowledge not as an asset but as a non-marketable ‘community resource’. In standard economics, this kind of phenomenon might be referred to as externalities or agglomeration effects. But since they represent ‘typical’ forms of economically important knowledge, their emergence and development need to be understood not as exceptional phenomena but as regular outcomes of socioeconomic processes. But the two perspectives raise similar questions. Can knowledge be transferred from one place to another? How difficult is it to transfer knowledge and what are the transfer mechanisms? Is it possible to change the form of knowledge (for instance through codification) so that it gets easier (more difficult) to transfer? One reason for the distinctions between different kinds of knowledge proposed below is that they help to sort out these questions. Responding to these questions is also a way of specifying how the knowledge base of the economy is constituted. If knowledge were completely public, it would be meaningful to speak of one common knowledge

290

Handbook of knowledge and economics

base for the whole economy and there would be a strong need for coordinating investments in knowledge production at the global level. If, conversely, knowledge were completely individual and private, there would be no common knowledge base at all and investment in knowledge production could be left to the individuals themselves. As we shall see below, as often is the case, reality is complex and most knowledge is neither completely public nor completely private. Some knowledge is ‘in the air’ locally but cannot easily be moved out of the local context. The knowledge base is fragmented and may best be illustrated as constituted by a number of semi-public ‘community pools’ with shared access regionally, professionally or through networking. Limited access means that some are excluded from even approaching these pools, while others with formal and legal access may lack the necessary tools to tap into them. The last decades have witnessed two different tendencies that tend to change the private/public character of knowledge. On the one hand, the widening use of information technology and the increasing importance of communicating scientific knowledge in the economy have given strong incentives as well as more efficient instruments when it comes to codifying knowledge and making it explicit. This tends to make knowledge more widely accessible. But on the other hand, and to some degree as a response to the codification trend, there has been a strong political push in favour of de facto and legal protection of intellectual property. Led by the big US companies that operate in science-based sectors, multinational companies worldwide have successfully lobbied for more broad and strict legal protection first at the national level and later at the global level. This means that there are at least two different types of barrier around knowledge pools. Some reflect that competence is unequally distributed in space and lack of absorptive capacity. Others reflect political power and legal institutions, denying access to those without formal ownership. The first type may in principle be overcome by investment in knowledge and competence building and by joining networks. The second type of barrier will reflect the use of political power in negotiations. Both these barriers tend to reinforce the inequality of the distribution of knowledge, and without a global ‘new new deal’ with focus on access to knowledge the world will tend to become increasingly polarized. 13.2.3

A Terminology of Knowledge

In 1987, Sidney Winter concluded an important contribution on knowledge and management by pointing out that there is ‘a paucity of language’ and ‘a serious dearth of appropriate terminology and conceptual schemes’

One knowledge base or many knowledge pools? 291 for analysing the role of knowledge in the economy (Winter, 1987). Since then, the number of relevant publications has grown immensely and some progress has been made (Foray, 2000; David and Foray, 2002; Amin and Cohendet, 2004), but as compared to Winter’s original analysis, little headway has been made in terms of a terminology acceptable to all. There is little agreement on questions such as: what is the meaning of knowledge? What separations and distinctions between different kinds of knowledge are most useful for understanding how knowledge affects economic development? 13.2.4

Knowledge of the Mind, Body and Soul

One classical distinction is the one between data, information, knowledge and wisdom. Data are the raw materials used to construct information. Knowledge is necessary to get meaning from information. Wisdom is needed in order to use knowledge. These distinctions may be useful for some purposes but the selection and ordering of them reflect a bias in the way knowledge is regarded. The focus is on mental cognition, where knowledge is something absorbed through access to information and analysed in the mind. An alternative understanding of knowledge is represented by the philosophy of Dewey and the pragmatist Chicago School (Campbell, 1995). Here knowledge is seen as emanating from practice and as layered not in a separate mind but in the body as a whole (Kolb, 1984). In modern society there are many specialized organizations and institutions that promote the training of the mind. Some of them are actually based on the assumption that knowledge is acquired through getting access to information, and those who operate them may see learning as a cognitive process separated from practice. But closer analysis of scientific work shows that practice is fundamental to learning and that expert knowledge is located in the body. Actually you might define expertise as a stage where you do not have to go through a stepwise analytical process but can draw directly upon ‘back-bone’ knowledge. There is a tension between mind and body in the historical and analytical understanding of knowledge. The distinction between explicit knowledge (mind) and tacit knowledge (body) illustrates this tension, and so does the organisation of economic activities where certain parts of organisations (R&D department) may be seen as mainly concerned with cognition (mind) while other parts (production departments) are seen as involved in practical action (body). In fact there is no learning and knowledge without involving the body and very limited learning in societies without cognitive attempts to make explicit elements of tacit knowledge. I

292

Handbook of knowledge and economics

shall use the distinction between tacit and codified knowledge below, but do so in order to demonstrate how the two are intertwined and mixed differently in different contexts. In order to see the close interdependence of the two dimensions, it is necessary to make the distinction between them. 13.2.5

The Social Character of Knowledge

Each individual may be seen as a ‘knowledge container’, but what she contains will be more or less useful and meaningful in different social contexts. The knowledge of a professor in physics may be of little use as participant in a safari in Africa, while a professional lion-hunter may be of little use at a scientific conference in Boston. So the context makes all the difference. Most meaningful knowledge is constructed in interaction with others and gets a meaning in interaction with others. Interaction involves communication and cooperation. Communication may be oral or take place via gestures or with the help of artefacts. It may take place in a more or less structured context – as speech or as conversation. Cooperation may be more or less purposeful as work or as play (Amin and Cohendet, 2004). Knowledge of importance for economic purposes may be rooted in the relationships within a team, in routines common to the firm or in a wider community extending outside the borders of the single firm. The team may be a formal working unit but it may also be a self-organizing community. Here I shall refer to it as a ‘community team’; I prefer to reserve ‘community’ tout court for social formations that cross the border of the firm. Community teams and communities may be based primarily upon epistemology or primarily upon practice. In the first case the major aim is to process and produce knowledge, while in the second it is to find workable solutions to a set of practical problems requiring skills that are interrelated (Wenger, 1998). But not all community knowledge remains a community resource. There are strong incentives for capitalist firms to transform community knowledge into private property. Inside organizations, codification of the skills, cooperation and interactions of employees may increase management control with the core knowledge of the firm. In relation to competitors there is a strong incentive both to protect key elements of the firm’s technology from access and more generally to block competitors’ access to strategic elements of knowledge. This lies behind one of the major contradictions of current capitalism. On the one hand, capital wants to subordinate knowledge under its own rule and transform it into private property. On the other hand, knowledge thrives in communities and communities cannot be fully subordinated to capital and their knowledge transformed into private property without

One knowledge base or many knowledge pools? 293 losing their effectiveness as sites for knowledge creation and reproduction. The contradiction between modes of innovation that are collective, open and interactive, and modes of appropriation of knowledge that aim at privatizing the outcomes of innovation is fundamental in the capitalist knowledge-based economy (Allen, 1983; Cowan and Jonard, 2003; Chesborough, 2003; von Hippel, 2005). While there are historical periods where we see tendencies toward knowledge sharing in certain areas characterized by rapid technical change or by rapid change in user needs, such tendencies may be reversed as the rate of change slows down. 13.2.6

Four Different Kinds of Knowledge

Following Lundvall and Johnson (1994), knowledge is here divided into four categories: ● ● ● ●

Know-what Know-why Know-how Know-who.

‘Know-what’ refers to knowledge about ‘facts’. The number of people who live in Beijing, the ingredients necessary to make pancakes and the year of the French Revolution are examples of this kind of knowledge. Here, knowledge is close to what is normally called information – it can easily be broken down into bits and communicated as data. ‘Know-why’ refers to knowledge about principles and laws of motion in nature, in the human mind and in society. This kind of knowledge has been extremely important for technological development in certain science-based areas, such as the chemical and electric/electronic industries. Access to this kind of knowledge will often make advances in technology more rapid and reduce the frequency of errors in procedures involving trial and error. ‘Know-how’ refers to skills – that is, the ability to do something. It may be related to the skills of artisans and production workers, but, actually, it plays a key role in all economic activities. The businessman judging the market prospects for a new product or the personnel manager selecting and training staff use their know-how. It would also be misleading to characterize know-how as practical rather than theoretical. One of the most interesting and profound analyses of the role and formation of knowhow is actually about scientists’ need for skill formation and personal knowledge (Polanyi, 1958/1978). Even finding the solution to complex mathematical problems is based on intuition and on skills related to

294

Handbook of knowledge and economics

pattern recognition rooted in experience-based learning rather than on the mechanical carrying out of a series of distinct logical operations (Ziman, 1979, pp. 101–2). Know-how is often developed individually through experience and kept within the borders of the firm or team. As specialization evolves and complexity of technology and science increases, however, cooperation between people and organizations and knowledge sharing become increasingly necessary (Pavitt, 1998). The more fine-grained and the deeper the division of labour among experts, the more crucial become the mechanisms that link different fields of expertise to each other. This is why ‘know-who’ becomes increasingly important. Know-who involves information about who knows what and who knows how to do things. But it also involves the social ability to cooperate and communicate with experts. Know-who makes it possible to draw upon intellectual capital through the means of social capital. 13.2.7

How Public or Private are the Four Kinds of Knowledge?

The public or private character of these kinds of knowledge differs in terms of both degree and form. Databases can bring together ‘know-what’ in a more or less user-friendly form. Information technology extends enormously the information potentially at the disposal of individual agents, although the information still has to be found and what is relevant selected. The effectiveness of search engines developed in connection with the Internet is highly relevant in this context, as it helps to specify how accessible the data actually are. Even with recent advances in this area, access to this kind of knowledge is still far from perfect (Shapiro and Varian, 1999). Still today, the most effective medium for obtaining pertinent facts may be through the ‘know-who’ channel, that is, contacting an outstanding expert in the field to obtain directions on where to look for a specific piece of information. Scientific work aims at producing theoretical models of the ‘know-why’ type, and historically much of this work is placed in the public domain. Academics have strong incentives to publish and make their results accessible. The Internet offers new possibilities for speedy electronic publishing. Open and public access is of course a misnomer, in that it often takes enormous investments in learning before the information has any meaning. Again, know-who, directed towards academia, may help the amateur to obtain a translation into something more comprehensible. In some areas where new technological opportunities evolve very quickly and technological competition is intense, the technical solutions introduced by engineers may get far ahead of academic know-why

One knowledge base or many knowledge pools? 295 (Vincenti, 1990). Technology may solve problems or perform functions without a clear understanding of why it works. Only later on, science may explain the causalities involved. Here, know-how comes before know-why. Know-how is the kind of knowledge with most limited public access and for which mediation is the most complex. The basic problem is the difficulty of separating the competence to act from the person or organization that acts. The outstanding expert – cook, violinist, manager – may write a book explaining how to do things, but what is done by the amateur on the basis of that explanation is, of course, less perfect than what the expert would produce. Attempts to use information technology to develop expert systems show that it is difficult and costly to transform expert skills into information that can be used by others. Know-who refers to a combination of information and social relationships. Telephone books listing professions and databases listing producers of certain goods and services are in the public domain and can, in principle, be accessed by anyone. In the economic sphere, however, it is often necessary to connect with specialized competencies and to find the most skilled and reliable expertise; hence the importance of good personal relationships with key persons one can trust. These social and personal relationships are by definition not public. They cannot be transferred and, more specifically, they cannot be bought or sold on the market. As Arrow (1971) points out, social norms are more efficient than markets because you cannot buy trust and, if you could, it would have very little value. This is fundamental because it implies that the economics of knowledge, in order to be relevant, needs to seek support in other social science disciplines. 13.2.8

Most Knowledge is neither Strictly Public nor Strictly Private

It is clear by now that very little knowledge is ‘perfectly public’. Even information of the know-what type is unavailable to those not connected to telecommunication or social networks. Moreover, the current state of information technology still limits access for those who are in fact connected. Scientific and other types of complex knowledge may be perfectly accessible, in principle, but for effective access the user must have invested in building absorptive capacity. Know-how is never fully transferable since how a person does things reflects that individual’s personality (even organizations have a ‘personality’ in this sense). On the other hand, little economically useful knowledge is completely private in the long run. Tricks of the trade are shared within the profession. Know-how can be taught and learnt in interaction between the master and the apprentice. New technological knowledge may be costly to

296

Handbook of knowledge and economics

imitate but, when it is much more efficient than the old, there are several ways to obtain it. Even when the owner of private knowledge does not want to share it with others, there are ways to obtain it, such as reverse engineering – taking products apart to find out how to produce them. If necessary, private agents will engage in intelligence activities aimed at getting access to competitors’ secrets. Different parts of economic theory handle this mixed situation differently. Underlying much of neoclassical theory of production and economic growth is the simplifying assumption that there is a global bank of blueprints from which anybody can get a copy to be used for starting up production. This ignores the fact that only skilled agents can use blueprints and that skills are unevenly distributed geographically and socially. Since skill cannot easily be transformed into blueprints, the idea of general access is not tenable. The resource-based view of the firm takes a different view and assumes that the competence of the firm determines the directions in which it may expand its activities (Penrose, 1959). It is the specificity of the knowledge base that determines the specific pattern of economic growth of the firm. In a long-term perspective this view leads to a more dynamic perspective – cf. the dynamic capability theory of the firm (Teece and Pisano, 1994). It points to the need for firms to engage in continuous creation of new competencies within the firm and it points to the need to develop ‘learning organizations’. Without such efforts, imitation and innovations among competitors would, sooner or later, erode the firm’s competencies.

13.3 ON TACITNESS AND CODIFICATION OF KNOWLEDGE There is currently a debate among economists about the role of tacit and codified knowledge (Cowan et al., 2000; Johnson et al., 2002). One reason for the interest is that tacit knowledge is definitely not a public good and cannot be transmitted as information. If transformed into explicit codes, it may become easier to ‘transfer’. The process of codification of knowledge is therefore important for understanding the ongoing transformation of the ‘knowledge base’. One of the important consequences of the information technology revolution is that it changes both the incentives and the tools for codification. It makes it more attractive to transform knowledge into information that can be entered into the Internet. At the same time, it also offers new tools to pursue codification and to extend it to more complex bodies of tacit knowledge. The questions to be discussed here are, first, to what degree codification makes knowledge part of a generic knowledge base and,

One knowledge base or many knowledge pools? 297 second, to what degree it makes knowledge more easily transferable across localities and firms. I shall argue that the impact of codification is ambiguous in both these respects. I shall also emphasize that codification of knowledge is not always a contribution to progress. While codification may be a key to advance at some stages of development and in some contexts, at later stages there is a need to give more room for the use of less structured knowledge. The impact of codification on global inequality is ambivalent. 13.3.1

Codification in the Academic Community

The context for codification may determine what direction and form it takes – here it is especially relevant to make the distinction between the ‘academic’ sphere and the ‘business’ sphere. Scientific progress as it is organized by academic institutions is highly dependent on codification and codification is a way to make progress more widely diffused and recognized within the academic community (Dasgupta and David, 1994). You might say that a high degree of codification is a prerequisite for scientific progress.1 Codification in the realm of academic research will typically make access to knowledge more global. Scientific knowledge comes closest to form worldwide common knowledge. But scientific progress goes hand in hand with increasing specialization and complexity. In order to access the most advanced knowledge in any specific scientific discipline, highly developed expertise and advanced infrastructure are needed. Therefore scientific knowledge is public only in a relative sense – there might be no barriers from the supply side but in order to establish effective demand, ‘absorptive capacity’ that requires substantial prior investments is needed. Second, the access to a specific knowledge field among academia does not guarantee that the knowledge can be easily transformed into economic results. For instance, to make economic use of the codified knowledge in physics would be dramatically more realistic in a local context where firms with R&D efforts in ICT sectors operate than in a region where no such firms are present. To have access to academic knowledge without competent industrial users corresponds to being rich in raw materials but lacking the tools and technology to exploit them. So even if academically organized knowledge in principle can be accessed everywhere, it has different economic value in different localities. Actually, an acceleration of the codification process in the academic sphere may increase global inequality since absorptive capacity is so different in the rich and in the poor countries. This lack of absorptive capacity emanates both from a more limited competence in the academic field

298

Handbook of knowledge and economics

and from a more limited competence in the business sector to effectively demand and use academic knowledge in innovation processes. Developing countries will need to invest heavily in academic science in order to be able to draw upon what appears to be ‘a common global stock of knowledge’ and, more importantly, they need to establish absorptive capacity in the private sector. The World Development Report from 1998/99 is incorrect when it starts with the following promising words: ‘Knowledge is like light. Weightless and intangible, it can easily travel the world, enlightening the lives of people everywhere. Yet billions of people still live in the darkness of poverty – unnecessarily.’ Not even the most accessible knowledge at the global level – that is, codified academic science – travels like light and it is much less the case for the codified knowledge produced for profit in the private sector. 13.3.2

Codification in the Business Sector

Firms use knowledge in the form of information as the basis for decision making and in the form of skills for solving problems and designing innovative solutions. Different firms may give different weight to codified versus tacit knowledge and they may make more or less big efforts to transform tacit knowledge into codified knowledge in relation to decision making, work organization and organization of innovation. The balance will reflect the technological and market context. But it may also reflect national/cultural context.2 Within a specific context, firms face important dilemmas when deciding how far to go in terms of codification. Codification offers potential benefits but it is costly and it has negative consequences. In what follows I shall make a distinction between codification in relation to respectively decision making, organization of work and organization of innovation process. I shall focus on how codification affects worldwide access to knowledge. 13.3.3

On the Use of Expert Systems within Firms

Efforts to develop expert systems and to codify specific technical processes may contribute to economic efficiency. In the process of codification the creation of ‘know-why’ knowledge may take place and transform trialand-error practices into a systematic understanding of the processes at stake (Lazaric et al., 2003). But in important fields of management practice there are limits to how far human know-how can be substituted for by expert systems. With the wide use of information technology one might expect firms to substitute for scarce management skills by management information

One knowledge base or many knowledge pools? 299 systems. The basic idea would be to develop expert systems that could be fed by relevant information about the business context and then left to a computer program to come up with the right decision. There is no doubt that there is some movement in this direction, but it is a movement characterized by hesitation and set-backs. Eliasson (1996) has illustrated the limits of using management information systems as a substitute for management skills by pointing to the strategic failures of leading producers of such systems as IBM and other big ICT firms. Know-how, if it is not economically trivial, easy to copy and routine, is never a completely public good and normally firms get access to it only by hiring experts or merging with companies with the knowledge they want. It has also been demonstrated that the transformation always involves changes in the content of the expert knowledge (Hatchuel and Weil, 1995).3 The reason for this is that strategic management decisions are based upon experience and that they make as much use of pattern recognition and intuition as they use analytical models. Decisions are normally not the outcome of a logical deductive process. What distinguishes the successful manager or team of managers from the less successful is that he/they can handle new and unforeseen problems as well as routine ones. This indicates one of the most important limits to the use of information management systems. They may be used successfully only for repetitive decisions and in an environment that remains stable over time. It has been argued that the more complex the problem, the more dangerous it is to rely on management information systems to find the right solution (Cowan, 2001). Complexity in itself might not be the major hindrance to codification, however. Other factors that refer to the social interaction among agents may be at least as important. From time to time new waves of codification in a field of knowledge may be triggered by new insights into technical causalities (Lazaric et al., 2003). The most important limit to what management is willing to invest in codification is related to the rate of change of the field of knowledge and of the problems that it has to tackle. If new types of problems appear frequently, codification may actually invalidate the capacity to deal with the problems. If problems remain structured in a similar way, the incentive to codify is strong. Therefore the impact of information technology on incentives for codification of expert knowledge has been contradictory. Information technology has made it more realistic to simulate and reproduce complex decision processes. But at the same time the extended use of information technology has led to an acceleration of change, making it less attractive for firms to get locked into pre-programmed routines. To this should be

300

Handbook of knowledge and economics

added that experts at all levels of organization will resist attempts to codify their knowledge and that of all categories managers are in the best position to make this resistance effective. This implies that management skills cannot be easily transferred from one context to another. One most effective way to develop and transfer good management principles may be to move managers from one organization to another. Data from Denmark indicate that small family-owned firms that are stuck with the same top management for a long period are less successful in terms of innovation and growth; the most successful firms are parts of industrial groups and multinational firms, and this reflects the fact that there is internal ‘job circulation’ among managers within such organizations (Lundvall and Nielsen, 2007). For less developed countries this means that developing local management skills and selectively learning from management abroad is necessary. There are certainly information systems supporting accounting procedures, customer information and procurement and stock control. As far as there is a scarcity of management skills, it might be rational to invest in such systems. However, these systems can be seen as supporting but not as substitutes for strategic decision making. And their intelligent use requires strategic leadership. 13.3.4

Codifying the Work Process

Work may be organized so that it is more or less based upon codification and built-in routines. The history of work organization involves stages of codifying skills and building them into machinery or into routines to be strictly followed by workers. Scientific management and Taylorist forms of organization have at times been highly productive and effective. Today there is a tendency to define ‘the high-performance workplace’ as one where workers have more autonomy – the work process is less structured and codified (Becker and Huselid, 1998). Lorenz and Valeyre (2004) have used employee survey data to develop a taxonomy of work processes in terms of the learning opportunities they offer to employees. They distinguish between four categories: ● ● ● ●

Simple production Taylorism Lean production learning Discretionary learning.

In their work the focus is on the content of learning of these four forms of work organization. But the four models may also be contrasted in terms

One knowledge base or many knowledge pools? 301 Table 13.1

National differences in organizational models (percentage of employees by organizational class) Discretionary learning model

Lean production

Taylorist organization

Simple organization

Denmark Netherlands

60.0 64.0

21.9 17.2

6.8 5.3

11.3 13.5

Germany France

44.3 38.0

19.6 33.3

14.3 11.1

21.9 17.7

Portugal Greece

26.1 18.7

28.1 25.6

23.0 28.0

22.8 27.7

EU-15

39.1

28.2

13.6

19.1

Source: European Foundation for the Improvement of Living and Working Conditions, Third Working Conditions Survey, 2000.

of to what degree they codify the skills of workers and build them into machinery or into strict routines. Simple production is a mixed category of old and new services but it is the least structured, while the Taylorist organization is the most highly codified. The lean production learning model makes use of modern management techniques such as job rotation and team work, but leaves little autonomy to the individual worker. This contrasts with the discretionary learning model, where the employees are given more freedom to make choices. I would argue that lean production is less structured than Taylorism but more structured than discretionary learning. On the basis of the data developed by Lorenz and Valeyre we can get an idea of how the workforce is distributed between the four forms in 15 European countries. It is interesting to note that the workforce of countries at different income levels is distributed differently over these four categories. To illustrate, I have ordered six European countries according to GNP per capita and shown how the workforce is distributed across the four archetypes of working organization (Table 13.1). The most important result is the complex relationship between level of economic development and degree of codification of the work process. At low-income levels the work process will become more codified until a certain point, and after that it will become less codified. Economic development that involves a growth of manufacturing activities in a context of skill shortages will increase the proportion of workers in Taylorist organizations (cf. Portugal and Greece). At a later stage of

302

Handbook of knowledge and economics

development more demanding in terms of flexibility and innovation, it becomes more rational to give workers more autonomy and the proportion of Taylorist work is reduced (Germany and France). In (small) economies with high income per capita, the need for continuous innovation and adaptation to change is even greater and here we find a movement toward discretionary learning (Netherlands and Denmark). Table 13.1 shows that there is not a linear relationship between degree of codification of work and the level of economic development. This pattern reflects that codification of workers’ skills has both advantages and drawbacks seen from the point of view of management. Codifying the skills of employees may make the firm less dependent on employees. If it is possible to build the skills into machinery and routines, the firm may succeed in establishing a situation where parts of the labour force may be substituted without negative effects on performance. In a context where there is a dramatic shortage of skilled labour, a Taylorist organization with much of the workers’ competence built into the machinery might be seen as attractive. The negative side of this strategy is that the labour force may remain unskilled and that upgrading of products and processes will be slow. Also the firm will be vulnerable to external shocks. When the context – technology or market – changes, the organization will not be able to adapt since it is designed to solve a constant and narrow set of problems. Only in a context of stable technology and stable demand will this kind of organization be attractive. Today this is typical of sectors with low value-added per employee. This is why developing countries with a shortage of skilled labour risk getting stuck in such activities. At the aggregate level of the national production system, a strong presence of such rigid organizations implies a high birth and death rate of organizations. The alternative to the adaptation of existing organizations is that they die when they cannot cope and that new ones appear. High frequencies of birth and death of firms are sometimes interpreted as signs of a vibrant entrepreneurial economy, but they also reflect the degree of rigidity of existing organizations. To close down an existing organization and establish a new one involves substantial ‘transformation costs’ in terms of lost community knowledge. This is why less developed economies specialized in the most Taylorist steps in global value chains may be victims of both lock-in into low-value-added activities and high vulnerability to external shocks. As explained in Box 13.1, in this case codification cannot be seen as a process that represents progress from a lower to a higher level of knowledge. Work process codification has made knowledge more accessible worldwide. Today it is possible to establish industrial processes in parts of the

One knowledge base or many knowledge pools? 303

BOX 13.1

CODIFICATION AS REFINING KNOWLEDGE AND TRANSFORMING IT TO A HIGHER FORM

There is a bias in Western philosophy favouring analytical and highly structured knowledge while regarding intuitive knowledge rooted in experience as being of a lower order (Nonaka and Takeuchi, 1995). Our analysis of codification of academic knowledge, expert knowledge, work processes and innovation does not support this view. The closer we get to the frontier of science, the more the scientist relies on experience and on his/her capacity to recognize patterns without being able to present his/her analysis as a logical sequence. The more complex the management task and the more dynamic the context, the less can the firm rely on expert systems. Highly codified work organizations such as those dominated by Taylorism characterize less developed economies where skills are scarce. In high-income countries, more and more employees are given free rein to engage in unstructured problem solving, individually and in teams. Innovation modes based upon science and giving major attention to codified knowledge cannot stand alone. They need to be supported by organizational forms that promote experience-based learning and outcomes, not in disembodied codified knowledge, but in new skills and new products. world where skilled labour is scarce. This is reflected in the tendency toward more and more developed global value chains and more generally in increasingly global competition, especially in standardized commodities. For developing countries it is a key problem to find a way of building change into such Taylorist organizations. One option is to develop elements of ‘lean production learning’. Lean production implies that workers may work in teams, change tasks and get some limited discretion in solving problems as they appear. This might give the minimum of space for learning that in the longer run can increase the value-added in such processes. But in large emerging economies such as China where the ambition is to develop home-spun innovations in highly dynamic technological fields, there might be a need to move directly toward discretionary learning in certain parts of the economy exposed to rapid change in technology and markets.

304

Handbook of knowledge and economics

13.3.5

Codification in Relation to the Innovation Process

Most authors using the concept of knowledge creation and knowledge production refer to technical innovation as the output of the process (Antonelli, 1992; Nonaka and Takeuchi, 1995). In new growth theory, the output of the R&D sector is viewed either as a blueprint for a new production process more efficient than the previous one or as a new semimanufactured goods process not easily copied by competitors (Romer, 1990; Verspagen, 1992, pp. 29–30). The process of innovation may be more or less codified. One example of a highly codified innovation process would be the development of a new pharmaceutical product where a new chemical formula is the basis of the innovation. One example of a less codified innovation would be the development of a new machine where the operator changes the machine on the basis of his own experience. Box 13.2 refers to Adam Smith, who made a similar distinction in the introduction to the classic, The Wealth of Nations. In Jensen et al. (2007) we distinguish between two different modes of innovation and show that firms successful in terms of innovation combine strong versions of both modes. The STI mode refers to the science– technology–innovation sequence and the process operates mainly on the basis of codified knowledge, while the DUI mode operates mainly on the basis of experience based on learning by doing, using and interaction. Here I shall consider how the two modes affect the access to the outcome of the innovation process. Is a codified outcome easier to transfer across organizational boundaries and geographical borders than the outcome of a DUI process? 13.3.6

The STI Mode

The dichotomy should not be taken to imply an absence of complementarities between the two modes. For instance, scientists operating at the frontier of their field in the R&D departments of large firms need to draw upon their tacit experience-based knowledge when making experiments and interpreting results, and specific R&D projects will often be triggered by problems emanating from practice. We may still define it as predominantly STI if immediate attempts are made to restate the problem in codified form. The R&D department may start going through its earlier work, looking for pieces of codified knowledge, as well as looking for codified knowledge that can be drawn from outside sources. In order to communicate with scientists and scientific institutions outside, it may be necessary to translate the problem into a formal scientific code.

One knowledge base or many knowledge pools? 305

BOX 13.2

ADAM SMITH AND THE TWO MODES OF INNOVATION – DUI AND STI

Adam Smith links the development of the division of labour to innovation in two different ways and in doing so he indicates two different modes of innovation. One is experience-based, and corresponds to DUI learning, while the other is science-based and corresponds to STI learning. Adam Smith (1776, p. 8) on the DUI mode of innovation: A great part of the machines made use of in those manufactures in which labour is most subdivided, were originally the inventions of common workmen, who, being each of them employed in some very simple operation, naturally turned their thoughts towards finding out easier and readier methods of performing it. Whoever has been much accustomed to visit such manufactures, must frequently have been shown very pretty machines, which were the inventions of such workmen, in order to facilitate and quicken their own particular part of the work.

Adam Smith (ibid., p. 9) on the STI mode of innovation: All the improvements in machinery, however, have by no means been the inventions of those who had occasion to use the machines. Many improvements have been made by the ingenuity of the makers of the machines, when to make them became the business of a peculiar trade; and some by that of those who are called philosophers or men of speculation, whose trade it is not to do any thing, but to observe every thing; and who, upon that account, are often capable of combining together the powers of the most distant and dissimilar objects.

All through the process, documenting results in a codified form remains important. It is not sufficient for the single scientist to keep results in his own memory as tacit knowledge. Often the project involves teamwork and modularization, where single results are used as building blocks for other members in the team. At the end of the process – if it is successful – a transfer of the results within the organization or across organizational borders will call for codified documentation as well. When a patent application is made the documentation needs to be made in a techno-scientific language that allows the patenting authority to judge the originality of the innovation. This means that, on balance, the STI mode of learning, even if it starts from a local problem, will make use of ‘global’ know-why knowledge

306

Handbook of knowledge and economics

all the way through and, ideally, it will end up with ‘potentially global knowledge’ – that is, knowledge that could be used widely if it were not protected by intellectual property rights. 13.3.7

The DUI Mode

The DUI mode of innovation refers to an innovation process where there is emphasis on the utilization of know-how and know-who that is tacit and often highly localized. This mode may result in incremental innovation in simple organizations but it is also present when it comes to realize radical innovation. Here it requires organizational structures and relationships that enhance and utilize learning by doing, using and interacting. In terms of the work organization patterns it implies a combination of discretionary and lean production learning. The DUI mode of learning is characterized by ongoing change that continuously confronts employees with new problems and incites learning by doing (Arrow, 1962a). Finding solutions to these problems enhances the skills of the employees and extends their repertoires. Some of the problems are specific while others are generic. Therefore learning may result in both specific and general competencies for the operator. When the process is complex – a good example is the learning-by-using of new models of aircraft – it will involve interaction within and between teams and may result in shared routines for the organization. As the whole organization gets more insight into the actual working of the system, it might find more efficient ways to organize work and solve problems as they emerge. This is the kind of case that Rosenberg (1982) uses to illustrate learning-by-using. Both learning-by-doing and learning-by-using normally involve interaction between people and departments, and this is why such practices as cross-functional groups and job rotation show positive relations between learning and performance. It has been argued that learning-by-doing and learning-by-using result only in ‘local’ knowledge and that without codification and transformation of the knowledge into codified knowledge the impact on the economy as a whole would remain limited. In a recent paper I have argued that this argument neglects the outcome of learning-by-interacting involving users and producers. The introduction of new products emanating from this kind of interaction is an alternative way of transforming local learning into more global knowledge. The new products will embody the experiences of several users. From the viewpoint of the whole economy, learning-by-interacting has the effect of transforming local learning into general knowledge embodied in for instance new machinery, new components, new software-systems or even new business solutions (Lundvall, 2006).

One knowledge base or many knowledge pools? 307

13.4 HOW DO THE STI AND THE DUI MODES OF INNOVATION AFFECT THE TRANSFERABILITY OF KNOWLEDGE? While the output of the DUI mode may be a tangible new product with embodied technical knowledge – such as a numerically controlled machine tool – the outcome of the STI process may be disembodied knowledge that can be widely distributed. But the more codified form also makes it easier to protect this kind of knowledge through intellectual property rights in the form of patents or licences. The codification process that results in a patent may be seen as contributing to the cumulative knowledge creation process by making explicit fundamental characteristics of the new product or process. This contrasts with the outcome of the DUI mode, where the new knowledge is embedded in the new product or process but not made explicit and not transformed into disembodied knowledge that can be traded in the market. If the outcome is a new product, reverse engineering may be an option for competitors. If it is a new process for internal use, the access will be limited. Here the mobility of employees may be an alternative transfer mechanism. Normally one would expect a codified output of the innovation process to be more accessible worldwide than embodied knowledge. In the current period, where protection of codified knowledge has become a major concern of firms that are world leaders in advanced technology, this might not be the case. The STI mode resulting in disembodied codified knowledge may actually result in more restrictive access than the DUI mode, where the final product is a new system or product with embodied but unprotected knowledge. If this is so, it might give the most advanced firms in different technological fields an incentive to go even further in the direction of codification but with an increasingly negative impact on the distribution of knowledge. Another important contradiction in modern capitalism is reflected in the codification process. While codification in principle makes knowledge disembodied and more accessible worldwide, it also makes it easier to exclude others from using it. This tension is reflected in knowledge politics and knowledge management. With the most recent developments in the field of intellectual property rights, where these tend to be stretched to cover new areas including living organisms as well as software, the net effect of codification on global access might actually be negative. The creativity of people and communities finds itself blocked by too many barriers.

308

Handbook of knowledge and economics

13.5 CONCLUSIONS The debate on codification has been complicated by the fact that different kinds of codes have been alluded to. Some codes are explicit and available in the form of textbooks, manuals, formulas and organizational diagrams. Others have developed spontaneously as a local means of communication within or between organizations (Arrow, 1974). Communities of practice may have their own codes that give outsiders privileged access to community knowledge. Epistemological communities that bring together scholars contributing to a specific scientific field certainly have their own code. In fact a great deal of economically relevant knowledge is communicated in such specialized and local codes. One of the most important sources of innovation is establishing social interaction and communication between such communities. We need more of the idealized agents referred to by Adam Smith (1776/1910, pp. 9–10): ‘philosophers or men of speculation, whose trade it is not to do any thing, but to observe every thing; and who, upon that account, are often capable of combining together the powers of the most distant and dissimilar objects’. The scientific as well as business community is highly specialized and there may be a scarcity of ‘men of speculation’ with the kind of background that makes it possible for them to combine distant and dissimilar objects. To foster such men and women may be a major challenge for education systems and for designing career paths in the labour market. Another challenge is to design legal systems and institutions so that they make it attractive to combine elements from separate knowledge pools in new ways. Another characteristic of intellectual capitalism is the growing urge to privatize what can be privatized and to transform shared knowledge into private and legally protected property. The ambivalence in the business community between the need to share knowledge with others and the need to protect their own knowledge from others has become biased in favour of privatization. To some degree this may be seen as a response to the tendency to make explicit and codify knowledge. Whatever the reason, it makes access to knowledge pools even more dependent on financial resources and political power. This chapter started from the idea of ‘the knowledge-based economy’. Our conclusion is that the idea of ‘one knowledge base’ for the economy is misleading and that the kind of knowledge that matters for the economy should rather be regarded as many separate ‘pools’, each with limited access. Using the standard terminology of economics, most knowledge is neither a strictly private nor a strictly public good. Most useful knowledge is a kind of ‘community resource’ – neither private nor public. In a capitalist regime there will be constant efforts to transform this kind of

One knowledge base or many knowledge pools? 309 knowledge into private property. But often it can neither be appropriated by individual firms nor be transformed into a commodity. In the rich part of the world, knowledge management and knowledge politics might therefore be seen as similar to the management of an ecosystem. Besides focusing on the growth and quality of each pool, it is necessary to stimulate the valorizing of diversity through innovation. Establishing new links between separate pools, for instance by letting experts with access to one pool get access to another pool, or stimulating experts with access to different pools to interact, are fundamental elements in an innovation-oriented knowledge policy. In less developed countries innovation policy is about tapping into foreign pools and linking them to domestic pools and transforming the combination either into innovation and market value or into social and collective use. To refer to this process as knowledge ‘transfer’ or knowledge ‘spillover’ is to underestimate the efforts necessary and the barriers that may exist. Often it is necessary for the less developed economy to go through a process of institutional transition in order to overcome the barriers. And some of the knowledge pools in the rich world are surrounded by high fences and guarded by company lawyers armed with law-books spelling out intellectual property rights. Others may be difficult to localize since they are integrated in networks of more or less invisible academies and communities of practice. There is a need for a global new new deal where the focus is upon giving less privileged parts of the world easier access to the pools of knowledge now controlled by the rich countries and by transnational companies. This implies both a reform of the intellectual property rights regime of the World Trade Organization and a major investment in competence building in the less developed economies.4 Without such an effort the gaps between rich and poor countries will grow, perhaps with the exception of a few large economies, such as China, where the current efforts to accumulate capital and invest in endogenous innovation is enormous (Gu and Lundvall, 2006).

NOTES 1. But not sufficient: it would be a serious mistake to assume that scientific work is based exclusively upon codified knowledge. The personal knowledge of the scientist cannot be fully codified and personal knowledge is distributed unevenly in space, reflecting local learning experiences. And therefore important elements of scientific knowledge are localized and embodied in scientists and scientific teams. These elements can be transferred from one place to another only through the movement of people. This is why star scientists can earn a lot of money.

310

Handbook of knowledge and economics

2. Arundel et al. (2007) demonstrate that the differences in these respects are also substantial within the group of the most developed European economies. 3. The difficulty/impossibility of making a one-to-one translation of expert knowledge into information systems does not mean that the effort to make such a translation is meaningless. In fact the effort may be helpful in imposing a more systematic approach rather than trusting in ‘trial and error’ (Lazaric et al., 2003). This is a point of more general relevance. In a similar vein, it might be very useful for management to make an annual report on the knowledge assets of the firm in spite of the fact that the outcome for the bottom line may be dubious and that the routinization of the reporting may be of little use (Edvinsson, 1997). 4. For an example of an effort to redistribute capacity building in the field of innovation research, see www.globelics.org.

REFERENCES Abramowitz, M. and David, P. (1996), ‘Technological change and the rise of intangible investments: the US economy’s growth path in the twentieth century’, in D. Foray and B.-Å. Lundvall (eds), Employment and Growth in the Knowledge-based Economy, Paris: OECD, pp. 11–32. Allen, Robert C. (1983), ‘Collective invention’, Journal of Economic Behaviour and Organisation, 4 (1), 1–24. Amin, A. and Cohendet, P. (2004), The Architectures of Knowledge: Firms, Capabilities and Communities, Oxford: Oxford University Press. Antonelli, C. (1992), The Economics of Information Networks, Amsterdam: Elsevier. Arrow, K.J. (1962a), ‘The economic implications of learning by doing’, Review of Economic Studies, XXIX (80), 155–73. Arrow, K.J. (1962b), ‘Economic welfare and the allocation of resources for invention’, in R.R. Nelson (ed.), The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton, NJ: Princeton University Press, pp. 609–25. Arrow, K.J. (1971), ‘Political and economic evaluation of social effects and externalities’, in M. Intrilligator (ed.), Frontiers of Quantitative Economics, Amsterdam: North-Holland, pp. 3–24. Arrow, K.J. (1974), The Limits of Organization, New York: W.W. Norton and Co. Arrow, K.J. (1994), ‘Methodological individualism and social knowledge’, Richard T. Ely Lecture, AEA Papers and proceedings, 84 (2), 1–9. Arundel, A., Lorenz, N., Lundvall, B.-Å. and Valeyre, A. (2007), ‘How Europe’s economies learn: a comparison of work organization and innovation mode for the EU-15’, Industrial and Corporate Change, 16, 1175–210. Becker, B. and Huselid, B. (1998), ‘High-performance work systems and firm performance: a synthesis of research and managerial implications’, in G. Ferris (ed.), Research in Personnel and Human Resources, 16, Greenwich, CT: JAI Press, pp. 53–102. Campbell, J. (1995), Understanding John Dewey: Nature and Cooperative Intelligence, Chicago, IL: Open Court Publishing. Chesborough, H.W. (2003), Open Innovation: The New Imperative for Creating and Profiting from Technology, Cambridge, MA: Harvard University Press. Cowan, D. (2001), ‘Expert systems: aspects of limitations to the codifiability of knowledge’, Research Policy, 23, 1355–72. Cowan, R. and Jonard, N. (2003), ‘The dynamics of collective invention’, Journal of Economic Behaviour & Organisation, 52, 513–32. Cowan, M., David, P. and Foray, D. (2001), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 9, 211–54. Dasgupta, P. and David, P.A. (1994), ‘Towards a new economy of science’, Research Policy, 23, 487–521.

One knowledge base or many knowledge pools? 311 David, P.A. and Foray, D. (2002), ‘Economic fundamentals of the knowledge society’, SIEPR Discussion Paper No. 01-14, Stanford University. Edvinsson, L. (1997), ‘Developing intellectual capital at Skandia’, Long Range Planning, 30, 366–73. Eliasson, G. (1996), Firm Objectives, Controls and Organization, Amsterdam: Kluwer Academic Publishers. Foray, D. (2000), The Economics of Knowledge, Cambridge, MA: MIT Press. Foray, D. and Lundvall, B.-Å. (1996), ‘From the economics of knowledge to the learning economy’, in D. Foray and B.-Å. Lundvall (eds), Employment and Growth in the Knowledge-Based Economy, Paris: OECD, pp. 11–32. Gu, S. and Lundvall, B.-Å. (2006), ‘China’s innovation system and the move towards harmonious growth and endogenous innovation’, Innovation: Management, Policy & Practice, 8, 1–26. Hatchuel, A. and Weil, B. (1995), Experts in Organisations, Berlin: Walter de Gruyter. Hayek, F.A. von (1937), ‘Economics of knowledge’, Economica, 4, 33–54. Jensen, M., Johnson, B. and Lundvall, B.-Å. (2007), ‘Forms of knowledge, modes of innovation and innovation systems’, Research Policy, 36, 680–93. Johnson, B., Lorenz, E. and Lundvall, B.-Å. (2002), ‘Why all this fuss about codified and tacit knowledge?’, Industrial and Corporate Change, 11, 245–62. Keynes, J.M. (1936), The General Theory of Employment, Interest, and Money, New York: Macmillan. Kolb, D.A. (1984), Experiential Learning, Englewood Cliffs, NJ: Prentice-Hall. Lazaric, N., Mangolte, P.A. and Massué, M.L. (2003), ‘Articulation and codification of collective know-how in the steel industry: evidence from blast furnace control in France’, Research Policy, 32, 1829–47. Lorenz, E. and Valeyre, A. (2004), ‘Organisational innovation in Europe: national models or the diffusion of a new “one best way”?’, paper presented at the 2004 DRUID Summer Conference, Copenhagen, 14–16 June. Lundvall, B.-Å. (ed.) (1992), National Innovation Systems: Towards a Theory of Innovation and Interactive Learning, London: Pinter Publishers. Lundvall, B.-Å. (2006), ‘Interactive learning, social capital and economic performance’, in D. Foray and B. Kahin (eds), Advancing Knowledge and the Knowledge Economy, Boston, MA: MIT Press, pp. 63–74. Lundvall, B.-Å. and Johnson, B. (1994), ‘The learning economy’, Journal of Industry Studies, 1 (2), 23–42. Lundvall, B.-Å. and Nielsen, P. (2007), ‘Knowledge management in the learning economy’, International Journal of Manpower, 28 (3–4), 207–23. Marshall, A.P. (1923), Industry and Trade, London: Macmillan. Mead, G.H. (1934), Mind, Self, and Society, Chicago, IL: University of Chicago Press. Nelson, R.R. (1959), ‘The simple economics of basic economic research’, Journal of Political Economy, 67, 323–48. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Nonaka, I. and Takeuchi, H. (1995), The Knowledge Creating Company, Oxford: Oxford University Press. OECD (2000), Knowledge Management in the Learning Society, Paris: OECD. Pavitt, K. (1998), ‘Technologies, products and organisation in the innovating firm: what Adam Smith tells us and Joseph Schumpeter doesn’t’, paper presented at the DRUID 1998 Summer conference, Bornholm, 9–11 June. Penrose, E. (1959/1995), The Theory of the Growth of the Firm, Oxford: Oxford University Press. Polanyi, M. (1958/1978), Personal Knowledge, London: Routledge and Kegan Paul. Romer, P.M. (1990), ‘Endogenous technological change’, Journal of Political Economy, 98, 71–102. Rosenberg, N. (1982), Inside the Black Box: Technology and Economics, Cambridge: Cambridge University Press.

312

Handbook of knowledge and economics

Shapiro, C. and Varian, H.R. (1999), Information Rules: A Strategic Guide to the Network Economy, Boston, MA: Harvard Business School Press. Simon, H.A. (1957), Administrative Behavior, Glencoe, IL: Free Press. Smith, A. (1776/1904), An Inquiry into the Nature and Causes of the Wealth of Nations, London: Methuen and Co., Ltd, ed. Edwin Cannan, 1904, 5th edn. Smith, A. (1776/1910), The Wealth of Nations, Vol. I, London: J.M. Dent & Sons Ltd. Teece, D. and Pisano, G. (1994), ‘The dynamic capabilities of firms: an introduction’, Industrial and Corporate Change, 3 (3), 537–56. Verspagen, B. (1992), Uneven Growth between Interdependent Economies, Maastricht: Faculty of Economics and Business Administration. Vincenti, W.G. (1990), What Engineers Know and How they Know it: Analytical Studies from the Aeronautical Industry, Baltimore, MD: Johns Hopkins University Press. Von Hippel, E. (2005), Democratizing Innovation, Boston, MA: MIT Press. Wenger, E. (1998), Communities of Practice: Learning, Meaning and Identity, Cambridge: Cambridge University Press. Winter, S. (1987), ‘Knowledge and competence as strategic assets’, in D. Teece (ed.), The Competitive Challenge: Strategy for Industrial Innovation and Renewal, Cambridge, MA: Ballinger, pp. 159–84. World Bank (1998/99), World Development Report: Knowledge for Development, Washington, DC: World Bank. Ziman, J. (1979), Reliable Knowledge, Cambridge: Cambridge University Press.

14 Knowledge in finance: objective value versus convention André Orléan

14.1 INTRODUCTION The idea that economics deals with objective values plays an absolutely fundamental role in the way economists consider the singularity of their discipline. In their eyes, economics can be distinguished from the other social sciences precisely because it deals with objective constraints, constraints of scarcity, which are imposed on all agents, whatever their beliefs, whereas disciplines like history or sociology study facts of opinion, the meaning of which varies from one group or society to another. It would not be an exaggeration to say that economists construct the core identity of their discipline against competing disciplines through the play of this paradigmatic opposition between the flexibility of opinions and the objectivity of values. In a 1908 conference at the French Society of Political Economy, on the question of ‘the position of political economy in the social sciences’, Émile Durkheim (1975/1908, pp. 219–20) pointed out all the limits of this opposition. In reply to the question raised, he started by stating: What makes this question difficult is that, at first sight, political economy appears to deal with facts of a very different nature from the other social sciences. Morality and law, which form the subject matter of the other specified social sciences, are essentially matters of opinion. Wealth, which is the subject of political economy, seems on the contrary to be essentially objective and independent of opinion. (My translation)

But he immediately followed this by affirming that, in his view, this distinction is not valid: However, the present speaker believes that economic facts can be approached from another viewpoint; they are also, to a degree that I will not attempt to define, a matter of opinion. For the value of things depends, not only on their objective properties, but also on the opinions held about them. Doubtless, these opinions are partly determined by the objective properties; but they are also shaped by many other influences. If religious opinion should forbid a certain drink – wine, for example – or a certain meat (pork), then wine or pork would lose some or all of their exchange value. Likewise, it is movements in opinion,

313

314

Handbook of knowledge and economics

in taste, which give their value to one fabric or precious stone rather than another . . . (Ibid.)

This led him to conclude: From this point of view, the relations between economic science and the other social sciences appear in a different light. They all deal with phenomena which, considered at least from certain angles, are homogeneous, because they are all, in certain respects, matters of opinion . . . Political economy thus loses the predominance it has invested itself with, to become a social science alongside the others, in a close relation of solidarity with them, with no valid claim to rule over them. (Ibid.)

In other words, Durkheim was warning economists that the hypothesis of objectivity should be treated with circumspection. He pointed out that the determination of many economic values is highly dependent on the social context. This is, in his eyes, an empirical fact that cannot be denied. He also observed that there was nothing dramatic about accepting this fact, but, on the contrary, it should help to bring economics and the other social sciences closer together, by establishing a ‘close relation of solidarity’ between these disciplines. According to Durkheim, the key to the rapprochement he was calling for lay in the acceptance by economists of the role played, even in economics, by phenomena of opinion. To his great surprise, this analysis was met with widespread and unequivocal rejection by the economists at the conference. M. Edmond Villey even said that he ‘felt rather scandalized’. He declared: ‘Opinion . . . does not determine value, which is determined by rigorous natural laws . . . it is always the law of supply and demand, completely independent of opinion, which determine prices as it determines all values’ (ibid., p. 223). M. Paul Leroy-Beaulieu went further: ‘political economy stands above the other social sciences: it is the only one to have an indestructible and positive foundation, and its laws are immutable, however opinions may vary’ (ibid., p. 225). This is a good illustration of the forceful resistance displayed by economists when the hypothesis of objectivity is brought into question or when the idea of opinion is introduced. Faced with the violence of this reaction, which he obviously had not expected, Durkheim pointed out that the concept of opinion should not be understood as a pejorative term, a synonym for prejudice or unconsidered feelings: ‘This would mean only seeing one aspect of opinion’, he said. ‘We must not forget that opinion is also the result of experiments that peoples have made over the course of centuries; and this fact does lend it some authority’ (ibid., p. 223). But this refinement of his argument had no effect.

Knowledge in finance

315

As far as we can generalize about modern economists, they hardly appear to have evolved on this matter. Their spontaneous philosophy remains unchanged, and the hypothesis of objectivity still has pride of place. This philosophy can be described as ‘objectivist’ or ‘naturalizing’ in that it considers economic and financial equilibrium as a successful adaptation to predefined natural constraints of scarcity. The evolution of prices is determined by the fundamentals (resources, technologies, consumer preferences, market structures). The opinions and beliefs of agents play little or no role in such an analytic framework. The concept of opinion is rejected because economic theory generally conceives of the economic world as an objective world, free of ambiguity, which can be known in the same way that physicists know the natural world. The idiosyncratic beliefs of different individuals are of little importance because, in such an analytic framework, the objectivity of the facts cannot fail to impose itself on rational, informed individuals. This criticism of modern economists may seem too unequivocal if one considers all the work carried out on ‘selffulfilling prophecies’ or ‘sunspot equilibria’. The particularity of this work resides precisely in the fact that it considers situations in which economic equilibria are dependent on the collective opinions of economic agents. Here, assuredly, lie avenues of research that could transform economic theory. In the current state of research, however, these avenues remain largely unexplored, and I believe that this is precisely because the majority of economists continue to share an objectivist view of the world, relegating these phenomena of collective belief to the outskirts of economic theory, more as curiosities than as hypotheses deserving serious exploration and deepening. The present text seeks to define the terms of our analysis more precisely, on the basis of a specific investigation of finance theory. I shall move away from propositions about the knowledge of economic agents in general to concentrate on the way in which this knowledge is envisioned in the realm of finance. In the first section, we shall return to the hypothesis of objectivity to bring to light the specific form it assumes in the sphere of financial relations. In this case, it is the question of the relation to the future that is central. I shall demonstrate that the hypotheses made about the way individual investors anticipate future yields from securities are highly conditioned by the hypotheses made about the very nature of the future. Because this future is conceived as existing objectively in a probabilistic form, it follows that an ‘accurate estimate’ of the value of the securities can be defined ex ante and without ambiguity. The concept of informational efficiency, which plays such an important role in finance theory, can be directly deduced from this. The market for security A is said to be informationally efficient if the price formed at time t is equal to the fundamental valuation

316

Handbook of knowledge and economics

of the security. In the second section, I shall show that the hypothesis of the objectivity of the future raises several problems, even in a probabilistic framework. This will lead to challenging the idea that it is possible to define, at time t, such a thing as a ‘true estimate’ of the value of securities. On the contrary, I contend that only subjective estimates can be made of this value, and that these estimates are inherently diverse and heterogeneous. This is a central point in my analysis: knowledge of the future cannot be objective; it is irreducibly subjective. To demonstrate this, I present a thought experiment in which I set two investors face to face, each defending a different estimate of the same security, and I shall show that these two agents, even if they are rational and perfectly informed, can perfectly well maintain their divergence, without there being any rational argument or objective information to sway them. I shall use the term ‘opinion’ to qualify these subjective, informed, rational beliefs. This idea that, in a financial market, diverse ‘opinions’ about the value of the same security can rationally coexist leads us to uphold that it is impossible to define ex ante such a thing as a unique ‘true estimate’ or fundamental valuation. This impossibility poses a radical challenge to the idea of informational efficiency, to the extent that it is no longer possible to determine ex ante an estimate to serve as a yardstick by which to measure the capacity of the market accurately to evaluate securities. Consequently, the concept of bubble, itself defined as a persistent gap between the fundamental value of the security and the price observed, also loses its meaning. Finally, in the third section, I shall seek to demonstrate that taking the concept of opinion into account does not strip finance theory of all meaning. I shall focus my attention more particularly on the market price itself. What is the nature of this price? What collective knowledge does it express? How does one move from a set of heterogeneous opinions to a unique market price? The most direct answer consists in interpreting this price as the sum total of individual subjective estimates, weighted by the capacity of each investor to influence prices. This has the advantage of simplicity, but it treats the market price as a pure ‘artefact’ emerging mechanically from competitive interactions, with which it is difficult to associate a specific vision of the future. Yet empirical observation of markets and economic theory underlines the central role that price plays as an expression and vector of a certain conception of the future, a conception to which all investors can refer, either to accept it or to reject it. These observations lead to discarding this first interpretation and to proposing an alternative analysis, in which the financial market is seen as a cognitive machine whose function is to produce a reference opinion, perceived by all the operators, not as the improbable product of a sum total of more or less well-informed opinions, but as an expression of ‘what the market thinks’. This is because of the self-referential nature of speculation, where each

Knowledge in finance

317

individual makes up his/her mind according to what he/she anticipates the majority opinion to be. Here we have an example of the Keynesian ‘beauty contest’ model. Basing my argument on the works of Mehta et al. (1994), I shall show that market price has the nature of a salient opinion that imposes itself on agents. It follows that the price can be considered as a convention.1 We are now far removed from the objectivist theories we started with. We have moved from a world in which a true, objectively given representation of the future is considered to exist to a world in which the representation of the future is not a fact of nature, but the result of a self-referential process of shared beliefs. This representation is neither natural nor objective, but historically and socially constructed. However, this does not mean that the standard analysis has been entirely rejected. Our paradigm retains the idea of competitive price formation and, consequently, the assumption of no arbitrage. What distinguishes the conventionalist approach is the way in which the collective knowledge produced by the market is apprehended. From this point of view, the standard analysis appears rather as a particular case within the conventionalist analysis, a particular case in which the convention is adhered to with such force and unanimity that each person is convinced of its absolute veracity, to the point of seeing it as an exact representation of the future. From the observer’s point of view, it may appear that the representation adopted by the market is chosen because of its objectivity. In conclusion, this reflection leads me to propose an analysis of financial markets fully in keeping with Durkheim’s observations. Instead of considering the representation of the future, conveyed by prices, as an objective fact, we should see it as the result of a self-referential process of opinion.

14.2 EFFICIENCY THEORY AND THE HYPOTHESIS OF THE OBJECTIVITY OF THE FUTURE The fundamental value of securities raises particularly difficult questions for economic theory, questions involving the nature of the knowledge that a human society can acquire about its own future. This becomes clear when we remember that the fundamental value of a share is grounded in the flow of future dividends that the possession of this share will provide for its owner. If we set aside the question of discounting, by assuming that it takes place at a constant rate, denoted r, then this fundamental value can be written formally as follows: [VF]

VFt 5

Dt11 Dt12 Dt1n 1 1 ... 1 1 . . . (14.1) (1 1 r) (1 1 r) 2 (1 1 r) n

318

Handbook of knowledge and economics

where VFt denotes the fundamental value of the share at time t and Dt+n denotes the dividend paid at time t + n by the company concerned. The nature of the estimate that can be made of this value, at time t, is highly dependent on the way we analyse, at the present time, the ‘reality’ of the future dividend at time t + n. To put it simply, modern finance theory holds fast to the hypothesis that the future is objectively given in a probabilistic form. This basic hypothesis is already present in the Arrow–Debreu model of general equilibrium. It states that the future can be represented in the form of an exhaustive list of exogenous events or states of the world, assumed to describe everything that it is relevant for an economic agent to know. A given value of the dividend paid is associated with every state of the world e. So a share is described by the payments it generates in each state: d(e) for e [ E, with E being the set of all states of the world. If the future is an objective fact like any other, this naturally leads us to suppose that rational, well-informed agents will necessarily end up knowing it. Consequently, we shall observe the convergence of individual representations towards the ‘correct’ representation, as long as the agents possess all the information and process it rationally. And this is the essential theoretical point that marks the specificity of the hypothesis of probabilistic objectivity of the future. The very existence of an objective future forms a reference that prevents the subjective drift of estimates by anchoring them in an objective foundation that rational activity cannot fail to recognize. Adopting this very strong hypothesis, it is possible to define an optimal expectation at time t, namely that which makes the best possible use of all the available, relevant information, and that is, consequently, independent of the idiosyncratic opinions of the agents. This expectation can be said to be ‘rational’. It is expressed mathematically, thanks to the conditional expectation operator, as follows: [AR]

E [ Dt1n 0 Wt ] 5 EtDt1n

(14.2)

It appears that the only relevant variable is Ωt, the available, relevant information at time t. The fact that subjective opinions can disappear in this way is, as I have already indicated, a direct consequence of the hypothesis that the future is imposed objectively on every rational individual in such a way that, making rational use of his/her faculties of judgement, he/she is necessarily led to adopt this estimate. Here, the fact that this objectivity is of a probabilistic nature is of secondary importance. Elsewhere, I have suggested the term ‘fundamentalist’ to describe the type of rationality at work here (Orléan, 1999, pp. 65–6). Turned towards nature, it has the declared aim of elucidating objective truths. It

Knowledge in finance

319

is this rationality that underpins the mathematical expression of equation (14.2). The hypothesis of informational efficiency (HIE) of financial markets is deduced from this. It states that the financial market is a space within which, thanks to the competition between rational agents, the price formed at any given moment is the best possible reflection of the fundamental value, given the information available. This can be simply written as follows: [HIE]

Pt 5 E [ VFt 0 Wt ] 5 Et (VFt)

(14.3)

This is exactly how Fama (1965) defined the concept of efficiency in financial markets: ‘In an efficient market at any point in time the actual price of a security will be a good estimate of its intrinsic value’ (Fama, 1965, p. 56). Or again: ‘In an efficient market, the actions of the many competing participants should cause the actual price of a security to wander randomly about its intrinsic value’ (ibid.). This conception is founded entirely on the assumption that it is possible to define, at time t, an accurate and unequivocal estimate of the intrinsic value. Otherwise, these definitions would make no sense. Within this theoretical framework, the fundamental value objectively pre-exists the financial markets, the central role of which is to provide the most reliable and precise estimate, in accordance with equation (14.3). Consequently, we can say that the hypothesis of informational efficiency sees finance as a faithful ‘reflection’ of the real economy. From this perspective, financial evaluation has no autonomy, and it is precisely for this reason that it can be put entirely at the service of the productive economy, to which it delivers the signals that enable capital to be invested wherever it will be most useful. Competition is no more than a stimulus to obtaining this result. The real cognitive force at work here, the one truly responsible for achieving this result, is fundamentalist rationality, in other words the capacity of investors to ‘defeat the dark forces of time and ignorance which envelop our future’ (Keynes, 1936, p. 145). This conception of an objective future enjoys a very wide consensus within the community of economists. Before presenting the reasons that lead us to reject it, I shall consider why economists are so deeply attached to this hypothesis of objectivity. There appear to be three motivations, of varying levels of importance to my purpose. The first, of a very general nature, has already been mentioned in the introduction. It involves the fundamental importance economists attach to the process of objectification in their very conception of successful modelling. A well-formed model is one that reconstructs the relation of individuals with their social and institutional environment in the form of

320

Handbook of knowledge and economics

a relation with objects or mechanisms. A prime example is provided by the neo-Walrasian analysis of market transaction, which treats the different parameters that define the goods in a perfectly symmetrical manner, whether they involve the quality, the location or the time at which they will be exchanged. In such an approach, the relation to time is conceived along exactly the same lines as the relation to space or quality: it is denied all specificity. Subjective beliefs can play no role in such a context, because the individual is facing a social world reduced to objects determined ex ante: individual cognition is strictly reduced to rational calculation alone. The second motivation concerns the belief, widely held in one form or another, that ‘without this hypothesis, there is no salvation’. It is often expressed in debates about the definition of uncertainty, debates that set the dominant probabilistic view, also called the risk approach, against the Keynesian or Knightian approach, which considers the possibility of radical uncertainty, where probabilities cannot be calculated. The idea is that in situations of radical uncertainty, the economist no longer has anything constructive to say. This is clearly expressed by Lucas in the following quotation: Unfortunately, the general hypothesis that economic agents are Bayesian decision makers has, in many applications, little empirical content; without some way of inferring what an agent’s subjective view of the future is, this hypothesis is of no help in understanding behaviour . . . John Muth (1961) proposed to resolve this problem by identifying agents’ subjective probabilities with observed frequencies of the events to be forecast, or with ‘true’ probabilities, calling the assumed coincidence of subjective and ‘true’ probabilities rational expectations . . . [This hypothesis will not be] applicable in situations in which one cannot guess which, if any, observable frequencies are relevant: situations which Knight called ‘uncertainty’. It will most likely be useful in situations in which the probabilities of interest concern a fairly well defined recurrent event, situations of ‘risk’ in Knight’s terminology .  .  . In cases of uncertainty, economic reasoning will be of no value. (Lucas, 1984, pp. 223–4)

In this passage, Lucas explicitly limits the validity of economic reasoning solely to situations of risk, in the Knightian sense of the term, in other words situations in which a certain stationary condition of the world prevails. In this case, it is possible to use the frequencies observed as a foundation for individual estimates. Such is the character of the hypothesis of rational expectation: it identifies subjective probabilities with ‘true’ probabilities. In the opposite case, what Lucas calls ‘uncertainty’ in the Knightian sense of the term, economic theory, unable to postulate anything definite about individual estimates, ‘will be of no value’. Here, even the Bayesian hypothesis is insufficient, as it tells us nothing about a priori probabilities. This analysis therefore leads economists to restrict their

Knowledge in finance

321

conception of the future to a stationary condition. This is hardly surprising. It is obvious that the hypothesis of probabilistic objectivity of the future is of practical use only when one assumes that the world exhibits a certain stationarity, without which it is hard to see how this objective future could be the object of positive knowledge. For this reason, the two hypotheses, although they are distinct, often go hand in hand. The third motivation is more complex. For our present purposes, it is the most interesting, even if no one has expressed it explicitly so far. As the positions involved are largely unformulated, the interpretation I wish to give is assuredly a perilous exercise. It may be that I am barking up the wrong tree and that these ideas exist only in my own mind. I shall try to explain what I believe to be the conception of certain colleagues in the domain of finance who adhere to the efficiency theory even though they are sceptical about the hypothesis of an objective fundamental value. They see this as a metaphysical question, of no practical consequence. Their implicit thesis is that the reasoning can be performed in terms of subjective estimates without any effect on the overall construction. Clearly, the concept of subjective estimate is an easier category to handle than that of objective estimate: one cannot doubt that it is within the capacity of individual agents to form, at time t, subjective estimates about a value that will only be determined and perfectly known at time t + n. To quote Lucas again: At a purely formal level, we know that a rational agent must formulate a subjective joint probability distribution over all unknown random variables which impinge on his present and future market opportunities. The link between this subjective view of the future and ‘reality’ is a most complex philosophical question, but the way it is solved has little effect on the structure of the decision problem as seen by an individual agent. In particular, any distinction between types of randomness (such as Knight’s (1921) distinction between ‘risk’ and ‘uncertainty’) is, at this level, meaningless. (Lucas, 1986, p. 223)

As we can see, Lucas contends that the exact nature of the uncertainty facing the agent is of no importance when it comes to subjective estimate. As far as individual estimate is concerned, Lucas may be right. But it cannot be concluded that this question is also of no importance for the theorist! The efficiency theory makes an absolute assumption that it is possible to define an ‘accurate’ or ‘exact’ estimate, failing which most of its statements would be meaningless. This estimate forms the yardstick against which the efficiency of market evaluations can be measured. From this point of view, the hypothesis of the objectivity of the future plays an essential role: it justifies the possibility that subjective estimates can converge, under the action of fundamentalist rationality alone, towards

322

Handbook of knowledge and economics

the same estimate, namely the optimal or rational estimate given by equation (14.2). In the absence of this hypothesis, the subjective estimate can be seen only as a pure opinion, inescapably subjected to the infinite variability of idiosyncratic interpretations of the world. It is then no longer possible to postulate the existence of a good estimate, one that is true to the objective structure of the economy. For this reason, resorting to subjective estimates alone is insufficient. What could ‘efficient’ mean, in such a context? Consequently, the efficiency theory requires more: it requires the possibility of defining the ‘good estimate’, as the above quotation from Fama underscores. There is no escaping from the hypothesis of the probabilistic objectivity of the future, despite its metaphysical character. So we can sum up the motivations that lead economists from efficiency to the hypothesis of an objective and stationary future in the following manner. It is a vision that corresponds with the most deeply rooted habits of economic modelling (Durkheim, 1975/1908), a vision to which the large majority of economists are all the more strongly attached as they believe that its rejection would inevitably lead economic theory into a dead end. In addition, it may appear to certain theorists that this hypothesis can be reduced to the simple postulation of subjective, rationally performed estimates. In this context, my approach consists in demonstrating that the hypothesis that the future can be known objectively doesn’t hold up, but that, despite the fears of economists, I can discard this hypothesis and still construct an analysis just as rich as that of the efficiency theory, and which contains this latter as a particular case. Further, I believe that this approach presents a much more realistic image of market finance. In this alternative construction, which can be described as ‘conventionalist’ or ‘self-referential’, collective representations play a central role. The two following sections are devoted to a presentation of this alternative approach, which entails a fresh analysis of the way agents form their conceptions and their knowledge.

14.3 THE IRREDUCIBLE SUBJECTIVITY OF FUNDAMENTALIST ESTIMATES To my mind, prevailing finance theory is unsatisfactory both in its conception of the future and in its conception of the knowledge that agents can form about that future. Several arguments can be put forward to support this criticism. The first consists in a direct attack on the idea of an objectively given future, on the grounds that this idea is incompatible with the free will of agents as modelled by this very same economic theory. If we believe in individual free will, then the future must be considered

Knowledge in finance

323

to be undetermined and contingent: it is the product of individual decisions, including those taken in the sphere of finance. To put it more precisely, how can we reconcile the hypothesis that the fundamental value is determined ex ante, even probabilistically, before the market transaction itself, with the supposed efficiency of securities markets in terms of the allocation of the economy’s resources? If this efficiency is to have any meaning, then the securities markets must necessarily have an impact on the determination of profits and dividends, at least for certain companies. But if this is the case, then the fundamental value of the companies concerned is no longer determined ex ante, because it depends on effects produced by the securities market  itself! Clearly, there is a fundamental logical contradiction in assuming simultaneously that the markets reflect a pre-existing reality and that their presence is capable of improving, and therefore transforming, the functioning of the economy. To resolve these difficulties, economic time must be considered as historic time, determined by the collective action of all individuals. Consequently, the idea of finance as a reflection of reality must be abandoned. This contradiction is only resolved in the very specific and perfectly unrealistic context of an Arrow–Debreu-type Walrasian equilibrium, where all values, both present and future, are determined simultaneously. However, even this doesn’t remove all the difficulties. In such a context, where profit is perfectly determined for every state of the world, what meaning can we give to security transactions? As Geanakoplos pointed out, ‘In Arrow–Debreu equilibrium, there is no trade in shares of firms . . . If there were a market for firm shares, there would not be any trade anyway, since ownership of the firm and the income necessary to purchase it would be perfect substitutes’ (Geanakoplos, 1987, p. 121). The hypothesis of a transparent future renders securities markets redundant. A second argument, of a Keynesian or Knightian nature, criticizes the unrealism specific to this conception: have we ever seen such a thing as a description of the future giving an exhaustive account all the events that might possibly occur? ‘About these matters there is no scientific basis on which to form any calculable probability whatever. We simply do not know’ (Keynes, 1937, p. 214). We can only conclude that estimates about the future have an inescapably subjective dimension; they are opinions. This is the argument I shall concentrate on in the rest of this chapter. Not only does it avoid the metaphysical or philosophical complexity of the previous argument, but it can be directly applied to the domain of finance. The inescapably subjective nature of personal estimates can be seen in the fact that no procedure can be described that would enable the convergence of individual estimates of the fundamental value. This is a reality that can be empirically observed: two rational individuals, perfectly

324

Handbook of knowledge and economics

informed and with access to the advice of as many experts as they wish, can maintain divergent estimates, without either acting irrationally or contradicting observed facts. This is essentially because agents can, to justify their point of view, resort to the hypothesis of the non-stationarity of the economy, which creates a link with the idea of Keynesian uncertainty. When we accept the idea that something radically new can appear, then each individual is ‘free’ to have his/her own personal vision of the future. This could be seen quite clearly during the ‘Internet bubble’, when totally wild estimates were justified on the basis of extravagant fundamentalist scenarios. Those who pointed out that the hypotheses contained in these scenarios required unprecedented growth rates or productivity levels2 were told that they were singularly lacking in imagination, and that just because something had never happened didn’t mean that it couldn’t happen in the future. An irrefutable argument, and how true it turned out to be! But if one is permitted to reject the lessons of the past on the grounds, perfectly valid in themselves, that the world is not stationary and that new things are continually appearing in it, then one can eliminate all objections. There ensues an irreducible subjectivity of the fundamentalist evaluation, which appears to us to describe very accurately the situation of real economies. This result leads us to affirm that the fundamentalist estimate should be conceived as being a pure opinion. The reality of the world of finance is characterized by a radical diversity of opinions. We can relate this result to the interesting reflections developed by Mordecai Kurz concerning what he calls ‘rational beliefs’ (Kurz, 1994 and 1996). This author demonstrates that, in the context of a nonstationary economy, perfectly rational agents, possessing the same information, in this case exhaustive information concerning all the economic variables since the economy came into existence, can nevertheless form divergent beliefs. For us, as for Kurz, the essential issue here is that of non-stationarity, because it allows for a multiplicity of interpretations compatible with the data observed. On this subject, Kurz (1994) observes that it is not even necessary for the economy to be truly non-stationary, as long as the agents believe that it may be, because ‘there does not exist any statistical means by which agents can ascertain that a stationary system is, in fact, stationary’. The direct consequence of these analyses is that there does not exist any means of defining ex ante a ‘good’ estimate of the fundamental value. There is only a set of divergent opinions, none of which can benefit from superior scientific legitimacy. In this connection, it may be relevant to note that credit rating agencies, whose financial evaluations can be considered the most objective possible, stress the fact that their ratings are no more than ‘opinions’, protected by the first amendment of the American

Knowledge in finance

325

Constitution. They go so far as to compare the ratings they give to a newspaper editorial, ‘the world’s shortest editorials’,3 and this is more than just legal quibbling to protect themselves against dissatisfied customers. In these conditions, the concept of efficiency is disproved. But for many theorists, as mentioned above, this represents nothing less than the outrageous provocation of a theoretical crisis. They believe that in such a state of affairs, it will no longer be possible to say anything positive. I hope to demonstrate that this is not the case: instead of assuming that there exists an objective representation of the future imposed on all the agents, we should rather consider the financial markets as the producers of conventional representations that serve as points of reference for investment decisions. As I see it, this is the only line of reasoning suited to the non-stationary nature of economic time. It is absurd to assume the ex ante existence of an objective knowledge of the future shared by all rational, informed economic agents. We must, on the contrary, suppose that this collective representation is produced by financial interactions, and varies with them. Before analysing the financial processes that lead to the emergence of the financial convention, I shall describe a few empirical intuitions about its contents. At the most basic level, the financial convention can be simply defined as a shared way of interpreting future economic developments. One example of this would be the ‘New Economy convention’, according to which the future of developed capitalist economies lay essentially in the diffusion of the new information and communication technologies (ICT). This convention prevailed in the international financial markets at the end of the 1990s. The belief was that with the appearance of ICT, capitalism was entering a new era of productivity marked by the end of traditional cycles. The result was a wave of excessive optimism about the future profitability of companies connected with e-business. As this event is undoubtedly still fresh in everyone’s memory, I shall not go into great detail here. What should be pointed out, however, is that a convention determines more than just the definition of a ‘scenario of reference’, however important that may be in the formation of expectations. We must go further, and also consider the battery of specific criteria it constructs to serve as a basis for the concrete valuation of companies. Thus, in the case of the ‘New Economy convention’, faced with the difficulty of accounting for stock market prices solely on the criterion of profits, as most ‘dot.com’ businesses were loss-making, a new basis for making estimates appeared, in the form of ‘value per user’. So the potential number of subscribers, visitors or customers was adopted as the strategic variable, supposed to enable the level of value creation to be assessed. This was a most unreliable hypothesis. Furthermore, the first company to develop the ‘e-commerce’ of any given product was believed to benefit from a prohibitive advantage,

326

Handbook of knowledge and economics

making it the natural leader in the sector and endowing it with a large premium in its valuation. This reasoning was equally unsound. To sum up, the financial convention comprises a certain interpretation of the future development of the economy, combined with a set of specific conventions about valuation. It serves as a reference to all investors, including for physical investment. It can be represented formally, in the Arrow–Debreu manner, as a distribution of probabilities of future events. However, instead of being considered as the ‘true’ distribution of probabilities, it should be perceived as a conjecture chosen by market convention at a given moment. From a conceptual point of view, this is a fundamental difference. We no longer assume that there exists only one relevant representation of the future, or only one possible valuation of securities. Because the future cannot be known objectively, there exists a plurality of possible legitimate interpretations. These depend on the opinions of the market, and they cannot be entirely justified by a fundamentalist type of analysis. A good example is provided by the retail toy market studied by Shiller (2001, p. 176). At the beginning of the 1990s, this market was dominated by the venerable, long-established company Toys ‘R’ Us. This domination was vigorously challenged by the newcomer eToys, created 1997, which threw itself headlong into the development of e-commerce. Comparing these two companies objectively, we have, on the one hand, a company with an undeniable wealth of savoir-faire and experience, making, in 1998, a profit of $376 million from a turnover of $11 billion in 1146 shops, and, on the other hand, a company with no experience, making a loss of $28 million in the same year from a turnover of $30 million. In other words, eToys weighed the equivalent of three Toys ‘R’ Us shops and made losses when its rival was making profits. And yet at the end of 1999, despite these eloquent figures, eToys was valued at 30 per cent more on the stock market than the American giant of the toy industry! To value these two firms in such an absurd manner, the market had to believe not only that the entire future of the economy lay in e-commerce and that the first to enter this market would possess structural advantages, but also that older firms would be incapable of adapting to the new situation. The following years were to prove how misguided both of these judgements were. eToys went bankrupt in 2001, with a share value of no more than a few cents, while Toys ‘R’ Us developed a successful Internet business by forming an alliance with Amazon. This is no isolated case. It provides a good illustration of the valuation criteria adopted by the ‘New Economy convention’, and of the fact that these criteria were based on a very arbitrary conception of future economic development. As we have already pointed out, this valuation contradicts neither the observed facts nor rationality, but it has the character

Knowledge in finance

327

of one conjecture among many. It has the status of opinion, of a belief in the anticipated omnipotence of e-commerce. In this sense, it should be described as conventional, which doesn’t mean that it can be just anything. We must not forget that, ex post, the investors can judge perfectly well whether or not the conventional predictions have proved to be accurate. This is an important fact, which greatly limits the arbitrariness of conventions. The interpretation of the future adopted by the convention must be backed up, if not by full verification, then at least by an absence of contradiction in the economic developments subsequently observed. For a convention to endure, the observed facts must be in keeping with the predicted facts. In other words, although the initial choice of the convention may be arbitrary to a certain degree, it must accord with economic reality to a certain extent, if it is to survive. This was demonstrated by the Internet bubble, which burst when the financial and technical efficiency of companies in the New Economy proved to be much weaker than expected. If, ex ante, ‘we simply do not know’, ex post we can judge the accuracy of conventional valuations. Today, for example, we can say with certainty that the ‘New Economy convention’ was erroneous, in that it resulted in an over-accumulation of capital in certain sectors. There is, however, nothing automatic about the falsification of a convention: it is only abandoned after a continual accumulation of anomalies. This is illustrated by the ‘Asian Miracle convention’, which dominated the valuation of SouthEast Asian countries during the mid-1990s. During the first half of 1997, it needed a whole catalogue of bad news from the region (trading deficits, resounding bankruptcies, an increase in doubtful credits) before investors finally lost their belief in the ‘Asian Miracle convention’. These reflections have led me to relate the concept of financial convention to the concept of paradigm developed by Thomas Kuhn (Orléan, 1999). Kuhn, an epistemologist, proposed this concept to analyse the way in which scientific communities organize themselves, choosing certain promising lines of research and rejecting others. The link with our own subject of reflection lies in the fact that these communities, like stock markets, are faced with Keynesian type uncertainty, as there is no objective basis for determining which fields of research or theoretical hypotheses should be favoured. Given these conditions, how does the scientific community organize its research work? Kuhn says that it chooses a ‘world view’, partly arbitrary, and sticks to it as long as it is not totally discredited by a persistent accumulation of anomalies. Contrary to the mechanical application of the Popperian idea of falsification, in the case of an isolated contradiction between the adopted theory and the observed facts, the paradigm is not immediately abandoned. Indeed, it is thanks to this property that the dynamism of scientific enterprise can find its full expression. Without it, research

328

Handbook of knowledge and economics

would constantly be impeded by the occurrence of problematic results, not immediately in keeping with the predictions of the paradigm. The paradigmatic organization, when faced with such problematic results, gambles that they are only local anomalies that will later be solved, without requiring the whole system to be called into question. This is a method of proven effectiveness. I believe that the financial community proceeds in a similar fashion. It does not abandon a convention at the first sign of incoherence. There must be a series of cumulative anomalies before investors react. This avoids the risk of discarding a convention that may otherwise give good results on the strength of a few isolated, ambiguous facts. This comparison brings out the fully rational character of the conventional organization of knowledge, as it draws its model from the most rational of all human activities, that of scientific research itself. This leads to an analysis that sees financial markets as cognitive structures, ‘collective cognitive systems’ (Favereau, 1989) that produce diverse conjectures, certain of which are then selected. There is an irreducible element of arbitrariness in this selection, deriving from the Knightian uncertainty surrounding the question involved. For this reason, it is not possible a priori to be sure of having made the right choice. Consequently, the choice of one convention rather than another necessarily takes the form of a gamble. This is in no way a constraint that can be surmounted, because in a domain dominated by Knightian uncertainty, it is the best one can do. Valuations and investments are produced on the basis of collective adherence to this gamble, and these, in return, shape economic reality. As long as what is produced is in accordance with predictions and satisfactory for investors, the convention endures. But when the observed facts are too far in contradiction with the prevailing conventional representation of the world, then anomalies accumulate and the market ends up abandoning this convention and seeking another. This historical dynamic that produces a partly arbitrary selection, temporarily confirmed in the facts that it helps to produce and eventually discarded when its time has passed, is similar to that described by Schumpeter in relation to the cycle of technological innovations. In both cases, the idea that criteria might exist enabling the right choice to be made ex ante and with certainty must be rejected. In other words, this analysis presents us with a fully historical temporality, made of trial, error and learning, in which there is no optimality and above all no ex ante optimality. The future is radically undetermined. It is the result of individual choices which are themselves dependent on the way in which agents conceive their future. As Schumpeter had already observed, the central error in the orthodox approach is the wish to defend the a priori optimality of choices in a domain – the evolution of human societies – where this makes no sense, unless we abandon the perspective of historical

Knowledge in finance

329

time for that of logical time. The model of trial, error and learning is the only appropriate model for a historical conception of time. In such a context, the development of individual and collective knowledge cannot be understood solely on the basis of the fundamental economic data, for it also depends on other factors, such as the various beliefs and values that structure the social environment.4 Where the traditional economic model considers the development of knowledge as a reflection of objective reality, our model considers collective knowledge as resulting from the financial interactions themselves. The central issue is to understand what it is that causes the emergence of one unique conventional valuation of reference out of a heterogeneous group of individual beliefs. And the key to this understanding lies in the concept of self-referentiality.

14.4 COLLECTIVE BELIEFS AND SALIENCE: THE SELF-REFERENTIAL APPROACH The empirical illustrations presented above are all founded on the implicit hypothesis that, at a given moment in time, a unique conception of the future development of the economy becomes established in the market, what I have called a ‘financial convention’. Within the Arrow–Debreu approach, this hypothesis can be directly deduced from the hypothesis of the objectivity of the future. This theoretical approach maintains that there exists an accurate representation of the future development of the economy, which is reflected in prices. How can the hypothesis of a single representation continue to be accepted within our analytical framework? To the extent that we hold knowledge of the future to be a matter of heterogeneous, idiosyncratic opinions, can it be justified to apprehend the market price as being the expression of a unique, coherent representation of the future rather than the chaotic reflection of divergent views? In other words, by what miracle does a unique representation end up prevailing in a world of heterogeneous opinions, all equally rational and well informed? This is the essential theoretical question, and it is of fundamental importance: we are judging the financial market’s ability to produce a coherent representation of the future, to serve as a reference for investors. One way to tackle this question is to assume that each agent fixes his demand for securities on the basis of his personal conception of the fundamentals, without any reference to other existing conceptions. Let Ri denote the representation specific to investor i. On the basis of this representation, the financial investor i can infer a certain distribution of the future price and dividend. To illustrate, let us consider a much-simplified situation in which the N investors only have the choice between a non-risk asset

330

Handbook of knowledge and economics

yielding the rate r and a share. Let us assume that the N investors all have a CARA-type utility function. Then the net demand of individual i for this share at time t, which we denote Xi(t), can be written: [D]

Xi (t) 5

di (t 1 1) 1 pi (t 1 1) 2 (1 1 r) p (t) aiVi (t 1 1)

(14.4)

where di(t+1) is individual i’s expectation, at time t, of the dividend d(t+1) that will be distributed at time (t+1); pi(t+1) is the expectation of the price p(t+1); Vi(t+1) is the expectation of the variance of price at time (t+1), and ai is the risk-aversion of this same investor i. On these bases, it is not very difficult to find the equilibrium price. This must satisfy the equation of the equality of supply, assumed to be fixed and equal to X securities, and total demand: [EQUI]

a Xi (t) 5 X

(14.5)

i

This gives an equilibrium price p(t) that depends, in particular, on pM(t+1), the market expectation, where this last value is the weighted average of pi(t+1), taking the inverse of aiVi(t+1), which measures the ‘weight’ of investor i, as the weighting coefficient. This can then be written: [PRIX]

pM (t 1 1) 5

1 i5N

1 a a Vi (t 1 1) i51 i

pi (t 1 1) a a Vi (t 1 1)

i5N i51

(14.6)

i

Now, if the weighted average of the prices is indeed a price, the same cannot be said of the representations Ri that underpin these prices. Let us assume, for example, that individual a reckons that future economic development will be focused on biotechnologies, and that on this basis he expects a price equal to 100, whereas individual b believes there will be a financial crisis resulting in a price of 50. Assuming the two investors are of equal ‘weight’, we can then calculate, for the sake of argument, an average price of 75. Yet this price of 75 does not correspond to any coherent representation of the future. It is impossible to calculate the average of distinct qualitative conceptions. What meaning could we possibly give to an economic situation resulting half from the development of biotechnologies and half from a financial crisis? For this reason, I have proposed a completely different approach, in which a coherent and specific representation of the future prevails. The initial idea consists in regarding the stock market as a public space of opinions and communication, in which ideas and conjectures compete

Knowledge in finance

331

with each other. Now, I believe that the characteristic feature of this financial competition is its self-referentiality (Orléan, 1999): in a financial market, each agent tries to predict as accurately as possible what the majority opinion will be. For this reason, the investor concerned with his financial profit is extremely attentive to the way in which the collective opinion is formed. The result is a structure of interactions far removed from the fundamentalist model, in that the norm adopted is not an objective reality exogenous to the market, that is, the fundamental value, but an endogenous variable, namely the opinion of the market as conveyed by the price. Contrary to the fundamentalist model, expectations are not oriented towards the real economy, but towards the expectations of the other agents and even, more precisely, towards the expectation of the market as an entity, which we have denoted pM. My thesis is that it is this self-referential process that produces the market opinion: because each agent must determine his/her position in relation to the market opinion, he/she must conjecture about what this opinion is; by doing so, they give it life. The importance of this result must be clearly understood: treating the market as a specific, autonomous entity is not the result of a proclivity for holism, but of the very rules of the stock market game, which lead every agent to position himself in relation to the market. In my analysis, in other words, pM is no longer simply the ex post result of individual behaviours, as in the previous model; it is now what the agents themselves expect. Consequently, we ought to replace the pi in equation 14.4 by (pM)i, because the agents determine their position according to the average expectation of opinion. We can see straight away that the process doesn’t stop at this first level of anticipation. For when the investors act according to (pM)i, a new average expectation of the market (pM)M is formed, which in turn leads to new expectations of the type [(pM)M]i, and so on. Such is the cognitive dynamics that we should analyse. To study this structure, Keynes proposed a simple and enlightening illustration: the celebrated ‘beauty contest’. He wrote: professional investment may be likened to those newspaper competitions in which the competitors have to pick out the (six) prettiest faces from a hundred photographs, the prize being awarded to the competitor whose choice nearly corresponds to the average preferences of the competitors as a whole. (Keynes, 1936, p. 146)

In this competition, each competitor’s opinion about which faces he truly thinks are the prettiest is of no importance. What matters is deciding how the others will approach the question, to get as close as possible to the majority view. However, if we assume that all the competitors are equally rational, it follows that the others’ opinions are also determined by what

332

Handbook of knowledge and economics

they themselves think the group opinion will be. As Keynes put it: ‘each competitor has to pick, not those faces which he himself finds prettiest, but those which he thinks likeliest to catch the fancy of the other competitors, all of whom are looking at the problem from the same point of view’ (ibid.). We are therefore dealing with a ‘specular’ structure: like the infinite reflections in a hall of mirrors, each competitor tries to read the minds of the others, who are themselves engaged in the same task. This forms a complex structure. Fortunately, thanks to the experimental work of Mehta et al. (1994), we now have a clearer idea of the way it works. We can sum up the main results as follows. The first empirical result these authors obtained, already highlighted in his time by Thomas Schelling (1960), is that when they are placed in this type of situation, a large proportion of agents succeed in converging on the same opinion. In other words, even in the absence of explicit communication between the players, one opinion emerges that attracts a high percentage of them. Here, I touch on one of the key elements of my approach: self-referential interactions possess the property of producing an opinion of reference, even when the initial, subjective opinions are widely dispersed. This is an oft-tested empirical fact: the degree of convergence in opinions shows a large increase. This is because all the players are seeking a consensual opinion, and they actually prove capable of causing such an opinion to emerge. How do they do it? What is the nature of this consensual opinion? One approach consists in examining the personal, subjective opinions of the players and determining which of them is the most popular – in other words, seeking the mode of distribution of the initial, subjective opinions. In the financial context that interests us here, the initial, subjective opinions are the a priori fundamentalist representations of the investors. However, the second fundamental result of Mehta et al. is that this is not the process the players actually follow. They do not try to determine each others’ personal or fundamental opinions, which remain opaque; they seek to discover a ‘salience’, that is, the opinion that imposes itself on the greatest number of players as expressing the opinion of the group in question. In other words, the cognitive work consists in examining the group itself and the way it is perceived by each player, so as to determine the opinion most characteristic of the group. In most situations, the salient opinion is clearly distinct from personal opinions. This result has important consequences when transposed to the domain of finance, as it tells us that the market price is not a direct expression of fundamentalist estimates. I have devoted several texts to the detailed analysis of this point (Orléan, 2004 and 2006b), and shall not return to it here. To conclude, I shall simply observe that the creation of a salient opinion is highly dependent on the historical and social context. Ex ante, the

Knowledge in finance

333

representations are not all of ‘equal power’: some of them are more likely than others to attract the consensus of the market. This particular power is what we might call their ‘symbolic force’. It is defined by the capacity to create salience. However, no simple model exists by which it can be evaluated. The most we can do is point out the large number of parameters that appear to play a part. Note for example, the role of precedents, already indicated by Schelling (1960). The fact that one or another representation has been adopted by the market some time in the past is an important element. It has been demonstrated, for example, on the basis of work by Shiller (1991), that the 1929 crash became established in 1987 as the salient model of financial crisis, provoking a severe panic among operators. But many other elements are also involved: the a priori beliefs and values of the agents; the degree of influence of economic players; whether or not scientific views contribute elements of confirmation. All in all, it appears that the tools and reflections required to understand this capacity to create salience exceed the traditional framework of economic analysis. We need to observe how the other social sciences, which have a longer experience of these phenomena, approach the question. Just as economists are at home in the quantitative sphere, they are still beginners in the analysis of representations. In writing a history of financial conventions, for example, one comes up against the difficulty that representations are not ‘stored’ in the same way as prices, the history of which can easily be found in databanks. This new collaboration with the other social sciences required for a deeper understanding of financial conventions brings me back to Durkheim’s analysis (1908). As shown in the introduction, if Durkheim emphasized how much he felt it necessary to distance himself from the hypothesis of objectivity, he didn’t see this as an insurmountable crisis for economics but, on the contrary, as a welcome opportunity to redefine its relations with the other social sciences, in the direction of greater solidarity. For it is by attaching central importance to the hypothesis of objectivity that economists have built a conceptual wall, enclosing economic reasoning within an isolated region of the social space. Replacing objectivity with the hypothesis of opinion and convention is the condition for economics to rediscover its natural place within the social sciences, in other words sciences concerned with social facts that are always facts of opinion. This is a theoretical revolution of prime importance (Orléan, 2006a).

14.5 CONCLUSION In this chapter, I have set out to demonstrate that finance theory has everything to gain from rejecting the hypothesis of the ex ante objectivity

334

Handbook of knowledge and economics

of the fundamental value, even if this requires us to abandon the idea of informational efficiency in its traditional form. It is not true, at any time, that the market forms the ‘best’ possible estimate of the value of companies, because this ‘best’ estimate does not exist. Prices are necessarily something of a gamble when we consider a non-stationary world dominated by Knightian uncertainty. Is this not borne out by the history of the Internet bubble? It would be absurd to judge the criteria of evaluation produced by the ‘New Economy convention’ as being optimal, given the information available. Nevertheless, they did possess a certain rationality. These were gambles justified by the computer revolution. Some of them proved to be accurate, for example those concerning Microsoft or other Internet-related companies, others were erroneous. This is the line of reasoning I have used to criticize the hypothesis of efficiency. In fact, however, as Hyme (2004) pointed out, there coexist two distinct definitions of the concept of efficiency. In the first, the emphasis is placed on the link between market prices and fundamental values. In the second, there is efficiency when the play of the stock market is fair. It is the first of these definitions that I have criticized. The second is perfectly compatible with the conventionalist approach, in that it can be deduced from the idea of competitive equilibrium. In other words, the divergence between the standard approach and the conventionalist approach does not revolve around the competition hypothesis, but around the knowledge that agents are capable of producing as to the future development of the economy. Instead of considering it as an a priori fact resulting from an objectively defined future, our whole conceptual effort has been aimed at rethinking this knowledge as the contingent product of opinion-based reasoning. This is our essential result. The self-referential market has the property of transforming a heterogeneous set of individual beliefs into a unique representation, perceived by each agent as an expression of what the market thinks. In this way, an estimate of reference is formed, from which each individual determines their own position. The self-referential hypothesis therefore reconciles the ex ante existence of a heterogeneous set of individual fundamentalist estimates and the ex post emergence of a unique representation that gives the price its significance. Far from being the chaotic summation of qualitatively diverse opinions, price is the expression of this convention. It is at this level that the essential function of the market operates. It is a cognitive function: supplying a representation of the future to facilitate investment decisions, and it is in this manner that the ‘dark forces of ignorance’ are temporarily vanquished. Because our analysis retains the idea that there exists a representation RM, on the basis of which the price becomes intelligible, it follows that it contains the traditional analysis as a particular case,

Knowledge in finance

335

defined as the exceptional situation in which individual opinions all adhere to the same model. We can now define an ‘accurate’ estimate, namely that which all the operators consider to be accurate. This then becomes the estimate on which the self-referential process converges. I have remained brief in my analysis of this self-referential process. The main emphasis has been placed on the fact that the convention possesses an arbitrary dimension. It is one among many possible conjectures. This is a consequence of the fact that ex ante optimal representations do not exist. All the same, this convention cannot be just anything, for two main reasons. First, ex post, investors can compare the conventional predictions with what has actually happened. Admittedly, there is no automatic falsification, but an accumulation of anomalies will result in the convention being rejected. There follows a process of trial and error, which is the form usually taken by the dynamics of knowledge in a world of uncertainty. Second, although the process leading ex ante to the emergence of the convention is only very partially founded on the fundamental data, it is not totally random. The concept of salience goes some way towards making this phenomenon intelligible, although it has not yet provided any very precise results: we can only hope that the fruits of collaboration between economists, sociologists and historians will shed more light on the mechanisms behind the process. From this point of view, once again, my model contains the traditional analysis as a particular case. The traditional model considers a world in which it is possible to decide ex ante between the different competing ideas solely on the basis of their adequacy to the facts. The real world is rarely like that, for the selection of one belief over and above the others cannot be explained by the facts alone. There is a residual part that comes down to the symbolic force of representations and the strength of influence of the economic agents who are the protagonists in this world.

NOTES 1. The concept of ‘convention’ appears in chapter 12 of the General Theory, but with a slightly different meaning from that used in the present text. For a discussion of the different meanings given to the concept of financial convention, see Orléan (1999, pp. 125–45). 2. The only scenarios we can reject are those that fail to respect fundamental economic constraints, such as those, for example, that assume a profits growth rate structurally higher than the rate of economic growth. When this has been taken into consideration, there remains a very wide diversity of estimates. 3. See the American Senate commission investigating the Enron affair, and more precisely the declaration of its chairman Lieberman, www.senate.gov/~gov_affairs/032002lieberman. htm. 4. Popper himself repeatedly stressed (e.g. Popper, 2002) that scientific theories are dependent on the metaphysical conceptions of their authors. There is no such thing as a direct expression of the facts.

336

Handbook of knowledge and economics

REFERENCES Durkheim, E. (1975), Éléments d’une théorie sociale, Paris: Les Éditions de Minuit, Collection: Le sens commun. Fama, E. (1965), ‘Random walks in stock market prices’, Financial Analysts Journal, 21 (5), 55–9. Favereau, O. (1989), ‘Marchés internes, marchés externes’, Revue Économique, 40 (2), 274–328. Geanakoplos, J. (1987), ‘Arrow–Debreu model of general equilibrium’, in John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave, 1, London: Macmillan Press, pp. 116–24. Hyme, Paulette (2004), ‘Des différentes acceptions de la “théorie des marchés efficients”’, Économie Appliquée, 57 (4), 43–57. Keynes, J.M. (1936), The General Theory of Employment, Interest and Money, London: Macmillan. Keynes, J.M. (1937), ‘The general theory of employment’, Quarterly Journal of Economics, 51 (2), 209–23. Knight, F.H. (1921), Risk, Uncertainty and Profit, Boston, MA and New York: Houghton Mifflin. Kurz, M. (1994), ‘On the structure and diversity of rational beliefs’, Economic Theory, 4, 877–900. Kurz, M. (1996), ‘Rational beliefs and endogenous uncertainty: introduction’, Economic Theory, 8, 383–97. Lucas, R.E. (1984), Studies in Business-Cycle Theory, Cambridge, MA and London: MIT Press. Mehta, J., Starmer, C. and Sugden, R. (1994), ‘The nature of salience: an experimental investigation of pure coordination games’, American Economic Review, 84 (2), 658–73. Orléan, A. (1999), Le pouvoir de la finance, Paris: Odile Jacob. Orléan, A. (2004), ‘What is a collective belief?’, in Paul Bourgine and Jean-Pierre Nadal (eds), Cognitive Economics, Berlin, Heidelberg and New York: Springer-Verlag, pp. 199–212. Orléan, A. (2006a), ‘La sociologie économique et la question de l’unité des sciences sociales’, ĽAnnée Sociologique, 55 (2), 279–306. Orléan, A. (2006b), ‘The cognitive turning point in economics: social beliefs and conventions’, in Richard Arena and Agnès Festré (eds), Knowledge, Beliefs and Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 269–99. Popper, Karl (2002), The Logic of Scientific Discovery, London: Routledge Classics. Schelling, Thomas (1977), The Strategy of Conflict, Oxford: Oxford University Press, first edition 1960. Shiller, Robert (1991), Market Volatility, Cambridge, MA and London: MIT Press. Shiller, Robert (2001), Irrational Exuberance, New York: Broadway Books.

PART III ECONOMICS, KNOWLEDGE AND ORGANIZATION

15 Embodied cognition, organization and innovation Bart Nooteboom

15.1 INTRODUCTION This chapter adopts a constructivist, interactionist perspective on knowledge that emerged from the developmental psychology of, in particular, Piaget and Vygotsky. According to this view, cognition is not only the basis of action but also a result of it, and ‘intelligence is internalized action’. We perceive, interpret and evaluate the world on the basis of cognitive categories that we construct in interaction with that world, particularly in interaction with other people. These categories constitute our ‘absorptive capacity’ (Cohen and Levinthal, 1990). As a result, people see the world differently to the extent that they have developed their cognition along different life trajectories, in different environments. In the literature on management and organization, this view has also been called the ‘activity view’ of knowledge (Blackler, 1995), and it has been widely adopted by many scholars of organization (e.g. Weick, 1979, 1995). This view also has roots in sociology, in particular in the ‘symbolic interactionism’ of G.H. Mead (1934). More recently it has received further substance from neural science, in what has come to be known as ‘embodied cognition’, which will be discussed in some detail in this chapter. This perspective has far-reaching implications for economics and management, and enables improved understanding of the ‘knowledge economy’ and the ‘network economy’, or what has recently received the fashionable label of ‘open innovation’ (Chesbrough, 2003). Clearly, if knowledge arises from interaction with others, in a ‘knowledge economy’ interaction between firms in networks becomes crucial. So the claim here is that this perspective of cognition is important for a proper understanding of the ‘knowledge’ and the ‘network’ economy. On a more fundamental level, with its view of the individual as mentally constituted in interaction with others, and hence socially, while preserving individual variety of cognition, this perspective enables us to transcend the ‘methodological individualism’ of economics as well as the ‘methodological collectivism’ of (some) sociology, in a new ‘methodological 339

340

Handbook of knowledge and economics

interactionism’, and thereby helps to make a novel combination of economics and sociology, in what may perhaps be seen as a newly emerging integrative behavioural science. Also, this perspective has implications for the theory of the firm, including a theory of inter-organizational relations (IOR). If people construct their cognition differently along different life paths, this yields ‘cognitive distance’, and this, in turn, yields the need for an organization to act as a ‘focusing device’ in order to sufficiently align cognition, by some ‘organizational cognitive focus’, to utilize opportunities for complementary capabilities. By definition, this yields some organizational myopia, and to compensate for this firms require outside relationships with other firms, at greater cognitive distance. This also has implications for the theory of innovation, particularly in innovation networks. The question there is to what extent the structure of a network and a firm’s position in it affect the variety of cognition and abilities to absorb it. The chapter consists of two sections. The first introduces embodied cognition, and specifies its contrast with the traditional ‘representational– computational’ view of cognition. The second analyses the implications for economics and management.

15.2. THE NEW PERSPECTIVE OF EMBODIED COGNITION1 15.2.1

The Traditional View

The perspective of embodied cognition stands in opposition to the ‘representational–computational’ (RC) view that has been dominant in cognitive science. That view assumes that knowledge is constituted by symbolic mental representations and that cognitive activity consists of the manipulation of (the symbols in) these representations, called computations (Shanon, 1988, p. 70). According to Shanon (1993), the representations according to the RC view are: 1. 2. 3. 4.

symbolic: in the use of signs there is a separation of a medium and the content conveyed in it; abstract: the medium is immaterial; its material realization (physiology) is of no relevance; canonical: there is a given, predetermined code, which is complete, exhaustive and determinate; structured/decomposable: well-defined atomic constituents yield wellformed composites;

Embodied cognition, organization and innovation 341 5.

static: mind is the totality of its representations; structure and process are well demarcated.

The basic intuition is that behaviour is based on beliefs, desires and goals, and representations are postulated as entities that specify them (Shanon, 1993, p. 9). The reconstruction of variety as variable, combinatorial operations on fixed elements is an ancient ploy: the ploy of decomposition. In formal grammar it yields the ‘standard principle of logic, . . . hardly ever discussed there, and almost always adhered to’ (Janssen, 1997, p. 419), that the meaning of a compound expression is a function (provided by rules of syntax) of the meanings of its parts. It was adopted by Frege, in his later work (Frege, 1892; Geach and Black, 1977; Thiel, 1965; Janssen, 1997). The motivation for this view is in a respectable scientific tradition to yield a parsimonious reconstruction, in terms of stable entities and procedures of composition of those entities into a variety of structures, to account for orderly and regular human behaviour across a large variety of contexts. It also explains how people can understand sentences they never heard before. A subsidiary motivation is that by interposing the cognitive as an intermediate, abstract level between psychological phenomenology and physiology, we can circumvent the need for a full reconstruction in terms of physiology, and we can thereby evade reductionism. However, there are empirical and theoretical objections to such a symbolic, semantic, representational view (Shanon, 1988; Hendriks-Jansen, 1996). If meanings of words were based on representations, it should be easy to retrieve them and give explicit definitions, but in empirical fact that is often very difficult. A second empirical point is that people are able to recategorize observed objects or phenomena, so that representations vary, if they exist, and then they are no longer determinate. Words generally have more than one meaning, and meanings vary across contexts. Closed, that is, exhaustive and universal, definitions that capture all possible contexts are often either infeasible or extremely cumbersome. For most definitions one can find a counter-example that defeats it. For example, what is the definition of ‘chair’? Should it have legs? No, some chairs have a solid base. Not all chairs have arm rests or back rests. Neither has a stool, but we distinguish it from a chair. A child’s buggy seat on a bike has a back rest, but is not called a chair. At least in some languages, a seat in a car is called a chair. A chair is used for sitting, but so is a horse. A cow is not a chair, but years ago I saw a newspaper item ‘watch him sitting in his cow’, with a picture of someone who used a stuffed cow for a chair. If it were customary for people living along a beach to collect flotsam to use for chairs, it would make sense, when walking along a

342

Handbook of knowledge and economics

beach, to point to a piece of flotsam and say ‘look – what an attractive chair’. Not to speak of professorial chairs. Another empirical point of fact, recognized by many (e.g. Putnam, 1975; Winograd, 1980), is that meanings are unbounded, and open-ended with respect to context. Novel contexts do not only select from a given range of potential meanings, but also evoke novel meanings. Novelty is produced in contextual variation (Nooteboom, 2000). Summing up, representations cannot be exhaustive, or determinate, or single-valued, or fixed. As Wittgenstein (1976) proposed in his Philosophical Investigations, in his notion of ‘meaning as use’, words are like tools: their use is adapted to the context, in the way that a screwdriver might be used as a hammer. One of the theoretical problems, recognized by Fodor (1975), who was a proponent of RC, is the following: if cognitive activity is executed by computation on mental representations, the initial state must also be specified in terms of those representations, so that all knowledge must be innate. That is preposterous, and certainly will not help to develop a theory of learning and innovation. Another theoretical objection is that if one admits that meaning is somehow context-dependent, as most cognitive scientists do, and if they are adherents of the RC view, then according to the RC view, context should be brought into the realm of representations and computations. Shanon (1993, p. 159) characterizes this as the opening of a ‘disastrous Pandora’s box’. To bring in all relevant contexts would defeat the purpose of reducing the multiplicity of cognitive and verbal behaviour to a limited set of elements that generate variety in the operations performed on them. Furthermore, we would get stuck in an infinite regress: how would we settle the context dependence of representations of contexts? Note that contexts in their turn are not objectively given, somehow, but subject to interpretation. As Shanon (1993, p. 160) put it, ‘If the representational characterization of single words is problematic, that of everything that encompasses them is hopeless.’ In recent developments in the logic of language, the notion has come up of ‘discourse representation theory’. In the words of van Eijck and Kamp (1997, p. 181), ‘Each new sentence S of a discourse is interpreted in the context provided by the sentences preceding it . . . The result of this interpretation is that the context is updated with the contribution made by S.’ The contribution from this theory is that it yields a dynamic perspective on semantics: truth conditions are defined in terms of context change. This theory can even be formalized so as to preserve compositionality (Janssen, 1997). However, I propose that the dynamic of interpretation and context is more creatively destructive than is modelled in discourse representation theory: the interpretation of a novel sentence can rearrange the perception of context and transform interpretations of past sentences.

Embodied cognition, organization and innovation 343 Summing up, compositionality is problematic due to context dependence plus the fact that contexts themselves are subject to interpretation and reinterpretation. Or, to put it differently: the meaning of the whole is not only determined by the meaning of the parts, but feeds back into shifts of meaning of the parts. 15.2.2

Situated Action

I don’t see how we can account for learning and innovation on the basis of representations that satisfy any, let alone all, of the assumptions of RC: separation of medium and content; a predetermined, complete, exhaustive and determinate code; well-defined and static constituents of composites. However, this does not mean that we need to throw out the notion of mental representations altogether. If we do not internalize experience by means of representations, and relegate it only to the outside world, how would cognition relate to that world? How can we conceptualize rational thought other than as some kind of tinkering with mental models, that is, representations that we make of the world? Despite his radical criticism of the RC view, even Shanon (1993, p. 162) recognized this: ‘On the one hand, context cannot be accounted for in terms of internal, mental representations . . .; on the other hand, context cannot be accounted for in terms of external states of affairs out there in the world . . .’. For a solution, he suggests (ibid., p. 163): Rather, context should be defined by means of a terminology that, by its very nature, is interactional. In other words, the basic terminology of context should be neither external nor internal, but rather one that pertains to the interface between the two and that brings them together.

Similar criticism and conclusions were offered by Hendriks-Jansen (1996), who concluded that we should take a view of ‘interactive emergence’, and Rose (1992), who proposed the view of ‘activity dependent selforganization’. This leads to the ‘situated action’ perspective. This perspective entails that rather than being fully available and complete prior to action and outside of context, mental structures (‘representations’) and meanings are formed by context-specific action. One could say that up to a point the situated action view goes back to early associationist theories of cognition, proposed, in various forms, by Berkeley, Hume, William James and the later behaviourist school of thought (Dellarosa, 1988, p. 28; Jorna, 1990). However, a crucial difference with behaviourism (notably the work of Skinner and his followers) is that here there is explicit concern with internal representation and mental processing, even though that does not satisfy the axioms of the RC view.

344

Handbook of knowledge and economics

Nevertheless, in some important respects the ‘situated action’ view seems opposite to the RC view. It proposes that action is not so much based on cognitive structure as the other way around: cognitive structure is based on action. However, the cognitive structuring that arises as a function of action provides the basis for further action. Thus both are true: action yields cognitive structuring, which provides a new basis for action. Rather than taking one or the other position, I take both, in a cycle of development. Knowledge and meaning constitute repertoires from which we select combinations in specific contexts, which yield novel combinations that may shift repertoires of knowledge and meaning. Such shifts of knowledge and meaning occur in interaction with the physical world, in technological tinkering, and in social interaction, on the basis of discourse (cf. Habermas’s 1982, 1984 notion of ‘communicative action’). Situated action entails that knowledge and meaning are embedded in specific contexts of action, which yield background knowledge, as part of absorptive capacity, which cannot be fully articulated, and always retain a ‘tacit dimension’ (Polanyi, 1962). This view is also adopted, in particular, in the literature on ‘communities of practice’ (COP) (Brown and Duguid, 1991, 1996; Lave and Wenger 1991; Wenger & Snyder, 2000). This is related to the notion of ‘background’ from Searle (1992). Interpretation of texts or pictures is based, to some extent, on unspecified, and incompletely specifiable, assumptions triggered in situated action. When in a restaurant one asks for a steak, it is taken for granted that it will not be delivered at home and will not be stuffed into one’s pockets or ears. As a result, canonical rules, that is, complete, all-encompassing and codified rules, for prescribing and executing work are an illusion, since they can never cover the richness and variability of situated practice, which require improvisation and workarounds that have a large tacit component that cannot be included in codification of rules, as recognized in the literature on COP (Brown and Duguid, 1991). The proof of this lies in the fact that ‘work to rule’ is a form of sabotage. 15.2.3

Internalized Action

According to developmental psychologists Piaget and Vygotsky, intelligence is internalized action. By interaction with the physical and social environment, the epistemological subject constructs mental entities that form the basis for virtual, internalized action and speech, which somehow form the basis for further action in the world. This internalized action is embodied in neural structures that can be seen as representations, in some sense, but not necessarily in the symbolic, canonical, decomposable, static sense of mainstream cognitive science. In contrast with Piaget, Vygotsky

Embodied cognition, organization and innovation 345 (1962) recognized not only the objective, physical world as a platform for cognitive construction, but also the social world with its affective loading. While according to Piaget a child moves outward from his cognitive constructs to recognition of the social other, according to Vygotsky the social other is the source of the acquisition of knowledge and language. Vygotsky proposed the notion of ZOPED: the zone of proximal development. This refers to the opportunity for educators to draw children out beyond their zone of current competence into a further stage of development. In language acquisition by children, a phenomenon on which Piaget and Vygotsky agreed was that at some point children engage in egocentric speech, oriented towards the self rather than social others, and that this subsequently declines. Piaget interpreted this as an outward movement from the self to the social other – a ‘decentration’ from the self. Vygotsky ascribed it to a continued movement into the self, in an ongoing process of formation and identification of the self and development of independent thought. The reason that egocentric speech declines is that overt speech is partly replaced by ‘inner speech’. Before that stage, however, speech is preceded by and based on sensori-motor actions of looking, gesturing, pointing, aimed at satisfying a want. Werner and Kaplan (1963) demonstrated ‘that reference is an outgrowth of motor–gestural behaviour. Reaching evolves into pointing, and calling-for into denoting.’ They note that ‘it is in the course of being shared with other people that symbols gain the denotative function’. Both Shanon and Hendriks-Jansen use the notion of the ‘scaffolding’ that the context yields. It is reminiscent of Vygotsky’s notion of ZOPED. Literally, a scaffold is used in the building of an arch: stones are aligned along a wooden scaffold until they support each other and the scaffold can be removed. The paradigmatic case in cognitive development of children is the support provided to the infant by its mother. According to the account given by Hendriks-Jansen (1996), infants do not have an innate language capability as claimed by Chomsky. Instead, they have innate repertoires of activity sequences, such as facial ‘expressions’, eye movements and myopic focusing, kicking movements, randomly intermittent bursts of sucking when feeding, random gropings. At the beginning these movements do not signify anything, nor do they seek to achieve anything, and they certainly do not express any internal representations of anything. The mother, however, instinctively assigns meanings and intentions where there are none, and this sets going a dynamic of interaction in which meanings and intentions get assigned to action sequences selected from existing repertoires on the occasion of specific contexts of interaction. Thus the random pauses in sucking are falsely picked up by the mother as indications of a need to jiggle the baby

346

Handbook of knowledge and economics

back into feeding action. In fact it is not the jiggling but on the contrary the stopping of it that prods the baby to resume the action. The taking turns in stops and jiggles does not serve any purpose of feeding, as the mother falsely thinks, but a quite different purpose, for which evolution has ‘hijacked’ what was thrown up by previous evolution. It is ‘used’ to ready the child for the ‘turn taking’ that is basic for communication: in communication one speaks and then stops to let the other speak. Here, the child acts, stops and triggers the mother to action, who jiggles and then stops and thereby triggers the baby to action. At first, the infant can focus vision only myopically, which serves to concentrate on the mother and her scaffolding, not to be swamped by impressions from afar. Later, the scope of focusing vision enlarges, and the infant randomly fixes its gaze on objects around it. The mother falsely interprets this as interest and hands an object to the infant, and thereby generates interest. The child is then prone to prod the mother’s hand into picking up objects, first without and later with looking at the mother. Groping and prodding develop into pointing, which forms the basis for reference that is the basis for meaning and language. While the child points and utters sounds, the mother responds with the correct words, and so language develops. In egocentric speech the child starts to provide his own scaffolding, which further contributes to the development of his own identity. Along these lines, meaning and intentionality do not form the basis for action but arise from it, with the aid of scaffolds from the context. As indicated, according to Vygotsky, overt speech is next internalized, to yield virtual speech, and cognitive constructs serve as a basis for virtual action: to explore potential actions mentally, by the construction of mental models, deduction, mental experiments. While cognition is not necessarily in terms of language, and can to some extent develop without it, its development is tremendously enhanced by language, in the development of internal speech. The notion of scaffolding lends further depth to the debate, in the COP literature, on the role of specific action contexts, in specifying and elaborating the meaning of words, and in generating new meanings. 15.2.4

Connectionism and Neural Darwinism

As indicated, the situated action view contests the idea of semantic representations as a necessary and universal basis for all knowledge, but it allows for representations in some sense as the basis for at least some behaviour. For example, it might be consistent with connectionism: the view that cognition is based on neural nets, which can generate systematic regularity without the explicit specification of generative rules

Embodied cognition, organization and innovation 347 in underlying representations. Such nets are representations in some sense, generated, by some mechanism, from experience in the world (cf. Smolensky, 1988). In parallel distributed processing (PDP: cf. Rumelhart and McClelland, 1987), two radical steps are taken. One is to no longer accept the computer metaphor of sequential processing according to some algorithm, but to approach knowledge and learning in terms of parallel processes that interact with each other. The second is that knowledge is not stored in units, to be retrieved from there, but in patterns of activation in connections between units. Knowledge is implicit in this pattern of connections rather than in the units themselves (Rumelhart et al., 1987, p. 75). What is stored is the connection strengths between the units that allow the patterns to be recreated (McClelland et al., 1987, p. 9). Edelman’s (1987, 1992) ‘neural Darwinism’ seems to yield a viable perspective for understanding how situated action might work in terms of neural networks (or ‘neuronal groups’, as he calls them). Here, the development of neural groupings, in patterns of connected neural activity, is seen as evolutionary, and more specifically as Darwinian, in that neuronal groups are to a large extent randomly generated, and then selected and reinforced according to success in the actions that they generate. In this way, the selection environment for individual action ‘generates’, by selection, mental structures in the way that in evolutionary economics the competitive and institutional environment ‘generate’, by selection, organizational structures. This yields an explanation of how activity gets internalized in the form of neural structures. As an example of context-dependence of cognition, according to Edelman, memory, both short and long term, is not the ‘retrieval’ of some entity, but a process of recategorization; of reactivating, and in the process possibly shifting, the process of selection among neuronal groups. Hence memory also is context-dependent, and the process of recall may affect the template of future recall. The difference between connectionist models of PDP and neural selectionism is that the former aims to operate on some notional, abstract level between symbols and neural networks (Smolensky, 1988), whereas the latter operates directly on the level of neuronal groups. PDP retains symbols as some higher level, aggregate, emergent outcome of lower-level processing. There is further evidence for the constructivist, activity-based view from other modern research of cognition, in addition to the work of Edelman.2 While the brain has some domain specificity, that is, localization in the brain of cognitive functions, this specificity is plastic, that is, it is not fixed prior to experience, but is constructed from input. For example,

348

Handbook of knowledge and economics

blind people have been shown, with brain imaging techniques, to employ the visual cortex for object recognition. Another result that illustrates activity-based cognition is that after people learn to use objects as tools, accompanied by activity, observed with brain imaging, in motor areas of the brain, the mere observation of the tools triggers brain activity not only in the visual cortex, but also in that motor area. It has been shown that people from different cultures focus on different parts of images, and observe change in patterns differently. In this way, Edelman’s work, and other results from recent cognitive research, underpin the activity-based, constructivist view and its criticism of the earlier representational–computational view, which was still part of Herbert Simon’s view, and of some contemporary artificial intelligence. The central point here is that a mechanism of selection among neuronal structures shows in what way performance may precede competence; how meanings may be constructed from discourse (sensemaking) and knowledge from action (intelligence as internalized action), and provide the basis for ongoing action. This account seems consistent with Johnson-Laird’s (1983) account of mental models and Hendriks-Jansen’s account of how children learn language. This approach indicates how mental structures might emerge from experience in a way that allows for openness and variability across contexts. It offers an evolutionary perspective rather than a perspective of rational design. The programmatic significance of evolutionary theory is that it forces us to explain development not as the result of conscious, goal-directed, top-down, rational design by decomposition of functions, but as selection from among a repertoire of activity sequences, on the occasion of the demands and opportunities of specific contexts. Summing up, as a basis for situated action theory, and the interactionist, constructivist view of knowledge and meaning that it supports, I employ an evolutionary, connectionist theory of cognitive development. On the occasion of experience, selections and recombinations are made from partly overlapping and competing patterns of neural connections (Rumelhart and McClelland, 1987; Edelman, 1987, 1992; Rose, 1992). According to these theories, performance, in interaction with and with support from the context, yields competence as much as competence yields the basis for performance. This underpins a principle of ‘methodological interactionism’. 15.2.5

Embodied Cognition

The principles of situated action, internalized action and neural Darwinism yield what has come to be known as the perspective of embodied cognition.

Embodied cognition, organization and innovation 349 Embodied cognition lends further support to an interactionist, constructivist theory of knowledge that is adopted, explicitly or implicitly, by most authors in the literature on organizational cognition and learning (for surveys, see Hedberg, 1981; Cohen and Sproull, 1996; Meindl et al., 1998). According to this view, people ● ●

construct their cognitive categories, or mental models, by which they perceive, interpret and evaluate phenomena, in interaction with their physical and, especially, their social environment.

This view also appears in the ‘symbolic interactionism’ of G.H. Mead (1934/1984), in sociology, and has later been called the ‘experiential’ view of knowledge (Kolb, 1984) and the ‘activity’ view (Blackler, 1995). In the organization literature, this view has been introduced, in particular, by Weick (1979, 1995), who reconstructed organization as a ‘sense-making system’. The mental frameworks that result from construction constitute ‘absorptive capacity’ (Cohen and Levinthal, 1990). People can turn information into knowledge only by assimilating it into those frameworks, and thereby they shape and mould it. Consequently, to the extent that people have developed their cognition in different environments or conditions, they interpret, understand and evaluate the world differently (Berger and Luckman, 1967). As a result, there is greater or lesser ‘cognitive distance’ between people (Nooteboom, 1992, 1999). A constructivist perspective can slide, and has done so, into radical post-modern relativism. According to the latter, the ‘social constructionist’ notion of knowledge entails that since knowledge is constructed rather than objectively given, any knowledge is a matter of opinion, and any opinion is as good as any other. This would lead to a breakdown of critical debate. Embodied realism saves us from such radical relativism in two ways. First, our cognitive construction builds on bodily functions developed in a shared evolution, and possibly also on psychological mechanisms inherited from evolution, as argued in evolutionary psychology (Barkow et. al., 1992). Second, by assumption we share the physical and to some extent also a social world on the basis of which we conduct cognitive construction. That constitutes a reality that is embodied (Lakoff and Johnson, 1999). As a result of shared psychological mechanisms of cognitive construction and a shared world from which such construction takes place, there is a basic structural similarity of cognition between people. This provides a basis for debate. Indeed, precisely because one cannot ‘climb down from one’s mind’ to assess whether one’s knowledge

350

Handbook of knowledge and economics

is properly ‘hooked on to the world’, the variety of perception and understanding offered by other people is the only source one has for correcting one’s errors. A key characteristic of embodied cognition is that it sees cognition as rooted in brain and body, which are in turn embedded in their external environment. This is consistent with the ‘situated action’ perspective indicated above. The embodiment of cognition entails a continuum rather than a Cartesian duality between rational evaluation, feelings and underlying physiological processes in the body. This view that cognitive functions build on noncognitive, bodily and emotional functions was also prominent in the philosophy of Merleau-Ponty (1942, 1964), and was present, to some extent, in the work of Simon (1983); see also Nussbaum (2001). It plays an important role in Lakoff and Johnson’s (1999) book on ‘philosophy in the flesh’. Building on the philosophy of Spinoza, Damasio (2003) demonstrated a hierarchy of cognition, where rationality is driven by feelings, which in turn have a substrate of physiology, in a ‘signalling from body to brain’. As a result, in this chapter, the terms ‘knowledge’ and ‘cognition’ have a wide meaning, going beyond rational calculation. They denote a broad range of mental activity, including proprioception (i.e. use of hands to grope, feel and grasp objects), perception, sense-making, categorization, inference, value judgements and emotions. Note that the notion of cognitive distance then also refers to a variety of dimensions of cognition. In particular, people may be close in their normative ideas about how people should deal with each other, while they are very different in their substantive knowledge. That is what we often find in organizations, where people with different, complementary competence come together with a shared purpose and style of interaction. In the construction of meaning from actions in the world people employ metaphors, as discussed by Lakoff and Johnson (1980). We grasp our actions in the physical world, in which we have learned to survive, to construct meanings of abstract categories, starting with ‘primary metaphors’ that build on proprioception. Thus, for example, good is ‘up’, because we stand up when alive and well, while we are prostrate when ill or dead. The analysis is important not only in showing how we cope in the world, but also in showing how metaphors can yield what Bachelard (1980) called ‘epistemological obstacles’. I suspect that the primary metaphors, informed by experience with objects in the world, yield a misleading conceptualization of meanings, for example, as objects. Since objects retain their identity when shifted in space, we find it difficult not to think of words retaining their meaning when shifted from sentence to sentence. Underlying this is the ‘museum metaphor’ of meaning: words are labels of

Embodied cognition, organization and innovation 351 exhibits that constitute their meaning, and the ‘pipeline metaphor of communication’: with words, meanings are shipped across a ‘communication channel’. Meanings and communication are not like that, but we find it difficult to conceptualize them differently. In short, in abstract thought, we suffer from an ‘object bias’.

15.3 IMPLICATIONS FOR METHODOLOGY, THEORY OF THE FIRM AND INNOVATION 15.3.1

Methodological Interactionism

If knowledge arises from assimilation of perceptions into cognitive structures that have been constructed in interaction with the world, the self is social, in that it is constructed from interaction, but also individual, in that its constructions and interpretations are to some extent idiosyncratic since they depend on individual life trajectories. This yields a principle of methodological interactionism that transcends both the methodological individualism that forms part of the ‘hard core’ (in the sense of Lakatos, 1970, 1978) of the ‘research programme’ of mainstream economics (Weintraub, 1988), and the methodological collectivism of (some) sociology, according to which individuals are programmed by their social environment. As a result, embodied cognition may yield a perspective for integrating economics and sociology in a new behavioural science. What this can yield, more specifically, in terms of the theory of the firm, including a theory of inter-organizational relationships, and theory of innovation in inter-organizational relationships, is shown, to some extent, in the remainder of this chapter. This chapter cannot exhaust the full potential of methodological interactionism. For example, that principle opens up economics and management to insights from social psychology in human interaction and decision making that are of great importance, particularly in matters that involve combinations of rationality and feelings, as in conflict resolution and the making and breaking of trust (Nooteboom, 2002; Six, 2004). Here, the focus is on more general theory of organization. For that, I make a connection with the sociology of Georg Simmel. Simmel (1950, first published 1917) and Maslow (1954) proposed that people have different levels of needs, motives and cognitive make-up, where lower-level needs must be satisfied before higher levels can come into play (called the principle of ‘pre-potency’), and people are more similar on the lower levels than on the higher levels. In the classic categorization of Maslow, on the lowest level we find the most instinctive, automatic, unreflected and difficult to control drives of bodily physiology,

352

Handbook of knowledge and economics

such as hunger and sexual appetite, which are highly similar between different people. Next, we find needs of shelter, safety and protection. Next, love and affection. Next, social recognition, esteem and legitimation. Finally, at the highest level, individual expression and self-actualization. Higher levels are more idiosyncratic, and hence show greater variety between people, than lower levels.While there is some empirical evidence for a hierarchy of needs (Hagerty, 1999), the principle of pre-potency in particular is far from accurate. The ‘higher-level’ need for esteem and self-actualization can lead people to make great sacrifices on the ‘lower levels’ of safety, shelter and food. Man has a strong, basic, and perhaps even instinctive drive, it appears, toward metaphysics, as exhibited in the form of religious rituals of burial in the earliest forms of Homo sapiens That may even be part of the characterization of our species, in distinction with earlier hominoids. Also, while people may have the same needs on the physiological level of food and sex, the foods and behaviours they choose in order to satisfy those needs vary greatly. Apparently, higher levels find their expression in a variety of ways of satisfying needs on lower levels, in different ‘lifestyles’. Nevertheless, in spite of these qualifications and additions, it still seems true that there are different levels of needs and motives, and that people are more similar on lower levels of more basic needs, perhaps including spiritual ones, and more varied on higher levels of more sophisticated needs. This connects with the notion of cognitive distance. If people make sense of the world on the basis of mental categories that are constructed from interaction with the world, they see and interpret the world differently to the extent that they have developed their cognition along different life paths. Cognition is more similar to the extent that the corresponding phenomena are similar, as in mechanics subject to laws of nature, and is more different in abstractions and in cultural and social life. Simmel (1950) proposed that, as a result, in a randomly composed group of people, what they have in common resides on lower, more basic, unreflected levels of needs and object-oriented cognition as the size of the group increases. What random masses have in common is basic needs and instincts. He also proposed that the larger and more heterogeneous the group, the more norms and rules of conduct for the group are negative, indicating what is forbidden or undesirable, rather than positive, indicating goals and actions to achieve them. The underlying principle of logic is similar to the principle that a theory (with universal propositions) can be falsified but not verified. It is more possible to specify what has been found to be false (in the current context: impermissible) than to specify all that may be possible (here: desirable). To specify what is forbidden entails freedom to do what is not forbidden, while to specify what may be done is

Embodied cognition, organization and innovation 353 either partial, leaving options open, and is then not very functional, or it forbids what is not specified, and then is inherently conservative. The phenomena of levels and variety of cognition have important implications for organizations and IORs. 15.3.2

Theory of the Firm

While the theory of the firm is a familiar branch of economics, it is more appropriate to develop a theory of organization more widely, in which the firm is a special case, as is the practice in the management literature. Several economic theories of organization, in particular transaction cost economics (TCE), look at organizations as systems for governance, to reduce transaction costs by means of incentives, monitoring and control. However, increasingly it is recognized that for a variety of reasons ex ante incentive design is problematic. Due to uncertainty concerning contingencies of collaboration, and limited opportunities for monitoring, ex ante measures of governance are seldom complete, and need to be supplemented with ex post adaptation. Such uncertainties proliferate under present conditions of professional work and, especially, under the conditions of innovation that form the focus of this chapter. Professional work requires considerable autonomy for its execution and is hard for managers to monitor and evaluate, let alone measure. Rapid innovation increases uncertainty and makes formal governance, especially governance by contract, difficult to specify, which increases the importance of collaboration on the basis of trust. If specification of detailed contracts is nevertheless undertaken, it threatens to form a straitjacket that constrains the scope for innovation. Furthermore, the attempt to use contracts to constrain opportunism tends to evoke mistrust that is retaliated by mistrust, while in view of uncertainty there is a need to use trust rather than contract. Beyond governance, we should look at competence or capability (Nooteboom, 2004). Inspired by the work of Penrose (1959), much research of management and organization sees the firm (or organization more widely) as generating firm-specific organizational capabilities. The present chapter can be seen as extending, or deepening, the Penrosian view on the basis of the perspective of embodied cognition. From a competence perspective, incentives by individual rewards may obstruct teamwork, while if the interactionist view of learning is valid, that is crucial for innovation. As noted above, if the situated action view of competence is valid, then canonical rules, that is, all-encompassing and codified rules, for executing work are an illusion, since they can never cover the richness and variability of situated practice, which require informal improvisation and workarounds that have a large tacit component that cannot be included in

354

Handbook of knowledge and economics

codification of rules, as recognized in the literature on COP (Brown and Duguid, 1991). In conclusion, there is a need for an alternative to ex ante incentive alignment, and a basis for ex post adaptation. Using the perspective of embodied cognition, the view in this chapter is that organization functions primarily as a cognitive ‘focusing device’, for reasons of both competence and governance. In order to achieve a specific joint goal, on a higher level than basic needs, the categories of thought (of perception, interpretation and value judgement) of the people involved must to some extent be aligned and lifted to a higher level than the basic instincts that a random group would share (Kogut and Zander, 1992; Nooteboom, 1992, 2000). Alignment entails that cognitive distance must be limited, to some extent and in some respects. More precisely, organizational focus has a dual purpose. The first is to raise shared cognition to a level higher than basic needs and instincts, consistent with, and supporting the goal of, the organization. Also, while outside an organization, in society more widely, norms or rules of conduct tend to be negative, indicating what actions are forbidden or undesirable, organizations need positive norms, indicating goals and ways of achieving them. The second purpose of organizational scope is to reduce cognitive distance, in order to achieve a sufficient alignment of mental categories, to understand each other, utilize complementary capabilities and achieve a common goal. Note that, given the wide notion of cognition used here, focus has perceptual, intellectual and normative content. It includes views of how people ‘deal with each other around here’. To achieve such focus, organizations develop their own specialized semiotic systems, in language, symbols, metaphors, myths and rituals. This is what we call organizational culture. This differs between organizations to the extent that they have different goals and have accumulated different experiences, in different industries, technologies and markets. Organizational culture incorporates fundamental views and intuitions regarding the relation between the firm and its environment (‘locus of control’: is the firm master or victim of its environment?), attitude to risk, the nature of knowledge (objective or constructed), the nature of man (loyal or self-interested) and of relations between people (rivalrous or collaborative), which inform content and process of strategy, organizational structure, and styles of decision making and coordination (Schein, 1985). Organizational focus also has a dual function, of selection and adaptation. In selection, it selects people, in recruitment but often on the basis of self-selection of personnel joining the organization because they feel affinity with it. In adaptation, it socializes incoming personnel with initiation, and focuses their capabilities in training. To perform these functions,

Embodied cognition, organization and innovation 355 focus must not only have cognitive content, but must also be embodied in some visible form. Such form is needed for several reasons. For people to share cognition they need expression in language or other signs. Form is also needed to stabilize the mental processes underlying organizational focus. As such, organizational focus has the same function as the body has for individual cognitive identity. In the theory of embodied cognition it has been recognized that cognition, with its drives of multiple feelings, is diverse and volatile, and often limitedly coherent, and lacks a clearly identifiable, stable, mental identity of the ego, and that such identity, in so far as it can be grasped, is due, in large part, to the body as a coherent source of feelings and their underlying physiology. Similarly, cognitive activities in an organization require some embodiment to crystallize, direct and stabilize cognition and communication within and outside the organization. To perform its functions, organizational form has a number of possible features, corresponding to different ways in which organizational focus can work. For both the internal function of adaptation, with expression, crystallization, stabilization and direction, and the external function of selection by signalling, we find symbols, such as logos, ‘mission statements’, advertisement and external reporting. For the internal function in particular we find the exemplary behaviour of organizational heroes, often a founder of the organization, corresponding myths and rituals. Culture, with its signs, heroes, myths and rituals, aims to represent and engender a certain style of behaviour (Simmel, 1950, p. 341) whereby the individual becomes part of a collective intentionality. More formalized forms of organization are procedures, for reporting, decision making, recruitment, contracting and the like. An important more formal organizational form is legal identity, aimed at securing the interests of different stakeholders. Legal identity varies with the focal stakeholders and their interests, and is needed to regulate ownership and decision rights, liability, contracting and the like. Here, firms distinguish themselves from organizations more generally. A firm is defined as an organization of capital and labour aimed at profit, in contrast with, for example, a foundation that is not aimed at profit, and where profits are re-absorbed in the organization. The legal identity of firms varies according to the regulation of liability, ownership, availability of shares, employment status, tax and the like. Here, there is still a connection with TCE, in that under the authority of an employment relationship norms and forms of conduct can be imposed without having to engage in a contract for each individual transaction. As such, legal identity functions to formalize and consolidate organizational culture. Elements of this idea of organization are not new. It connects with the idea, in the organization literature, that the crux of an organization is to serve as a ‘sensemaking system’ (Weick, 1979, 1995), a ‘system of shared

356

Handbook of knowledge and economics

meaning’ (Smircich, 1983) or ‘interpretation system’ (Choo, 1998). I propose that this yields a more fundamental reason for firms to exist than the reduction of transaction costs, although transaction costs are also part of the story (Nooteboom, 2000). In a firm, people need to achieve a common purpose, and for this they need some more or less tacit shared ways of seeing and interpreting the world and regulating collaboration. 15.3.3

Boundaries of the Firm

A theory of the firm, or of organizations more widely, should account for the boundaries of the organization and for inter-organizational relationships (IORs). Effects of firm size have formed a central subject in economics. On the competence side are effects of static, productive efficiency, in scale and scope, which can take a variety of forms, including division of labour, and effects of dynamic efficiency in Schumpeterian debates on whether large or small firms are the more innovative. A discussion of the literatures on those subjects lies beyond the scope of the present chapter. The point to be made here is that debates on effects of firm size have turned out to be somewhat misguided in that small firms may compensate for their weaknesses by collaboration in networks (or clusters, or industrial districts), while large firms may compensate for their weaknesses by a greater or lesser disintegration or decentralization of units within the firm. Here, clusters of firms are also forms of organization, but they are not firms, while large firms can operate more or less as clusters. The difference lies in the ‘cohesiveness’ of cognitive focus and in legal identity. Here I propose the general principle that the boundary of an organizational entity in general, and of a firm in particular, is determined by the cohesiveness of the focus, in combination with its legal formalization. Note that the notion of organizational focus does not entail the need for people to agree on everything, or see everything the same way. Indeed, such lack of diversity would preclude both division of labour and innovation within the firm. As discussed in Nooteboom (1999), there is a trade-off between cognitive distance, needed for variety and novelty of cognition, and cognitive proximity, needed for mutual understanding and agreement. In fact, different people in a firm will to a greater or lesser extent introduce elements of novelty from their outside lives and experience, and this is a source of both error and innovation (DiMaggio, 1997). Nevertheless, there are some things they have to agree on, and some views, often tacit, that they need to share, on goals of the organization, norms, values, standards, outputs, competencies and ways of doing things. The cohesiveness of organizational focus has two dimensions, at least, of inclusiveness, or scope, and tightness. If the life world of people

Embodied cognition, organization and innovation 357 has many dimensions, inclusiveness denotes the number of dimensions included in organizational focus. This is closely related to the point, made earlier, that cognitive distance entails difference in a variety of dimensions of cognition. Tightness denotes similarity, or proximity, in the dimensions involved. A large inclusiveness or scope of focus entails that there is alignment on many issues, and tightness entails that on each issue there is little ambiguity and variety of meaning, norms and standards. A highly inclusive scope entails that more of a person’s life world is included in the organization, in ‘thick’ relationships, carrying many aspects of the life world and of personality, and a less inclusive scope entails less personalized, ‘thinner’ relationships. The notion of the cohesiveness of focus connects with the distinction Simmel (1950) made between a person’s function in an organization, which takes up only part of his personality, and his full personality. This is echoed in the distinction that Ring and van de Ven (1994) made between roles that people play and behaviour ‘qua persona’. In a cohesive focus, role and persona get closer. With a highly cohesive focus, the liberty of people and variety among them are constrained. Extremes of this are found in cliques, and especially in clandestine, secluded or secret societies (Simmel, 1950, pp. 345–76). Outside freedom to engage in external relationships is constrained by the high inclusiveness of organizational focus, by which there are few dimensions of the life world left that are not in some way already regulated within the group. Inside freedom is constrained by the tightness of focus, with little room to deviate from narrow norms. Both inside and outside sources of variety, and hence of innovation, are highly constrained. While inclusiveness and tightness are separate dimensions of scope, high inclusiveness does tend to generate tightness, as follows. When high inclusiveness forms an obstacle to outside relationships, which occurs to the extent that organizational focus imposes meanings not shared outside, then people are cut off from sources of fresh or different ideas, and they will tend to gravitate towards meanings shared inside the organization, which increases tightness, not because it is imposed by focus, but because it emerges from decreasing cognitive distance. Thus large scope and tightness together tend to reinforce themselves. An implication of the notion of a focusing device is that the need to achieve a focus entails a risk of myopia: relevant threats and opportunities to the firm are not perceived. To compensate for this, people, and firms, need complementary sources of outside intelligence, to utilize ‘external economy of cognitive scope’ (Nooteboom, 1992). This yields a new perspective on inter-organizational relationships, next to the usual considerations, known from the alliance literature. It also fits well with the prevalent

358

Handbook of knowledge and economics

idea in the literature on innovation systems that innovation derives primarily from interaction between firms (Lundvall, 1988). The notion of a firm as a focusing device yields an alternative to TCE for an explanation of the boundaries of the firm. The present theory yields a prediction that is opposite to that of classical transaction cost economics, and which is particularly relevant in innovation. With increasing uncertainty, in terms of volatility of technology and markets, firms should not integrate activities more, as transaction cost theory predicts, but less, because the need to utilize outside complementary cognition is greater. The argument from TCE was that under uncertainty one needs the greater power of management by fiat within a firm, to monitor behaviour and resolve conflicts. Here, the counter-argument is that under the volatility of innovation the risk of organizational myopia is greater and hence there is a greater need for outside complementary cognition, with ‘external economy of cognitive scope’. The prediction of less rather than more integration under uncertainties of innovation has been confirmed empirically by Colombo and Garrone (1998), who found that in technologically volatile industries, as measured by patent intensity, the likelihood of alliances rather than mergers and acquisitions is higher than in the absence of such volatility. 15.3.4

Communities and Firms

A firm may consist of a single community of practice (COP), as the smallest unit of organization, typically with a relatively cohesive focus, or it may be a ‘community of communities’ (Amin and Cohendet, 2003), with a less cohesive focus. Recall that a focus can be cohesive in either or both of two ways: relationships in the community are thick, comprising many dimensions of personality, or they are tight, with little variety in the relevant dimensions, or both. In fact, the notion of COP allows for a great variety of different kinds of community (Bogenrieder and Nooteboom, 2004). One type, apparently closest to the original idea of a community of ‘practice’ (Brown and Duguid, 1991; Lave and Wenger, 1991) is a thick and tight community where people interact on many issues (highly inclusive), on a daily basis, with little ambiguity of meanings (tight), in the execution of a practice. Another type is that of a community of professionals from different contexts of action, who exchange knowledge, such as scholars at a conference, for example. Here, focus is narrow in scope but often tight, with people talking precisely about few things. One can have a group with wide focus and little tightness, with people talking vaguely about many things, such as practitioners from different practices talking about many aspects of their practice. Strangers typically talk vaguely about few

Embodied cognition, organization and innovation 359 things. A cohesive group, with small internal distance in many dimensions of cognition, is likely to be very efficient in a static sense, or in ‘exploitation’, but inefficient in a dynamic sense, or in ‘exploration’ (March, 1991; Nooteboom, 2002). To keep cohesive groups from cognitive inertia, it may be needed to rotate people across them, to keep up variety, that is, to maintain some cognitive distance. A cohesive group can increase variety on the basis of outside contacts, but must then relax its tightness, to allow for variety of meaning. The size of the smallest community depends on how ‘systemic’, as opposed to ‘stand-alone’ (Teece, 1986; Langlois and Robertson, 1995; Postrel, 2002), the structure of the activity is. Exploitation is systemic when there is a complex division of labour, with many elements and a dense structure of connections between them, with tight constraints on their interfaces. These connections yield interdependence, with different types, as recognized by Thompson (1967): sequential (output of one step is input for the next), reciprocal (inputs and outputs both ways) and pooled interdependence (common use of a shared resource). An example of high ‘systemicness’ is an oil refinery. In more stand-alone systems, elements of the system are connected with few other elements, and connections are loose, allowing for some ambiguity and deviation from standards on interfaces. This allows for separate communities with a high degree of autonomy. An example is a consultancy firm. A third type is a modular system. Here, there are also multiple, connected elements, as in the systemic case, but different elements embody different, separable functions, and standards on interfaces between them allow for variety, where different modules can be plugged into the system as alternative ways of performing specific functions. Then, modules may be separate communities. The small firm often has a limited range or portfolio of technologies, products and competencies. The firm typically coincides with a single COP, with this smaller unit having separate legal identity, while in large organizations the smaller communities do not have such separate legal identity. As a result, they are vulnerable, with limited diversification of risks, limited specialization in functions, limited economies of scale and scope, and limited career perspectives. They also have both the potential advantages and disadvantages of a cohesive scope. They often, though not always and not necessarily, have a cohesive focus, in relatively thick and tight, often highly personalized relationships. One cause of high inclusiveness, and perhaps the most important one, lies in limited specialization of labour, due to lack of scale and scope. As indicated above, high inclusiveness generates tightness if inside norms or meanings are incompatible with outside ones. This may result from radical novelty, where the firm is generating new, unfamiliar meanings. High tightness may also result from

360

Handbook of knowledge and economics

the cognitive stamp that the entrepreneur puts on his small organization, where he interacts directly with his personnel. In this way, radically innovative, small firms may isolate themselves and thereby close themselves off from the sources of application and further innovation. This yields one of several paradoxes of the small firm. On the one hand, small-size, personalized, thick, informal relationships, integration of tasks among few people and direct contacts, internally and outside, for example with customers, enable high flexibility and motivational power of identification with the firm. On the other hand there is potential for suppression of freedom and variety, and of isolation from the environment (cf. Nooteboom, 1994). The large firm typically has a wider range of activities, with more or less cohesive focus in internal communities with a narrower range of activities, and an overall organizational focus that has limited scope, with people having less in common across the organization, but possibly with a high degree of tightness in a few aspects of cognition, particularly concerning normative issues of overall purpose and style of interaction. Small firms may compensate for their weaknesses with collaboration in networks or industrial districts to spread risks, and to obtain economies of scale and scope, mimicking large firms in some of their features. Somewhat perversely, perhaps, for dynamic efficiency, the greatest benefit of industrial districts may be that there is flexibility of configuration from the fact that firms that do not fit in new constellations can more easily be dropped than can departments of firms. Large firms can obtain the benefits of smallness by mimicking small firms in COP, while maintaining benefits of integration in the form of economies of scale and scope, and diversification of risks. Their limited scope of overall organizational focus allows for a wide variety of competence, and for personal freedom. However, the tightness of focus needed for governing a wide range of activity, in combination with limited opportunities for shedding parts of the organization, due to their inclusion in an overall legal entity, may inhibit the flexibility of configuration and variety of purpose that is conducive to radical innovation. 15.3.5

Innovation by Interaction

Especially from an evolutionary perspective (Nelson and Winter, 1982), heterogeneity or variety is a crucial source of innovation, and this has been taken up in the alliance literature (Stuart and Podolny, 1996; Almeida and Kogut, 1999; Rosenkopf and Nerkar, 2001; Fleming, 2001; Rosenkopf and Almeida, 2003; Ahuja and Katila, 2004). However, that literature does not explain how, precisely, heterogeneity produces innovation. Furthermore,

Embodied cognition, organization and innovation 361 heterogeneity in networks has two dimensions that are seldom explicitly distinguished. One is the number of firms involved, and the pattern of ties between them, and the other is the difference, in particular cognitive distance, between them. Between firms, in contrast with people, cognitive distance is the difference between the cognitive foci of firms, with two main dimensions of technological knowledge/competence and moral principles for internal governance. A large stream of literature has indicated only the problems rather than also the benefits of such cognitive distance. In a study on alliance formation in the semi-conductor industry, Stuart (1998) argued that the most valuable alliances are those between firms with similar technological foci and/or operating in similar markets, whereas distant firms are inhibited from cooperating effectively. In a similar vein, the diversification literature argues that most is to be learned from alliance partners with related knowledge and skills (Tanriverdi and Venkatraman, 2005), or from areas that firms already possess capabilities in (Penner-Hahn and Shaver, 2005). In the literature on international business also, a pervasive view is that cognitive distance is a problem to be overcome. Johanson and Vahlne (1977, 1990) employed the notion of ‘psychological distance’, which is seen as having an adverse effect on cross-cultural communication. When learning is discussed, in that literature, it is mostly seen as learning to cope with transnational differences by accumulating experience in cross-border collaboration (e.g. Barkema et al., 1997), rather than taking those differences as a potential source of learning to change home-country products or practices. Nooteboom (1999) proposed an interaction between the advantages and disadvantages of distance, as follows. The ability to understand each other (in absorptive capacity) and to collaborate declines with cognitive distance, whereas the novelty value of the relationship, that is, its potential to generate Schumpeterian novel combinations, increases with distance. If the two effects are linear with respect to distance, and if learning or innovation performance of the relationship is proportional to the mathematical product of novelty value and mutual absorptive capacity, the result is an inverted-U-shaped performance as a function of distance, as illustrated in Figure 15.1. This implies an optimal cognitive distance, which is large enough for partners to offer each other something new, but not so large that they cannot understand each other or come to agreement. In Figure 15.1, the downward-sloping line of absorptive capacity is not fixed. It is subject to an upward shift, as a function of the accumulation of knowledge in relevant fields and experience in IORs. That yields a shift to higher optimal cognitive distance, as illustrated in Figure 15.1. Wuyts et al. (2005) put the hypothesis of optimal cognitive distance to two empirical, econometric tests. The first test was conducted on a

362

Handbook of knowledge and economics

Absorptive capacity

Novelty value

Learning

Optimal cognitive distance

Cognitive distance

Source: Nooteboom (1999).

Figure 15.1

Optimal cognitive distance

combination of the basic hypothesis of optimal cognitive distance with the second hypothesis that cognitive distance decreases with increased frequency and duration of interaction. As argued by Gulati (1995) and others (Simmel, 1950; McAllister, 1995; Lewicki and Bunker, 1996), familiarity may breed trust, which is good for governance. However, it may also reduce variety of knowledge, which is bad for innovative performance. This yields the hypothesis of an inverted-U-shaped relation between radical technological innovation and the extent to which firms ally with the same partners over time. That hypothesis was tested on data on vertical alliances between biotech and pharma companies, and was supported. In fact, the derived hypothesis is subject to nuance. If two partners have access to other, non-overlapping partners, so that they are continually being refreshed with new, non-overlapping knowledge, cognitive distance between them is maintained, so that the relationship may remain innovative even when it lasts long. This is, in fact, the point, or part of the point, of Burt’s (1992) notion of bridging structural holes. The second test by Wuyts et al. was conducted on a combination of the basic hypothesis of optimal cognitive distance with a second hypothesis that the likelihood of a collaborative alliance increases with the expected

Embodied cognition, organization and innovation 363 performance of collaborative innovation. This yielded the derived hypothesis that the likelihood of an alliance for innovation has an inverted-Ushaped relation with cognitive distance. That hypothesis was tested on data on horizontal alliances in ICT industries. Cognitive distance was measured by differences in degrees of specialization in different dimensions of technology, inferred from patent data. Partial support was found. Technology-related measures of cognitive distance were not found to have any significant effect, but several indicators of differences in firms’ organizational characteristics proved to have the expected inverted-U-shaped effect. Several considerations were offered to explain why organizational aspects turned out to be more important than technological ones in ICT industries. Nooteboom et al. (2005) conducted a more complete empirical, econometric test, on the basis of a large set of data on inter-firm alliances over a ten-year period, in a variety of industries. Cognitive distance was reduced to technological distance, which was measured on the basis of correlation between profiles of technological knowledge composed from patent data. Innovative performance was measured as new patents, in successive years, with a distinction between exploratory patents, in new patent classes, and exploitative patents, in patent classes in which a firm already has patents. Absorptive capacity was made endogenous, in that the downward-sloping line of absorptive capacity (cf. Figure 15.1) was taken as a function of cumulative past R&D. The hypothesis of performance as an inverse-Ushaped function of cognitive distance was confirmed, including the further hypothesis that optimal distance is higher for exploration than for exploitation. The latter can be attributed to a higher slope of the novelty line in Figure 15.1. The study also tested for an effect of cumulative experience in alliances on absorptive capacity, but found none. It did yield an additional effect that was not hypothesized. The results indicated that cumulative experience in R&D not only raises the level of absorptive capacity (in an upward shift of the corresponding line; see Figure 15.1), but also that the upward slope of the line denoting novelty value decreased. This implies a principle of decreasing returns to knowledge, or a ‘boredom effect’. The more knowledge one accumulates, the further afield one has to go, to more exotic sources or partners, to still learn something new.

15.4 CONCLUSION This chapter explores the implications of a theory of embodied cognition for the fields of economics and management in general and for the

364

Handbook of knowledge and economics

theory of the firm and theory of innovation in particular. In general, embodied cognition yields a perspective for transcending the methodological individualism of economics and the methodological collectivism of sociology, in a principle of methodological interactionism, which may enable an integration of economics and sociology, in a new behavioural science. Embodied cognition yields the notion of cognitive distance between people, which poses a problem of organization. In order to achieve a shared purpose, cognition must be aligned to some extent. This yields the notion of organization as a focusing device, in a reduction of cognitive distance between people. The boundary of an organization in general and a firm in particular is determined by the cohesiveness of organizational focus, that is, the scope of cognition involved, and the tightness of cognitive alignment, together with the legal form in which focus is embodied. Organizations vary in size, according to their range of activities and the cohesiveness of their scope. Organizational focus yields myopia, which generally needs to be compensated in relationships with outside organizations, at a greater cognitive distance than within the organizations. Here, between organizations, cognitive distance is defined as differences in organizational focus. For learning by interaction, there is a trade-off between negative effects of cognitive distance, in lack of mutual understanding and ability to collaborate, and positive effects, in yielding cognitive variety as a source of Schumpeterian novel combinations. This yields the notion of optimal cognitive distance in interaction for innovation. Many extensions and related issues could not be discussed in this chapter. The analysis of innovation by collaboration between firms extends into the analysis of network effects of structure and strength of ties (Gilsing and Nooteboom, 2005). There is much more to be said about instruments of governance within and between firms. The analysis far from exhausts the potential of methodological interactionism. Embodied cognition yields the basis for a theory of invention as a function of shifts in context (Nooteboom, 2000). It also yields an opening for the integration of insights from social psychology (Nooteboom, 2002; Six, 2004).

NOTES 1. Much of this first section is taken, in an abbreviated form, from Nooteboom (2000, ch.6). 2. Source: ‘Cognition: From Molecules to Mind’, conference at the Royal Dutch Academy of Arts and Sciences, Amsterdam, 29 March 2006.

Embodied cognition, organization and innovation 365

REFERENCES Ahuja, G. and Katila, R. (2004), ‘Where do resources come from? The role of idiosyncratic situations’, Strategic Management Journal, 25 (8–9), 887–907. Almeida, P. and Kogut, B. (1999), ‘Localization of knowledge and the mobility of engineers in regional networks’, Management Science, 45 (7), 905–17. Amin, A. and Cohendet, P. (2003), Architectures of Knowledge: Firms, Capabilities and Communities, Oxford: Oxford University Press. Bachelard, G. (1980), La formation de l’esprit scientifique, 11th edn, Paris: J. Vrin. Barkema, H.G., Shenkar, O., Vermeulen, F. and Bell, J.H.J. (1997), ‘Working abroad, working with others: how firms learn to operate international joint ventures’, Academy ofManagement Journal, 40, 426–42. Barkow, J., Cosmides, L. and Tooby, J. (1992), The Adapted Mind: Evolutionary Psychology and the Generation of Culture, Oxford: Oxford University Press. Berger, P. and Luckmann, T. (1967), The Social Construction of Reality, New York: Doubleday. Blackler, F. (1995), ‘Knowledge, knowledge work and organizations: an overview and interpretation’, Organization Studies, 16 (6), 1021–46. Bogenrieder, I. and Nooteboom, B. (2004), ‘Learning groups: what types are there?’, Organization Studies, 25 (2), 287–314. Brown, J.S. and Duguid, P. (1991), ‘Organizational learning and communities of practice: toward a unified view of working, learning, and innovation’, Organization Science, 2 (1), 40–57. Brown, J.S. and Duguid, P. (1996), ‘Organizational learning and communities of practice’, in M.D. Cohen and L.S. Sproull (eds), Organizational Learning, London: Sage, pp. 58–82. Burt, R. (1992), ‘Structural holes: the social structure of competition’, in N. Nohria and R. Eccles (eds), Networks and Organizations: Structure, Form and Action, Boston, MA: Harvard Business School Press, pp. 57–91. Chesbrough, H.W. (2003), Open Innovation: The New Imperative for Creating and Profiting from Technology, Boston, MA: Harvard Business School Press. Choo, C.W. (1998), The Knowing Organization, Oxford: Oxford University Press. Cohen, M.D. and Levinthal, D.A. (1990), ‘Absorptive capacity: a new perspective on learning innovation’, Administrative Science Quarterly, 35, 128–52. Cohen, M.D. and Sproull, L.S. (eds) (1996), Organizational Learning, London: Sage, pp. 58–82. Colombo, M.G. and Garrone, P. (1998), ‘Common carriers entry into multimedia services’, Information Economics and Policy, 10, 77–105. Damasio, A.R. (1995), Descartes’ Error: Emotion, Reason and the Human Brain, London: Picador. Damasio, A.R. (2003), Looking for Spinoza, Orlando, FL: Harcourt. Dellarosa, D. (1988), ‘The psychological appeal of connectionism’, Behavioral and Brain Sciences, 11 (1), 28–9. DiMaggio, P. (1997), ‘Culture and cognition’, Annual Review of Sociology, 23, 263–87. Edelman, G.M. (1987), Neural Darwinism: The Theory of Neuronal Group Selection, New York: Basic Books. Edelman, G.M. (1992), Bright Air, Brilliant Fire: On the Matter of Mind, London: Penguin. Eijck, J. van, and Kamp, H. (1997), ‘Representing discourse in context’, in J. van Benthem and A. ter Meulen, Handbook of Logic and Language, Amsterdam: Elsevier, pp. 179–237. Fleming, L. (2001), ‘Recombinant uncertainty in technological search’, Management Science, 47 (1), 117–32. Fodor, J.A. (1975), The Language of Thought, New York: Thomas Y. Crowell. Frege, G. (1892), ‘On sense and reference’ (in German), Zeitschrift fűr Philosophie und philosophische Kritik, 100, 25–50. Geach, P. and Black, M. (1977), Philosophical Writings of Gottlob Frege, Oxford: Blackwell. Gilsing, V.A. and Nooteboom, B. (2005), ‘Density and strength of ties in innovation

366

Handbook of knowledge and economics

networks: an analysis of multi-media and biotechnology’, European Management Review, 2, 179–97. Gulati, R. (1995), ‘Familiarity breeds trust? The implications of repeated ties on contractual choice in alliances’, Academy of Management Journal, 38, 85–112. Habermas, J. (1982), Theorie des kommunikativen Handelns, Teil 1 und 2 (Theory of communicative action), Frankfurt: Suhrkamp. Habermas, J. (1984), Vorstudien und Ergänzungen zur Theorie des kommunikatieven Handelns (Preliminary studies and elaborations to the theory of communicative action), Frankfurt: Suhrkamp. Hagerty, M.R. (1999), ‘Testing Maslow’s hierarchy of needs: national quality of life across time’, Social Indicators Research, 46 (3), 249–71. Hedberg, B.L.T. (1981), ‘How organizations learn and unlearn’, in P.C. Nystrom and W.H. Starbuck (eds), Handbook of Organizational Design, New York: Oxford University Press, pp. 3–27. Hendriks-Jansen, H. (1996), Catching Ourselves in the Act: Situated Activity, Interactive Emergence, Evolution and Human Thought, Cambridge, MA: MIT Press. Janssen, Th.M.V. (1997), ‘Compositionality’, in J. van Benthem and A. ter Meulen (eds), Handbook of Logic and Language, Amsterdam: Elsevier, pp. 417–73. Johanson, J. and Vahlne, J. (1977), ‘The internationalization process of the firm – a model of knowledge development and increasing foreign market commitment’, Journal of International Business Studies (spring/summer), 23–32. Johanson, J. and Vahlne, J. (1990), ‘The mechanism of internationalization’, International Marketing Review, 7 (4), 11–24. Johnson-Laird, P.N. (1983), Mental Models, Cambridge: Cambridge University Press. Jorna, R. (1990), Knowledge Representations and Symbols in the Mind, Tübingen: Stauffenburg Verlag. Kogut, B. and Zander, U. (1992), ‘Knowledge of the firm, combinative capabilities, and the replication of technology’, Organization Science, 3, 383–97. Kolb, D. (1984), Experiential Learning: Experience as the Source of Learning and Development, Englewood Cliffs, NJ: Prentice-Hall. Lakatos, I. (1970), ‘Falsification and the methodology of scientific research programmes’, in I. Lakatos and A. Musgrave (eds), Criticism and the Growth of Knowledge, Cambridge: Cambridge University Press, pp. 91–196. Lakatos, I. (1978), The Methodology of Scientific Research Programmes; Philosophical Papers, Vols 1 and 2, ed. J. Worrall and G. Currie, Cambridge: Cambridge University Press. Lakoff, G. and Johnson, M. (1980), Metaphors we Live by, Chicago, IL: University of Chicago Press. Lakoff, G. and Johnson, M. (1999), Philosophy in the Flesh, New York: Basic Books. Langlois, R.N. and Robertson, P.L. (1995), Firms, Markets and Economic Change, London: Routledge. Lave, J. and Wenger, E. (1991), Situated Learning: Legitimate Peripheral Participation, Cambridge: Cambridge University Press. Lewicki, R.J. and Bunker, B.B. (1996), ‘Developing and maintaining trust in work relationships’, in R.M. Kramer and T.R. Tyler (eds), Trust in Organizations: Frontiers of Theory Research, Thousand Oaks, CA: Sage Publications, pp. 114–39. Lundvall, B.-Å. (1988), ‘Innovation as an interactive process: from user–producer interaction to national systems of innovation’, in G. Dosi, C. Freeman and R. Nelson (eds), Technical Change and Economic Theory, London: Pinter, pp. 349–69. March, J.G (1991), ‘Exploration and exploitation in organizational learning’, Organization Science, 2 (1), 101–23. Maslow, A. (1954), Motivation and Personality, New York: Harper. McAllister, D.J. (1995), ‘Affect- and cognition-based trust as foundations for interpersonal cooperation in organizations’, Academy of Management Journal, 38 (1), 24–59. McClelland, J.L., Rumelhart, D.E. and Hinton, G.E. (1987), ‘The appeal of parallel distributed processing’, in D.E. Rumelhart, J.L. McClelland and the PDP research group (eds),

Embodied cognition, organization and innovation 367 Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations, Cambridge, MA: MIT Press, pp. 3–44. Mead, G.H. (1934), Mind, Self and Society: From the Standpoint of a Social Behaviorist, Chicago, IL: University of Chicago Press. Meindl, J.R., Stubbart, C. and Porac, J.F. (eds) (1998), Cognition Within and Between Organisations, London: Sage. Merleau-Ponty, M. (1942), La structure du comportement, Paris: Presses universitaires de France. Merleau-Ponty, M. (1964), Le visible et l’invisible, Paris: Gallimard. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge: Cambridge University Press. Nooteboom, B. (1992), ‘Towards a dynamic theory of transactions’, Journal of Evolutionary Economics, 2, 281–99. Nooteboom, B. (1994), ‘Innovation and diffusion in small business: theory and empirical evidence’, Small Business Economics, 6, 327–47. Reprinted in N. Krueger (ed.), Entrepreneurship: Critical Perspectives on Business and Management, III, London: Routledge, 2002, pp. 327–47. Nooteboom, B. (1999), Inter-firm Alliances: Analysis and Design, London: Routledge. Nooteboom, B. (2000), Learning and Innovation in Organizations and Economies, Oxford: Oxford University Press. Nooteboom, B. (2002), Trust: Forms, Foundations, Functions, Failures and Figures, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Nooteboom, B. (2004), ‘Governance and competence, how can they be combined?’, Cambridge Journal of Economics, 28 (4), 505–26. Nooteboom, B., Van Haverbeke, W., Duijsters, G., Gilsing, V. and Oord, A. van der (2005), ‘Optimal cognitive distance and absorptive capacity’, Research Policy, 36, 1016–34. Nussbaum, M.C. (2001), Upheavals of Thought: The Intelligence of Emotions, Cambridge: Cambridge University Press. Penner-Hahn, J. and Myles Shaver, J. (2005), ‘Does international research and development increase patent output? An analysis of Japanese pharmaceutical firms’, Strategic Management Journal, 26, 121–40. Penrose, E. (1959), The Theory of the Growth of the Firm, New York: Wiley. Polanyi, M. (1962), Personal Knowledge, London: Routledge. Postrel, S. (2002), ‘Islands of shared knowledge: specialization and mutual understanding in problem-solving teams’, Organization Science, 13 (3), 303–20. Putnam, H. (1957), Mind, Language and Reality: Philosophical Papers, 2, Cambridge: Cambridge University Press. Ring, P. and van de Ven, A. (1994), ‘Developmental processes of cooperative interorganizational relationships’, Academy of Management Review, 19 (1), 90–118. Rose, S. (1992), The Making of Memory, New York: Doubleday. Rosenkopf, L. and Nerkar, A. (2001), ‘Beyond local search: boundary-spanning, exploration, and impact in the optical disc industry’, Strategic Management Journal, 22, 287–306. Rosenkopf, L. and Almeida, P. (2003), ‘Overcoming local search through alliances and mobility’, Management Science, 49, 751–66. Rumelhart, D.E., McClelland, J.L. and the PDP research group (1987), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations, Cambridge MA: MIT Press. Rumelhart, D.E., Hinton, G.E. and McClelland, J.L. (1987), ‘A general framework for PDP’, in D.E. Rumelhart, J.L. McClelland and the PDP research group (eds), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations, Cambridge MA: MIT Press, pp. 45–76. Schein, E.H. (1985), Organizational Culture and Leadership, San Francisco, CA: Jossey-Bass. Searle, J.R. (1992), The Rediscovery of the Mind, Cambridge, MA: MIT Press. Shanon, B. (1988), ‘Semantic representation of meaning: a critique’, Psychological Bulletin, 104 (1), 70–83.

368

Handbook of knowledge and economics

Shanon, B. (1993), The Representational and the Presentational, New York: HarvesterWheatsheaf. Simmel, G. (1950), The Sociology of Georg Simmel, trans. Kurt Wolff, Glencoe, IL: The Free Press. Simon, H.A. (1983), Reason in Human Affairs, Oxford: Basil Blackwell. Six, F. (2004), Trust and Trouble; Building Interpersonal Trust within Organizations, PhD dissertation, Erasmus University Rotterdam. Smircich, L. (1983), ‘Organization as shared meaning’, in L.R. Pondy, P.J. Frost, G. Morgan and T.C. Dandridge (eds), Organizational Symbolism, Greenwich, CT: JAI Press, pp. 55–65. Smolensky, P. (1988), ‘On the proper treatment of connectionism’, Behavioral and Brain Sciences, 11, 1–74. Stuart, T. (1998), ‘Network positions and propensities to collaborate: an investigation of strategic alliance formation in a high-technology industry’, Administrative Science Quarterly, 43, 637–68. Stuart, T. and Podolny, J. (1996), ‘Local search and the evolution of technological capabilities’, Strategic Management Journal, 17 (Special Issue), 21–38. Tanriverdi, H. and Venkatraman, N. (2005), ‘Knowledge relatedness and the performance of multibusiness firms’, Strategic Management Journal, 26, 97–119. Teece, D.J. (1986), ‘Profiting from technological innovation: implications for integration, collaboration, licensing and public policy’, Research Policy, 15, 285–305. Thiel, C. (1965), Sinn und Bedeutung in der Logik Gottlob Freges, Meisenheim am Glan: Anton Hain. Thompson, J.D. (1967), Organizations in Action, New York: McGraw-Hill. Vygotsky, L. (1962), Thought and Language, trans. E. Hanfmann and G. Varkar (eds), Cambridge, MA: MIT Press. Weick, K.F. (1979), The Social Psychology of Organizing, Reading, MA: Addison-Wesley. Weick, K.F. (1995), Sensemaking in Organizations, Thousand Oaks, CA: Sage. Weintraub, E.R. (1988), ‘The neo-Walrasian program is progressive’, in N. de Marchi (ed.), The Popperian Legacy in Economics, Cambridge: Cambridge University Press, pp. 213–30. Wenger, E. and Snyder, W.M. (2000), ‘Communities of practice: the organizational frontier’, Harvard Business Review (Jan.–Feb.), 139–45. Werner, H. and Kaplan, B. (1963), Symbol Formation, New York: Wiley. Winograd, T. (1980), ‘What does it mean to understand language?’, Cognitive Science, 4, 209–41. Wittgenstein, L. (1976), Philosophical Investigations (first published in 1953), Oxford: Basil Blackwell. Wuyts, S., Colombo, M.G., Dutta, S. and Nooteboom, B. (2005), ‘Empirical tests of optimal cognitive distance’, Journal of Economic Behavior and Organization, 58 (2), 277–302.

16 Knowledge and its economic characteristics: a conceptual clarification Ulrich Witt, Tom Broekel and Thomas Brenner

16.1 INTRODUCTION It is broadly acknowledged now that knowledge plays a crucial role in understanding technological change and its impact on economic activities (cf., e.g., Feldman, 1994; Nonaka et al., 2000; Antonelli, 2001; Mokyr, 2002; Murmann, 2003). With a growing body of research in economics focusing on the role of knowledge, a growing number of concepts and classifications of knowledge has been developed since the first elaborate attempt in that direction by Machlup (1980). One distinction that, among others, will serve as a reference here goes back to Polanyi (1958). It is the widely used distinction between tacit and overt, or implicit and explicit, knowledge (see Cowan and Foray, 1997; Cowan et al., 2000 and the literature cited there). While knowledge is considered overt when it is openly accessible to everybody and understandable, for tacit knowledge this is considered not to be the case (Cowan and Foray, 1997; Zack, 1999; Cowan et al., 2000; Lissoni, 2001). A distinction like this is made with the intention to bring out characteristics of knowledge that are regarded as significant. Yet, what the significant characteristics are depends on the context. Distinguishing tacit from overt knowledge does not seem of much help, for instance, in deciding whether or not knowledge has a public-good character – a question that figures prominently in the economic literature (e.g. in welfare theory; see Arrow, 1962, or new growth theory; see Langlois, 2001). Some authors claim that, once knowledge exists, it has properties of a public good (Nelson, 1959; Arrow, 1962; Foray, 2004; Brusoni et al., 2005) or at least of a local public good. Others consider knowledge a private good (e.g. Callon and Bowker, 1994). It seems worthwhile, therefore, to clarify how different characteristics of knowledge relate to each other and which ones matter when. This chapter is devoted to exploring this question. We shall argue that the different characteristics of knowledge – its overtness or tacitness and its public- or private-good property – depend equally on the state of the knowledge technology, that is, on how knowledge can be acquired, stored, used and communicated (cf. Nelson and Nelson, 2002). Different 369

370

Handbook of knowledge and economics

specifications of the characteristics therefore correspond to different assumptions about the knowledge technology, assumptions that are not always made explicit. To make our point, we shall start with a brief outline of the technological underpinnings of knowledge storage, expression and transfer in Section 16.2. The insights gained at that level suggest modifications regarding the distinction between overt and tacit knowledge on the one hand and encoded and non-encoded knowledge on the other that are discussed in Section 16.3. Section 16.4 then turns to the public- versus private-good characteristics of knowledge and derives the implications of our interpretation. Section 16.5 presents the conclusions.

16.2 KNOWLEDGE STORAGE, EXPRESSION AND TRANSFER Knowledge is accumulated over time at the level of each human being, at the level of groups of individuals as in a firm or organization, and even at the level of the entire economy. The accumulated body of knowledge can be stored in a variety of ways, from individuals’ memory to diverse technical media, and it can be embodied in artefacts. As long as in human history there was no possibility to encode knowledge, knowledge could only be stored internally, that is, in individuals’ memory. This kind of storage ‘technology’ humans are naturally endowed with has a limited capacity, accessability and reliability, and suffers, moreover, from the constrained lifetime of the individual agents. Once ways had been found to encode knowledge on suitable artificial media, this meant that an external storage technology had become available that was not subject to the same constraints as human memory. On the other hand, with the physical separation of the storage of knowledge in artificial media and its use, which still requires the involvement of individual human minds, it becomes evident that stored knowledge as such does not produce any effect. Knowledge becomes effective – and economically relevant – only when it is accessed and processed by a human mind and eventually ‘expressed’ by some action. The processing involves more or less idiosyncratic, subjective interpretations. For this reason, knowledge, even when stored as a given body on artificial media, can find expression in different thoughts and actions by different agents. To enable the agents to activate knowledge stored in codified form external to the individual brain and express it by their actions, two preconditions must be met. The agents need to have access to the stored knowledge, and they must have a cognitive ‘absorptive capacity’1 that allows them to understand the code and interpret context and meaning of what is stored. (Note that the encoding and storage of

Knowledge and its economic characteristics 371 knowledge or its embodiment in artefacts is also a form of expressing knowledge.) Access to externally stored knowledge is already an aspect of knowledge communication in which there are transmitters and recipients. It is useful to distinguish here between direct and indirect communication and between whether knowledge is intentionally or non-intentionally communicated. Knowledge is communicated directly only in oral or visual transmissions requiring face-to-face contact between transmitter and recipient – the communication technology that humans are naturally endowed with. It determined the constraints on knowledge transfer for the part of human history during which the encoding of knowledge was unknown. An indirect communication of knowledge is mediated by encoding the knowledge and transmitting it via, or storing it on, an artificial medium that can (later) be accessed by the recipient(s).2 Indirect knowledge transmission relies on optical, acoustic or electronic signals. Examples of communication by means of intermediate knowledge storage are written documents and visual and acoustic displays. Indirect communication thus presupposes codification of knowledge and an external storage technology. Knowledge is intentionally communicated when a sender deliberately transmits knowledge to some recipient(s), be it directly or indirectly. Knowledge is communicated unintentionally in two ways: ●

●

in direct communication when it is implicitly expressed by the action of an agent that, if observed, allows the observer to make an inference from the action to the underlying knowledge; in indirect communication when agents gain access to knowledge encoded on, and/or transmitted by, an artificial medium whose access has not been intended by the transmitter and/or the legal owner of the storage medium. Unintentional knowledge transmission is often involved when action or artefacts are imitated.

Indirect communication making use of technical media enables a more powerful knowledge transfer than direct communication. The reason is that several individuals can have a multiple, often even parallel, access to the storage medium. Moreover, the way in which knowledge can be communicated becomes independent of the physical presence of particular agents who hold that knowledge in their memory. Furthermore, in a technical storage medium, knowledge can be structured and retrieved in a much more systematic way and by the effort of several people.3 However, since indirect communication hinges on knowledge codification, its power is constrained by the state of, and the historical progress in, the encoding technology. For this reason, accelerations in the rate of knowledge

372

Handbook of knowledge and economics

communication and accumulation are correlated with major technical breakthroughs in codification – from the invention of writing to that of the printing press to the development of electronic encoding and automatized processing of the code (Dudley, 1999). However, independent of the great leaps forward in the encoding technology and the enormous growth of the body of encoded knowledge, in order for knowledge to become effective and economically relevant it still needs to be expressed by individual minds. This means that, because of their limited lifetime, each human generation has to acquire the growing knowledge anew in an incessant intergenerational knowledge transfer.

16.3 THREE TYPES OF KNOWLEDGE There seems to be a consensus that overtness of knowledge means that knowledge is, in principle, freely accessible by everyone and – if encoded – has a commonly understood code. (Actual accessibility depends, of course, on what proprietary regime can be enforced; see the next section.) In an interpretation in which overt and tacit knowledge are polar cases, the degree of overtness and, conversely, of tacitness can vary between the extremes (cf., e.g., Saviotti, 1998). As Nelson and Winter (1982, p. 78) note, ‘tacitness . . . is a matter of degree [and] . . . the same knowledge, apparently, is more tacit for some people than for others’. However, in contrast to the notion of overt knowledge for which the characteristics of accessibility and intelligibility are decisive, it is not clear whether tacitness of knowledge means just the opposite with respect to these two characteristics or whether it refers to some other characteristic(s). Indeed, it seems rather controversial what precisely tacitness of knowledge means.4 In the original statement in Polanyi (1958), tacitness refers to knowledge that is ill defined, cannot be articulated, and may not even be fully recognized. A frequently given example of tacit knowledge is the knowing how to keep balance in riding a bicycle. As everyone who has tried to teach someone else riding a bicycle is aware of, this know-how cannot be described verbally. It has to be acquired through trial-end-error learning by each individual anew. (Some hints that may ease the learning process may, of course, be helpful, as may be the observation of how others bike.) As Polanyi (1966, p. 4) put it, ‘we know more than we can tell’. This means that even though tacit knowledge can be expressed by an action (biking), it defies a verbal articulation (cf. Winter, 1987). As a consequence, it cannot be encoded, that is, given a (complete) description of how to do it. Hence it is not in the first place a matter of accessibility and intelligibility that makes tacit knowledge differ from overt knowledge. Rather it is the fact

Knowledge and its economic characteristics 373 that tacit knowledge does not lend itself to encoding and, for that reason, is not (easily) accessible and intelligible. Thus, in our understanding of tacit knowledge, what is causal for the difficulties in accessibility and intelligibility that impede the transfer of tacit knowledge is a lacking encoding which, in turn, follows from a lacking articulability. With a somewhat different emphasis, Brusoni et al. (2005) suggest an interpretation of tacit knowledge as ‘the inarticulable contextual framework(s) that provides individuals’ cognitive processes with the background within which to focus and to attribute meaning to contingent statements’. Encaoua et al. (2000, p. 193) argue that tacit knowledge is characterized by ‘a higher degree of uncertainty and the precise meaning is more interpretative and is not easily conveyed in a standardized medium’. Obviously, some additional characteristics are introduced in these interpretations while the precise relationship between tacitness and the encoding problem remains unclear. The characteristics Encaoua et al. (2000) attribute to tacit knowledge can also be satisfied by knowledge that is for some reason or other not (yet) encoded. However, knowledge that is not encoded does not necessarily have to be tacit. There is a difference between non-codified knowledge that can, in principle, be encoded and knowledge that cannot (Cowan et al., 2000; Balconi, 2002). Knowledge that cannot be articulated verbally surely is ‘uncodifiable knowledge’ (Grimaldi and Torrisi, 2001). People may not even be aware of it (Nelson, 1982). This indeed comes close to Polanyi’s original notion of tacit knowledge as ‘distinct from, but complementary to, the knowledge explicit in conscious cognitive processes’ (Cowan et al., 2000, p. 211). Thus, while tacitness of knowledge implies that this knowledge is not encoded, the converse is not true. Knowledge that, at some point in time, for some reason is not codified can be codified at a later point in time unless it is inherently non-codifiable. Only in the latter case it is tacit knowledge. As mentioned, the reason for the non-codifiability of tacit knowledge is that it is not articulable. However, even this characteristic may in some cases disappear over time. As a result of discoveries and systematic scientific research, previously tacit knowledge may become articulable, as Lazaric et al. (2003) have found in their case study of the French steel industry. Moreover, progress in production organization and increasing automatization often makes tacit knowledge obsolete (Balconi, 2002). However, this is not to claim that tacit knowledge is generally of declining importance. There are many cases where tacit knowledge can be expected to become neither articulable nor obsolete – particularly where sensory capacities and intuition matter (as in many cognitive tasks) or where subconscious control of motor skills is an important part of tacit knowledge (as in biking and many other cases of learning-by-doing).

374

Handbook of knowledge and economics

Hence we arrive at three distinct types of knowledge. The first is codified knowledge. The second is non-codified but, in principle, codifiable knowledge. The third is knowledge that is inherently non-codifiable (at least at the given time) and therefore tacit. In practice, the three types of knowledge may be activated simultaneously or sequentially. Imagine, for example, an agent who needs to work with some new laboratory equipment. A large part of knowledge on how to use the equipment can be obtained from the manual (the codified knowledge part). Additional non-codified knowledge can be obtained by asking experts who have already collected experiences with the same equipment. The user knowledge they communicate can be more specific than the manual. (This additional user knowledge could, in principle, be encoded, if there were incentives to do so.) Finally, by working with the equipment on the job, the agent may acquire an own-user knowhow, for example with respect to handling the equipment. It allows the agent to run down a learning curve in operating the equipment. The latter kind of know-how may in part be unconscious. As in the balance-on-thebicycle example, such know-how or parts of it cannot be articulated at all. It is the tacit, non-codifiable part of the agent’s knowledge. The differences in articulation and codification determine the specific storage and transfer conditions of the three types of knowledge. Concerning the storage aspect, unlike in the case of codified knowledge, all non-codified knowledge, tacit or not, can only be stored in human memory. When it is sometimes argued that non-codified knowledge is stored in organizations and institutions (Cowan and Foray, 1997) or in organizational routines (Nelson and Winter, 1982; Winter, 1987), this is therefore only a short-hand way of saying that people involved in organizational routines and interactions within an organization such as a firm hold in their individual memory a shared knowledge of how to interact. Concerning the knowledge transfer aspect, unlike in the case of codified knowledge, all non-codified knowledge, except tacit knowledge, can only be transmitted directly in a face-to-face interaction. Tacit knowledge cannot be interpersonally transmitted, not even through observational learning. Seeing someone expressing her knowledge of how to keep balance while riding a bicycle is not sufficient to acquire that know-how. This has to be done through one’s own trial-and-error learning.

16.4 WHEN IS KNOWLEDGE INDEED A PUBLIC GOOD? A question economists have been concerned with for a long time is whether knowledge is a public or a private good. As is well known, a pure

Knowledge and its economic characteristics 375 public good is characterized by two criteria. There is no rivalry in its use – the fact that the good is used by some agent does not affect the utility of other agents who also use the good – and there is no possibility of excluding anyone who wishes to use the good (no private appropriability). For a purely private good the opposite is true. The two criteria are independent of each other, and there are many goods that have only one of the two properties: non-rivalry in use or no exclusion of potential users (so-called impure public goods; cf. Sandler, 1975). It is often argued that knowledge has the properties of a public good (Nelson, 1959; Arrow, 1962; Teece, 1986; Callon and Bowker, 1994; Beise and Stahl, 1999; Roberts, 2001; Foray, 2004, ch. 6). This is not the place to appraise all facets of that debate. The aim rather is to reappraise the role of the technical terms by which knowledge can be stored, accessed and transferred for assessing the public- versus private-good nature of the different types of knowledge. The role seems quite obvious as far as the rivalry-in-use criterion is concerned. A significant feature of knowledge encoded on an artificial medium is that it can be reused many times and by all agents with the corresponding absorptive capacity without being instantaneously erased or degraded (Langlois, 2001). The more often the technical features of a storage medium allow the stored knowledge to be re-read, the less physical rivalry in use, that is, in getting hold of stored knowledge, is to be expected. Take, for example, the knowledge encoded in a book or on an electronic device such as a hard disc. As the knowledge stored does not disappear by being read, re-reading of the content of such storage media is possible many times. Their reuse may be subject to some tear and wear, so that their physical life time is limited. On the other hand, the same content can usually be copied on to an arbitrary number of additional carriers (books, CDs) at minor cost.5 However, the absence of physical rivalry in accessing encoded knowledge is not sufficient to satisfy the public-good property. As mentioned, it is not the stored knowledge as such that is economically relevant, but the expression of that knowledge in individual thought and action. The utility or the economic value derived from ideas and/or actions thus generated may, of course, decline the more frequently they are expressed. Imagine, for instance, that the knowledge about how to produce a certain chemical is described (encoded) in a scientific book. After reading the book the know-how can be expressed by, and may become commercially valuable to, every agent who has the suitable absorptive capacity and who engages in a production effort. The more agents indeed start producing the chemical, and the more intense competition therefore becomes, the less commercially valuable the know-how encoded in the scientific book of course

376

Handbook of knowledge and economics

becomes. The profitable business opportunity it initially offered can be competed away. In cases in which the value of knowledge declines with its repeated expression, knowledge obviously fails to satisfy the non-rivalry-in-use criterion. (Note, however, that the value of knowledge does not generally vary inversely with the number of knowledge users. There are significant cases in which the value of a piece of knowledge even increases with the number of agents applying it, as is known from the literature on increasing returns to the adoption of network goods; see Arthur, 1994.) For this reason alone, knowledge cannot generally be considered a pure public good. Technical media on which knowledge is encoded can usually be accessed with low or no physical rivalry. Yet, precisely because of this fact, wherever the value of a specific piece of knowledge declines with its diffusion, the technically feasible, frequently repeated expression of the stored knowledge induces rivalry in (the value of its) use. In this respect, knowledge assumes the characteristics of a private good. A similarly mixed result obtains with respect to the exclusion criterion. Again, it is useful to first have a look at the technical terms on which some particular piece of knowledge can be accessed. Agents can always express the knowledge they hold in their own memory. They can also express knowledge stored extra-personally, provided they get access to it and can absorb it. One case in point is indirect communication in which, as discussed above, agents access knowledge encoded on an artificial medium (which, for the present context, implies that the agent is not excluded by another party from accessing the medium). Another case is that of direct communication – be it intentional or unintentional. An agent then gains access to knowledge either by deliberate communication, for example by other agents giving her or him instructions, or knowledge is conveyed unintentionally when the agent observes its expression in the actions of others. Unlike in the case of deliberate communication where exclusion is not intended, unintentionally communicated knowledge is often the result of failure in excluding others. Indeed, the extent to which exclusion can be practised in direct or indirect communication depends on several factors. Actions can have inherent features that make it more or less difficult to infer the knowledge implicitly expressed by them from observing the action. Or there may be technical devices and/or legal regulations actually preventing the observation of actions. Finally, the absorptive capacity and the memory capacity of observers matters. In any case, access by others to knowledge implicitly conveyed by actions – and thus the imitation of the knowledge expression – can only safely be prevented through secrecy, that is, through keeping observers out. Precisely this, of course, is often

Knowledge and its economic characteristics 377 incompatible with the way in which knowledge can be commercially exploited. The access conditions (conditions under which exclusion cannot be practised, as required for a purely public good) depend on the type of knowledge discussed in Section 16.3. By its very nature, tacit knowledge does not satisfy the conditions for a purely public good. The reason is simply that tacit knowledge needs to be produced by each agent by her or his own trial-and-error learning – a natural barrier to a costless access. This restriction does not exclude the possibility, of course, that other agents, who observe someone exercising her or his tacit knowledge, are induced to start their own trial-and-error process that would neither have been induced nor perhaps have had a chance to be successful without such an observation. Observing the actions by which other agents express their tacit knowledge does not suffice, however, to convey that tacit knowledge. Such actions rather convey related, not (yet) codified knowledge that could, however, be verbally described and, hence, be codified. With respect to non-codified, but codifiable knowledge, the question of whether and how exclusion is possible requires a different answer. As long as it is not (yet) codified, this kind of knowledge is stored in the memory of some agent(s). It can be accessed by other agents only if it is revealed in direct communication either intentionally or unintentionally. Since by intentional communication access is deliberately granted, the crucial case is that of unintentional communication. As discussed above, the degree of exclusion here depends either on how far other agents can be prevented from observing the action potentially conveying the noncodified knowledge – the case of secrecy – or it depends on the extent to which observers are unable to infer the knowledge underlying an action, for example because they command an insufficient absorptive capacity. Similar conditions hold for codified knowledge when potential users can physically be excluded from the artificial medium on which it is stored, but not from observing the expression of the knowledge in corresponding actions. If the knowledge is conveyed by the observation, this amounts to the case of unintentional direct communication. An example may help to illustrate the point. Imagine a public music performance in which an artist plays a newly composed piece of music on her instrument. (The very art of playing the instrument to perform the music is based on tacit knowledge that, for the reasons explained, is no pure public good.) Assume that the piece is played from a sheet of music (codified knowledge) inaccessible to the audience or that it is played by the artist by heart (non-codified, but codifiable knowledge). In both cases no transfer of knowledge (of the piece of music) by indirect communication can occur. However, by the very nature of the public music performance as an act of direct communication,

378

Handbook of knowledge and economics

the audience can gain access to the knowledge by listening to the piece. The condition again is that the audience has a sufficient absorptive capacity – in this case sufficient to recollect the music from hearing the performance. The example and the conditions that need to be met can be generalized for knowledge, whether codified or not, to become something of a ‘local public good’ (cf. Cornes and Sandler, 1991), that is, a good freely accessible and reproducible by all agents involved in direct communication in a given location, a sufficient absorptive capacity presumed. Concerning codified knowledge, physical exclusion is possible whenever physical access to the medium can be prevented and unintentional direct communication via the observation of the knowledge expression does not occur or can also be prevented. Where these conditions are met, encoded knowledge is not a purely public good even if there is no rivalry in use. (Where these conditions are not met, encoded knowledge can retain its character as a private good if those who can gain access lack the necessary absorptive capacity to grasp the knowledge.) In the literature, the encoding of knowledge is sometimes held responsible for a metamorphosis of knowledge from a private to a public good. For example, Saviotti (1998, p. 852) writes: ‘it seems intuitively clear that when a piece of knowledge is completely codified, and when all agents know the code, then every agent in the population is capable of rapidly retrieving and using such a piece of knowledge’. Yet, in the light of our discussion this is true only if the codified knowledge is freely accessible. This condition is met in the case of published scientific knowledge, a case that writers like Saviotti perhaps presuppose. But the condition is certainly far from being generally valid. Codification is but the prerequisite for storing knowledge outside human memory. How that external storage is made use of, and by whom, does not hinge on codification, but on the technical features and costs of controlling access to the storage. A degradation of the private-good character of knowledge, codified or not, can however occur through an uncontrollable diffusion. Indeed, this seems to be the most important, though of course not generally valid, argument in favour of assigning knowledge the character of a public good. If the only way of commercially exploiting a piece of knowledge is to express it in the presence of a paying audience or to sell it in an encoded form, for example on print media or electronic media, a further diffusion of that piece of knowledge beyond the original transaction partners may be difficult to keep under the control of the original proprietor. It is particularly difficult if the piece of knowledge can not only be re-accessed over and again via the same technical storage medium, but also be easily copied on to additional media. Nonetheless, lack of control over the further diffusion of a piece of knowledge, which can gradually turn it into a public

Knowledge and its economic characteristics 379 good, does not hinge on whether knowledge is codified or non-codified. It is a problem of unintended communication and the technical means to prevent it.6

16.5 CONCLUSION This chapter started with a brief outline of the technological underpinnings of knowledge storage, access and transfer. The discussion suggested reconsidering the frequently used distinctions of overt versus tacit knowledge and encoded versus non-encoded knowledge and the relationships between them. As a result we found it necessary to distinguish at least three types of knowledge. The first type is encoded knowledge, which can be considered as overt knowledge if and only if the medium on which the knowledge is encoded is freely accessible and the code is commonly known. The second type is non-encoded knowledge, which can, however, be articulated and, hence, be encoded in principle. As long as it is not yet encoded, this type of knowledge can be considered overt only if two conditions are satisfied: when somebody who holds that knowledge expresses it through an action (e.g. articulating it), this must be observable, and it must be possible to infer the knowledge from the observed action. The first condition presupposes an opportunity to participate in what we called direct, face-to-face-based communication that must, moreover, be non-discriminatory. The second condition is implicitly assumed in the literature to be satisfied precisely because the knowledge is articulable. It is the latter condition that makes the difference for the third type of knowledge that we distinguish. This is (at least at the given point in time) non-articulable and therefore inherently non-codifiable or, in short, tacit knowledge. This kind of knowledge cannot be inferred from observing someone expressing it, as is epitomized by the example of keeping balance in riding a bike. It has furthermore been shown in the chapter that the technical terms of storing, accessing and transferring knowledge also affect the key characteristics of knowledge regarding its public-good character. Concerning the characteristics of rivalry in, and exclusion from, the use of knowledge, both have been argued to depend on the locus and the technology of storing and communicating knowledge. Where and how knowledge is stored and can be accessed decides how and by whom it can be ‘expressed’ (used). For this reason, the three different types of knowledge that have been distinguished have to be assessed differently with respect to their public- versus private-good character. As long as knowledge is exclusively stored in the memory of one individual, it is a purely private good in a

380

Handbook of knowledge and economics

trivial sense. Once it is released, this can, but does not have to be, different. People may still be excluded from the communication in which an agent expresses the knowledge she or he holds. Similarly, when knowledge is stored outside the individual’s memory – which presupposes its encoding – storage technology may still allow control of access by others. Furthermore, only if people have an absorptive capacity that allows them to acquire communicated or accessed knowledge and to exploit it without destroying it (no rivalry in use), may that knowledge indeed take on the characteristics of a public good.

NOTES 1. Cohen and Levinthal (1989). Unlike in the present individualistic interpretation, they do not consider storage, access and communication of knowledge in detail, but define conditions under which (firm) organizations are capable of absorbing new scientific developments into their R&D activities. 2. Something similar holds for user knowledge inherently embodied in artefacts, for example tools like a knife, a bow or an axe. If such an artefact is unknown to an agent, she may be unable to produce or even design it by herself. When finding such an artefact, however, it may be possible for the agent to infer its purpose (what it is good for) from its features – in which case the knowledge of its producer is indirectly conveyed. The argumentation relating to artificial media in this chapter may therefore be extended mutatis mutandis to artefacts embodying knowledge. 3. Cf. Foray (2004, ch. 4), who considers this the most important advantage of encoded knowledge. 4. Cowan et al. (2000) express concerns that the variety goes so far that the very notion of tacitness becomes blurred. 5. For expository convenience we do not account here for bottlenecks in accessing a particular storage medium like a book or a hard disc, which may, of course, ration their use and, thus, the knowledge expression at any point in time. 6. The value of knowledge stored on a movie DVD is, for example, almost non-appropriable these days, because it can be reproduced at almost no cost. But when a certain encryption is used, reproduction costs rise significantly and so do the costs of accessing the DVD content – without the original information (knowledge) being changed at all. The encryption code is a feature of the communication process, not of the knowledge itself.

REFERENCES Antonelli, C. (2001), The Microeconomics of Technological Systems, Oxford: Oxford University Press. Arrow, K.J. (1962), ‘Economic welfare and the allocation of resources for invention’, in R.R. Nelson (ed.), The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton, NJ: Princeton University Press, pp. 609–26. Arthur, B. (1994), Increasing Returns and Path Dependence in the Economy, Ann Arbor, MI: University of Michigan Press. Balconi, M. (2002), ‘Tacitness, codification of technological knowledge and the organization of industry’, Research Policy, 31 (3), 357–79.

Knowledge and its economic characteristics 381 Beise, M. and Stahl, H. (1999), ‘Public research and industrial innovations in Germany’, Research Policy, 28 (4), 397–422. Brusoni, S., Marsili, O. and Salter, A. (2005), ‘The role of codified sources of knowledge in innovation: empirical evidence from Dutch manufacturing’, Journal of Evolutionary Economics, 15 (2), 211–31. Callon, M. and Bowker, G. (1994), ‘Is science a public good?’, Fifth Mullins Lecture, Virginia Polytechnic Institute, Science, Technology, and Human Values, 19 (4), 395–424. Cohen, W. and Levinthal, D. (1989), ‘Innovation and learning: the two faces of R&D’, Economic Journal, 99, 569–96. Cornes, R. and Sandler, T. (1991), The Theory of Externalities, Public Goods, and Club Goods, Cambridge: Cambridge University Press. Cowan, R. and Foray, D. (1997), ‘The economics of codification and the diffusion of knowledge’, Industrial and Corporate Change, 6, 595–622. Cowan, R., David, P.A. and Foray, D. (2000), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 9, 211–53. Dudley, L. (1999), ‘Communications and economic growth’, European Economic Review, 43 (3), 595–619. Encaoua, D., Hall, B. and Laisney, F. (2000), The Economics and Econometrics of Innovation, Boston, MA: Kluwer Academic Publishers. Feldman, M. (1994), The Geography of Innovation. Economics of Science, Technology and Innovation, 2, Dordrecht: Kluwer Academic Publishers. Foray, D. (2004), Economics of Knowledge, Cambridge, MA: MIT Press. Grimaldi, R. and Torrisi, S. (2001), ‘Codified-tacit and general-specific knowledge in the division of labour among firms: a study of the software industry’, Research Policy, 30 (9), 1425–42. Langlois, R.N. (2001), ‘Knowledge, consumption and endogenous growth’, Journal of Evolutionary Economics, 11, 77–93. Lazaric, N., Mangolte, P.-A. and Massué, M.-L. (2003), ‘Articulation and codification of collective know-how in the steel industry: evidence from blast furnace control in France’, Research Policy, 32, 1829–47. Lissoni, F. (2001), ‘Knowledge codification and the geography of innovation: the case of Brescia mechanical cluster’, Research Policy, 30 (9), 1479–500. Machlup, F. (1980), Knowledge and Knowledge Production, Princeton, NJ: Princeton University Press. Mokyr, J. (2002), The Gifts of Athena – Historical Origins of the Knowledge Economy, Princeton, NJ: Princeton University Press. Murmann, J.P. (2003), Knowledge and Competitive Advantage, Cambridge: Cambridge University Press. Nelson, K. and Nelson, R.R. (2002), ‘On the nature and evolution of human know-how’, Research Policy, 31, 719–33. Nelson, R.R. (1959), ‘The simple economics of basic scientific research’, Journal of Political Economy, 67 (5), 297–306. Nelson, R.R. (1982), ‘The role of knowledge in R&D efficiency’, Quarterly Journal of Economics, 97 (3), 453–70. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Nonaka, I., Toyama, R. and Nagata, A. (2000), ‘A firm as a knowledge-creating entity: new perspectives on the theory of a firm’, Industrial and Corporate Change, 9 (1), 1–20. Polanyi, M. (1958), Personal Knowledge: Towards a Post-Critical Philosophy, London: Routledge and Kegan Paul. Polanyi, M. (1966), The Tacit Dimension, reprinted 1983, Gloucester: Peter Smith. Roberts, J. (2001), ‘The drive to codify: implications for the knowledge-based economy’, Prometheus, 9 (2), 99–115. Sandler, T. (1975), ‘Pareto optimality, pure public goods, impure public goods and multiregional spillovers’, Scottish Journal of Political Economy, 22 (1), 25–38.

382

Handbook of knowledge and economics

Saviotti, P.P. (1998), ‘On the dynamics of appropriability, of tacit and of codified knowledge’, Research Policy, 26 (7–8), 843–56. Teece, D.J. (1986), ‘Profiting from technological innovation: implications for integration, collaboration, licensing and public policy’, Research Policy, 15 (6), 285–305. Winter, S.G. (1987), ‘Knowledge and competence as strategic assets’, in D.J. Teece (ed.), The Competitive Challenge: Strategies for Industrial Innovation and Renewal, Cambridge, MA: Ballinger, pp. 159–84. Zack, M.H. (1999), ‘Managing codified knowledge’, Knowledge Management and Information Technologies, 40 (4), 45–58.

17 Tacit knowledge Paul Nightingale

17.1 INTRODUCTION What is tacit knowledge and why should economists care about it? This chapter explains what tacit knowledge is – basically a category of unconscious neurophysiological causation that provides the basis and context to actions and conscious mental states – and why the concept is both useful and potentially dangerous. This emphasis on the negative features of the concept is needed because tacit knowledge has recently become extremely fashionable in a range of policy debates and academic disciplines, where its usefulness can, and often is, overplayed. The concept of tacit knowledge has been around for many years, particularly within the technical change and innovation literature (Nelson and Winter, 1982; Dosi, 1982, 1988a, 1988b; Freeman, 1982; Pavitt, 1987; Senker, 1995; Nightingale, 1998) that draws heavily on Polanyi’s (1969) work to conceptualize the empirically robust finding that technological knowledge involves knowledge that cannot be codified or reduced to information. Because knowledge contains an element that cannot be reduced to information, it cannot be traded like information, and requires face-to-face interaction to transfer. As a result, it can be used to explain the localized, uncertain and path-dependent nature of technical learning within people (Pavitt, 1987; Freeman, 1982). Since the routines that firms use to produce technologies depend on these tacit skills, tacit knowledge can be used to explain why firms and nations have persistently different performance over time (Nelson and Winter, 1982; Pavitt, 1984, p. 343; Dosi et al., 1989; Nelson, 1991; Dosi, 1988a, p. 224). Within this literature, tacit knowledge is regarded as a useful but not particularly welldeveloped concept that explains many empirically robust features that cannot be easily accommodated in an information-processing framework (Nightingale, 2003). As such, its use within the innovation literature shows similarities with the psychology literature, where the concept of tacit or implicit knowledge is used to explain a range of empirically robust phenomena, particularly about learning, but is regarded as a not particularly well-defined concept (Berry, 1994; Reber, 1989a, 1989b). Both these literatures draw on the seminal work of the chemistphilosopher Michael Polanyi (1962, 1967, 1969), who argued that scientific 383

384

Handbook of knowledge and economics

knowledge had a tacit component that made it inherently personal and impossible to fully write down. Polanyi needed tacit knowledge to be difficult to operationalize because he was engaged in a public debate with J.D. Bernal about whether science could or should be centrally planned towards political ends. Polanyi was suspicious of Bernal’s Marxism and wanted to protect science from its influence. Much like Hayek’s emphasis on the tacit nature of norms and rules, tacit knowledge was developed as part of a political project against central planning that argued that, since part of science could not be articulated and entered into planning calculations, knowledge production could not, and therefore should not, be managed. In the light of the success of the Manhattan Project, Polanyi’s argument that science cannot be managed has not been taken seriously by many in the science policy community. The real issue is taken to be ‘what degree of management and direction produces the best results?’, rather than whether science can be managed at all. The choice between only markets or only central planning was a false choice with obvious rhetorical appeal. The policy community is interested in the middle ground and it remains an open and contested question if Polanyi’s concept can shed any light on it. Given tacit knowledge’s birthplace in a rather obscure and largely forgotten debate, it seems surprising that it still survives as a concept. However, it has recently moved centre stage in academic and policy debates on geographical localization, technology transfer, ‘competitiveness’ and particularly knowledge management. Part of this rise in prominence is due to the usefulness of the concept, and part is due to changes in external demand factors such as academic fashion and changing policy fads. The very flexibility of the concept that allows it to be applied usefully in some areas also allows it to be applied badly in others. The academic business and management literature is one of the main users of the concept of tacit knowledge, particularly in the literature that tends to follow, rather than lead, business practice. In the mid-1990s firms and consultants realized that the previous business process re-engineering (BPR) management fad of the early 1990s had reduced the workforce through downsizing so much that valuable, but often previously unnoticed, knowledge had been lost from the firm. As BPR approached the end of its life cycle, a new set of practices called ‘knowledge management’ (KM) took its place to help firms perform the tasks that had been disrupted by downsizing. Tacit knowledge played a central role in this literature (Leonard, 1995; Sveiby, 1997; Davenport and Prusak, 1998). Putting aside the obvious point that managers manage people, not knowledge, the KM literature jumped on the notion of tacit knowledge, drawing heavily on research on Japanese management practices.

Tacit knowledge 385 Research during the previous two decades, when the Japanese economy was performing very well, had shown that Japanese firms managed tacit knowledge in different ways to American and European firms, and this was used to explain their superior performance (Nonaka and Takeuchi, 1995; Lam, 1997; Hedlund and Nonaka, 1993). The suggestion was that the culture and management practices in Japanese firms allowed them to develop, codify and share tacit knowledge better than other firms in other countries. Moreover, these practices could potentially be transferred to the West, unlike the institutional configurations, such as lifetime employment, highlighted by Dore (2000) and Whitley (1999). However, there are a number of problems with this suggestion. First, the emphasis on the codification of tacit knowledge would seem to confuse unarticulated with inarticulable knowledge and contrasts sharply with Polanyi’s definition of tacit knowledge as knowledge that cannot be codified. Second, as a universal explanation it overlooks how superior Japanese performance was concentrated in sectors such as automobiles and consumer electronics. By contrast, Japanese performance in sectors such as pharmaceuticals and investment banking, which are known to have a strong dependence on tacit knowledge, is often less good. It would seem that tacit knowledge and its management may be a necessary, but not sufficient, explanatory variable. Third, while many Japanese firms did well in the 1980s and attracted considerable academic interest, by the time their management techniques became fashionable in the mid-1990s, Japan was in a major economic recession. There is clearly not a direct relationship between management practices, tacit knowledge and superior economic performance. It would seem, however, that these problems did not prevent tacit knowledge being used by policy entrepreneurs and management consultants. In the 1990s tacit knowledge, and the idea that it can be codified, became important academic research topics and were used by IT consultants to sell knowledge management services. Over time, as the usefulness of IT-driven KM was questioned, the concept of tacit knowledge was fluid enough to pass into more recent socially focused work on KM (see Tsoukas, 2002 for a critical examination). This conceptual flexibility was also seen during the 1990s in policy making, where tacit knowledge was seen as an important concept that would help policy makers solve some of the policy problems of the ‘knowledge economy’. Research has shown that the rise of the knowledge economy is a real phenomenon, and not just policy hype and a bubble in the high-tech stock market. There is now robust empirical evidence to support Daniel Bell’s claim that theoretical knowledge is increasingly important to modern economies (Bell, 1999; Hicks et al., 2001; Klevorick

386

Handbook of knowledge and economics

et al., 1987). In the 1990s regional, national and supranational policy makers across the political spectrum found the notion of tacit knowledge useful because of its analytical flexibility. The localized nature of tacit knowledge provided a language that both the left and right could use to justify improved training and infrastructure, and subsidies for R&D and industrial support, while maintaining a free-market rhetoric. Despite its extensive use in policy making, critical questions remain about the usefulness of the concept of tacit knowledge. Its current use seems strange given that Polanyi developed the concept to specifically argue that knowledge cannot be managed and policy cannot and should not influence national competitiveness. This chapter aims to address the usefulness of the concept and provide a brief overview of the literature that tries to explain how tacit knowledge went from being a concept used to argue against knowledge management and competitiveness to become a central feature in their respective literatures. The following section (17.2) explains what tacit knowledge is, and relates it to research in psychology and particularly Michael Polanyi’s phenomenology. Section 17.3 explores the implications of tacit knowledge for our understanding of rationality, the embodied nature of learning, and the process of technical change. The implications of these micro-effects at the more macro and meso levels are discussed in Section 17.4, which looks at technology transfer, localization and innovation. The final section (17.5) looks at the dangers of the concept and where it can be misapplied. It argues that the inherent flexibility of the concept makes it prone to over-extension.

17.2 TACIT KNOWLEDGE The classic work on tacit knowledge was by Michael Polanyi (1969). Based on his experience as a chemist, Polanyi argued that knowing involves an actor physically engaging in ‘skilful action’. This skilful action necessarily involves personal knowledge that cannot be articulated, hence his famous aphorism ‘we know more than we can tell’ (Polanyi, 1967, p. 4). Central to this conception of knowledge was the notion that written rules or commands need to be complemented by something else in order for an embodied agent to use and make sense of them. Polanyi highlighted that knowledge is centred on an agent’s body that is physically positioned and interacting with the world. This agent has a front and a back, a left and a right and a sensation of up and down as they move in the world (1969, p. 147). To interact with the world and articulate words and symbols requires action and skill. Since this skill is required to use

Tacit knowledge 387 articulated and codified rules, it is prior to the rules and cannot be reduced to rules. Thus Polanyi argued that skill requires unconscious trial and error learning as we ‘feel our way to success’ (1962, p. 62). This can be contrasted with cognitivist perspectives that typically treat knowledge in terms of abstract information processing that does not have this personal or embodied aspect (Nightingale, 2003). This emphasis on embodiment and movement in skilful action led Polanyi to make a distinction between our subsidiary and focal awareness. When we are engaged in a skilful performance, many of our actions become almost automatic and un-thought. We may be aware of them, but they do not occupy the central or focal part of our awareness. As you are reading this you are probably aware of your feet as part of your subsidiary awareness, but until now were not focusing your attention on them. This shifting in awareness exists because we can only focus our conscious attention in detail on one thing at a time. When we are not focusing on something it recedes into our subsidiary background awareness. Importantly, this also applies to our interaction with tools and technologies, where practice and the development of tacit skilful performance means that they become extensions of our body and we tacitly operate through them. For example, a pianist playing a well-known composition will not think about the movements of each individual finger and key but has learnt to regard them as extensions of her body governed by subsidiary awareness (Polanyi, 1969, p. 148). Skilled performers don’t concentrate excessively on their task, but let their bodies take over. This skilful performance is inherently tacit and cannot be expressed or transmitted in words, as to concentrate on it brings it from subsidiary awareness to focal awareness, where its tacit nature is changed. This can be seen when people attempt to explain how they undertake a skilful performance and lose the ability to perform well as they focus on what they are doing. Within psychology similar concepts are used to conceptualize forms of implicit knowledge that can only be recalled by doing (Anderson, 1983), the distinction between fact memory and skill memory (Schacter, 1992), and the very complex memory systems that go beyond the traditional procedural, semantic and episodic systems (Tulving, 1983; Squire and Butters, 1984). A considerable amount of empirical work has now been undertaken supporting Polanyi’s view that much of our learning and problem solving ability is tacit (or implicit) (Lewicki, 1986; Sternberg, 1986; Lihlstrom, 1987; Reber, 1989a, 1992; Dixon, 1971; Cheeseman and Merikle, 1984; Merikle, 1992; Berry, 1994; Carlo Umilta and Moscovitch, 1994; and Buckner et al., 1995). Research has also highlighted the particular importance of tacit skills and informal learning in professional work (Sternberg, 1997; Eraut, 2000).

388

Handbook of knowledge and economics

This research has shown that people can learn, yet not know they have learned or be able to articulate that understanding. Even simple examples, like riding a bicycle, reveal the tacit nature of knowledge. People generally think that to turn left on a bicycle they turn the handlebars to the left, as seems most obvious. However, if one were to attach a piece of string to the handlebars and pull them to the left, the bicycle would wobble uncontrollably and move to the right (Brown and Duguid, 2000). What people actually do when they turn left is to give a very slight pull to the right, which causes the bicycle to tip to the left and which allows them to move into the turn and be pulled around as the handlebars are then turned to the left. This knowledge is typically unarticulated, and from personal experience can be entirely unknown to the person who can skilfully turn a bicycle. When one tries to bring a subsidiary skill like riding a bicycle into focal awareness, the skill can be lost and has to be relearned. 17.2.1

The Biology of Tacit Knowledge

More recent research in neurology has given this psychology and the concept of tacit knowledge a more rigorous foundation in the biology. This research links neurological causal process to subjective mental states and actions in the world to confirm Reber’s (1992) conjecture that tacit knowledge is an older, more primitive form of knowledge that complements and underpins later evolutionary developments such as consciousness and language. The basic argument is that human cognitive capacities are rooted in our biology, which has developed by adapting pre-existing features to new functions as the primitive mechanisms of homeostasis that maintain organisms within the narrow range of parameters consistent with life have evolved. Tacit knowledge is therefore a reflection of the primitive, unconscious foundations for our more sophisticated cognitive processes (Damasio, 1994). Simple organisms maintain themselves using self-correcting biochemical ‘buffer solutions’, but as organisms become more complex, more sophisticated responses to external stimuli are needed, such as the secretion of specialized chemicals and the use of more sophisticated neurological mechanisms that automatically moderate behaviour through reflexes (Damasio, 1994, 1999). Both these interrelated biochemical and neurological systems can be improved by being linked to memories, to allow responses to be modified through learning (Edelman, 1992). This links the unconscious limbic brain-stem system to the later-evolving cortical systems of the senses that categorize the world (Edelman, 1992). Together these systems produce responses to neural images (either perceived or imagined) in the form of chemical and neurological changes within the body. These responses can

Tacit knowledge 389 be entirely unconscious, which is why even primitive organisms, which lack consciousness, are able to learn. Lesion studies have shown that learning and categorization are compromised when the unconscious parts of these systems are disabled (Damasio, 1994). Other studies have shown that it is possible to learn without being conscious that one is learning (Edelman, 1992; Reber, 1989b; Lihlstrom, 1987; Lewicki et al., 1992). This is possible because neural images require more permanence (500 ms) to become conscious than they do to influence learning (150 ms) (Tononi and Edelman, 1999, p. 1848; Libet, 1992). Consciousness and conscious learning requires additional infrastructure that relates neural images of changes in the body to neural images of the external things that produce those responses in the body (Damasio, 1999). This second-order neural architecture therefore produces an image of your body changing in response to an external object, allowing you to feel the changes it produces in you (Damasio, 1999). This produces a conscious sense of self as a subjective, inner qualitative state (Searle, 1998). Tononi et al. (1999, p. 1849) have produced evidence for the existence of specific neural systems and mechanisms within the brain that act as a searchlight over unattended mental images and bring subsidiary images into conscious, focal awareness (Edelman, 1989; Polanyi, 1967). Such neural systems allow selected images to be set out and brought from subsidiary awareness to conscious focal awareness (Posner, 1994). This ability to concentrate attention has obvious evolutionary advantages, for example, concentrating on escaping from a predator, or concentrating on attacking prey. These advantages can be improved further if attention can be linked to memory and categorization to allow learning from errors. This allows higher primates, and a range of other animals, to live in what Edelman (1989) has called the ‘remembered present’, where our attention to external objects is related to previous memories, learnt categories and their unconscious responses. Imaging studies have shown that learning of this kind changes the distribution of neural activity within the brain. Activity moves from being widely spread across the brain when consciously following explicit rules to become more localized outside the regions of the brain accessible to consciousness, as actions become increasingly tacit and automated (Tononi and Edelman, 1999, p. 1847). This functional isolation produces a ‘gain in speed and precision, but a loss in context-sensitivity, accessibility, and flexibility’ (ibid.). As a result, expert knowledge is often difficult to articulate. This process also creates a distinction between know-what and practicebased, largely tacit, know-how. For example, when someone is learning how to ski they may start consciously following explicit rules such as being told to ‘bend the knees’

390

Handbook of knowledge and economics

and ‘lean into the mountain’. However, as they learn the brain changes to automate the procedures and the explicit rules become unnecessary. As Searle suggests (1983, p. 150): As the skier gets better he does not internalize the rules better, but rather the rules become progressively irrelevant. The rules do not become ‘wired in’ as unconscious intentional contents, but the repeated experiences create physical capacities, presumably realized as neural pathways, that make the rules simply irrelevant. ‘Practice makes perfect’ not because practice results in a perfect memorization of the rules, but because repeated practice enables the body to take over and the rules to recede into the [tacit] Background.

What these, and many other, studies show is that knowledge is embodied, partly tacit, and that learning through practice builds up increasingly automatic responses. Consequently there are forms of knowledge, particularly based around skilful performance, that take many years to learn and cannot be fully articulated or codified (Nightingale, 2003).

17.3 IMPLICATIONS The existence of tacit knowledge is now something that is difficult to argue against given the extent of the psychological and neurological evidence. One can, and should, however, ask what it tells us of any great importance about the economy. Polanyi’s concept of tacit knowledge may now have a foundation in biology, but his larger project involved legitimizing a vision of science where individual scientists were unmanaged and unaccountable. It is hard to see how this larger project receives much additional support because while tacit knowledge can be used to generate an argument against central planning, that does not establish that complete autonomy is the best option. On the other hand, it does seem that the concept of tacit knowledge has some interesting implications for economics. The key insight is that if tacit knowledge exists, then ‘information processing’-based theories that ignore it may be problematic. This insight is particularly important in three areas. First, the link between decision making and embodied unconscious responses – what Damasio (1996, 1997) calls ‘somatic markers’ – means that many of the empirical features of decision making that are anomalies for traditional text-book approaches in economics end up being predictions of the tacit alternative. For example, Kahneman and Tversky’s (1974) finding that even statisticians would prefer an operation in which they had a 90 per cent chance of surviving than one in which they had a

Tacit knowledge 391 10 per cent chance of dying is not (from a tacit knowledge perspective) an unexpected choice. Choices are governed by somatic markers, which are biochemical responses attached to neural images that link to the body’s emotional system (the emotional system is a technical term for the neurological and biochemical mechanisms that regulate body behaviour) (Damasio, 1994, 1996, 1997). Choices with negative connotations have somatic markers that make them less likely to be chosen. Since these somatic markers are a product of evolution they are not necessarily optimal, which is why there are many areas where choices do not match a theoretical ‘rational’ model and might be seen as sub-optimal. This is particularly true where agents have emotional attachments to the subject of their choices, and we would not expect people to make unbiased choices about themselves, their families or external institutions that they use to define themselves, such as nation states. Nor would we expect them to be able to deal well with small risks or large implications. Inferior performance in some areas, however, is made up for by superior performance in others (Damasio, 1996, 1997). The potential overlap between behavioural and evolutionary economics in this area may represent an interesting and potentially fruitful avenue of future research. Second, the existence of tacit knowledge has implications for our understanding of the relationship between information and knowledge. One of the most dangerous ‘bad ideas’ in the policy arena is the notion that knowledge can be assumed to behave like information (Pavitt, 1987, 1996). The existence of tacit knowledge means that there is a part of skilful performance that cannot be codified and will remain embodied within people. As a consequence, it will move with them and may require a range of non-market processes for its efficient development and transfer, that may involve faceto-face interactions and significant levels of trust and experience. The tacit element of knowledge therefore seems to imply that learning will be pathdependent, localized and inherently social. As such, it is unlike the costless, instant process of buying and building up additional information. Because much knowledge cannot be said, but may only be shown, learning involves socialized interactions as individuals are taught by example and have their attempts corrected and modified (Schön, 1987; Bruner, 1990). This learning by doing (Arrow, 1962) can take considerable time (typically five to eight years to develop expert knowledge) and resources. Third, tacit knowledge has particular implications for technological learning and the development of technical capabilities. Since variation in tacit learning influences technological capabilities, we would expect differences in the social processes of technical learning to generate persistent differences in performance between firms over time (Nelson and Winter, 1982; Pavitt, 1984, p. 343; Dosi et al., 1989; Nelson, 1991; Dosi, 1988a, p. 224).

392

Handbook of knowledge and economics

Given the absence of well-functioning markets in embodied knowledge these differences may be difficult to resolve using market transactions. Tacit knowledge might also help explain why markets in technological capabilities do not work efficiently, which helps firms to exploit economic rents while their competitors develop their own capabilities. The uncertain and cumulative nature of technological learning makes it path dependent (Rosenberg, 1976) and can lead to sub-optimal technology choices (David, 1985). This cumulative path-dependence is given considerable emphasis within the theory of the firm where attention is given to how firms change their capabilities in response to changes in technology and the market (Penrose, 1959; Teece, 1977; Teece et al., 1997). However, a cautionary note should be sounded. While tacit knowledge is an explanation for these empirical differences in performance, it is not the only explanation and may not be the correct explanation. As Pavitt and Patel (1997) have shown, firms are diverse in their performance, but have much less diversity in their technological capabilities. This suggests that while it may be true that capabilities take time to develop, turning them into products that are successful in the market once you have them is far from easy. Concentrating exclusively on tacit knowledge may therefore cause one to overlook more important organizational constraints on firm behaviour. Furthermore, one can question how difficult it is to buy tacit skills in the market. Are most labour markets, to some extent, markets in tacit skills? The issue is not necessarily driven by the absolute costs of acquiring tacit knowledge, but is relative to the balance between costs and benefits. Markets in tacit skills may function very well, but firms will keep skills in house if the benefits of the ‘visible hand’ of managerial coordination are superior to the benefits of market coordination (Lazonick, 1991). This potential superior performance of organizations over markets, so that firms should explicitly not be seen as market failures, creates the danger that tacit knowledge can be used as a catch-all term for all the benefits that organizations have. It may therefore conflate various non-cognitive causes of superior performance under a cognitive term. For example, more efficient divisions of labour, incremental innovations to capital goods or better capacity utilization might all be driving superior performance, and could all be mistaken for tacit learning.

17.4 INNOVATION, TECHNOLOGICAL TRANSFER AND LOCALIZATION The emphasis on the impact of tacit knowledge on the relative costs and benefits of different organizational forms in the transfer and coordination

Tacit knowledge 393 of knowledge has meant that the concept has played an important role in technology transfer, economic geography and innovation. These areas are related by the notion that the tacit nature of knowledge makes knowledge difficult to exchange over large geographical, cultural or organizational distances. Within the literature there has been a gradual shift in analytical focus from individual learning, to group learning, then firm learning and finally regional and national learning. These huge differences in scale suggest some scepticism may be in order about the usefulness of conceptualizing such diverse phenomena using a single term such as tacit knowledge. What holds all these different levels of analysis together is the notion that a capability that includes tacit knowledge is required to use technology, and consequently that technology transfer is extremely costly without it. This notion is well established in the innovation literature (Pavitt, 1987; Alic, 1993; Teece, 1977; Vincenti, 1990). Early empirical work on the factors that differentiate between successful and unsuccessful innovation found that technical knowledge had an inherently tacit component and that managerial attention was needed to ensure that various specialized knowledge bases within and between firms were successfully coordinated (Rothwell et al., 1974; Tidd et al., 1997; Senker, 1995). These tacit skills are difficult to articulate and identify (Hicks, 1995) but are needed because technical change is typically uncertain and dependent on skilful interaction with people and technology. As a result, at the micro level organizations rely on communities of practice that share an ‘epistemology of practice’ and enable actors to coordinate within groups that have enough shared understanding to allow effective communication (Cook and Brown, 1999). Nelson and Winter’s (1982) evolutionary theory drew on similar ideas and highlighted the tacit nature of routine activity within firms. They argued that when firms confronted an uncertain future they used partly tacit routines to solve problems. These problem solving heuristics never generate perfectly optimal solutions because of the fundamental and irreconcilable uncertainty firms face, and because of the differences in perceptions about choices that exist. Over time, these partly tacit routines improved and firms developed capabilities that allowed them to solve technical problems. The ‘sticky’ and uncertain nature of tacit skills meant that they are generated far more efficiently through organizational learning than through markets (Penrose, 1959) partly because a degree of tacit knowledge is needed in order for learning to begin. Consequently, starting from scratch without any absorptive capacity makes learning from external sources and taking part in markets for knowledge difficult (Cohen and Levinthal, 1989, 1990). As highlighted previously, this produces persistent differences in the performance of firms (Penrose, 1959; Chandler, 1977;

394

Handbook of knowledge and economics

Nelson and Winter, 1982; Helfat, 1994). Tacit knowledge is therefore linked into the capabilities literature that has developed from Penrose’s (1959) argument that the ability of firms to use ‘bundles of resources’ to produce economically useful ‘services’ depends on their tacit knowledge, which is improved and increased over time, making firms inherently dynamic. The literature that has followed Penrose has used tacit knowledge in a number of ways that are often incompatible with each other. Within this literature, tacit knowledge is regarded as important for varying reasons – for example, because of its importance to uncertain technical change, because innovation requires good internal coordination and trust, because technical change often involves user–producer interactions, and so on. There is also a subset of the literature that argues that codified knowledge can be best coordinated by markets, and tacit knowledge by firms, and within this literature a subsection that argues that IT allows tacit knowledge to be codified, which, almost by definition, should be impossible. Transferring technology is not a simple process as often the technology and its new environment must be adapted to one another (Rogers, 1995). It is no surprise, therefore, that the technology transfer and innovation literatures draw on one another and on the concept of tacit knowledge. What is clear from the empirical evidence is that while we may be experiencing a globalization of trade, markets and regulation, we are not seeing such an extensive process of globalization with R&D (Pavitt and Patel, 1991). Following the distinction between products and technologies (Pavitt, 1996; Freeman, 1982), a distinction can be made between technology transfer within firms and technology transfer as the trade in products between firms. The literature on the diffusion of products (focused primarily on capital goods transferred between firms) has highlighted the need for complementary tacit skills (Teece, 1977; Alic, 1993; Metcalfe, 1988) and the building up of trust for transfer to be successful (Collins, 1990). This goes back to the Industrial Revolution, when English factory owners had to import skilled German craftsmen to get their blast furnaces to work. Even today Arora (1996) has shown that technology transfer to developing countries is being facilitated by packaging licences and tacit know-how together. Despite much good research, it is still unclear whether the difficulties of technology transfer are due to uncertainty, the inherent tacitness of technical knowledge, or because diffusion of technology requires uncertain complementary innovations to the technology and changes within the organization receiving it. When looking at technology transfer within the same organization, tacit knowledge is again considered important. Kogut and Zander (1993) argue that multinational corporations do not exist because of market failures

Tacit knowledge 395 and transaction costs as was traditionally supposed, but instead, they exist because multinational firms have superior ability to transfer knowledge across international borders. Similarly, Powell et al. (2002) found a close spatial relationship between venture capital networks and biotechnology firms, which they attribute to the need for face-to-face contact when transferring tacit knowledge. More than half of the biotechnology firms in their sample received local funding. They point out that the highly complex nature of biotechnology innovation makes the importance of tacit knowledge in their sample higher than may be found elsewhere. This is supported by Pavitt and Patel’s (1991) findings on the non-globalization of R&D, and Cantwell and Santangelo’s (2000) finding that organizational networks are particularly important in science-based industries that are dependent on tacit knowledge.1 The notion that tacit knowledge keeps capabilities localized can easily be extended to the regional level to explain differences in regional performance. However, there is already a substantial and very old literature on economic localization, such as Marshall’s Principles of Economics (1890), which only tangentially addresses tacit knowledge, and explains localization in terms of skilled labour, ancillary trades and specialization in the production process between firms. More recent work by Markusen (1998), Scott (1998) and Storper (1997), for example, has looked at the localization processes in a similar vein and has highlighted the variety of different forms of localization and the variety of reasons why they might form. What is clear from their work is that the localization of production cannot be explained simply in terms of ‘tacit knowledge’ and is dependent on a range of external institutional interactions and internal changes within firms, as highlighted in the national systems of innovation literature (Nelson, 1991; Lundvall, 1992). This literature needs to be distinguished from the literature on ‘clusters’ that draws on the work of Michael Porter (1990) and focuses on the competitiveness of regions (Martin and Sunley, 2001). While there is clearly some overlap, this second literature tends towards tautological explanations whereby ‘clusters’ are ‘defined a priori as beneficial groupings of firms and their existence is then explained in terms of these benefits’ (ibid., p. 24; Storper, 1997). The benefits of localization may be due to collective learning, which is influenced by tacit knowledge being easier to coordinate locally than at a distance (Steiner, 1998; Keeble and Wilkinson, 1999). However, Breschi and Lissoni (2001) question the implicit link between tacit knowledge and localization, and argue that there is little reason to believe that tacit knowledge will ‘spill over’ and ‘hang in the air’. Lawson and Lorenz (1999) draw on the capabilities literature and extend it from firms to the region, highlighting the role of tacit knowledge,

396

Handbook of knowledge and economics

embedded in firm routines in keeping capabilities localized. Their argument is not that regions have localized tacit knowledge, but that firms within regions have tacit knowledge embedded in their operational routines. This emphasis on learning is repeated in Amin and Cohendet (1999), who again question the implicit hard distinction between tacit and codified knowledge in much of the literature and argue instead for the importance of both in developing capabilities at the local level (Gertler, 2001). An insightful view on localization comes from Florida (2002), who argues that tacit knowledge is embodied in people and that these talented people are highly mobile. However, they and their tacit knowledge become localized through their ability to select where they live and work. High quality-of-life locations, which Florida associates as locations having a social milieu that is welcoming to newcomers, attract and retain people, and this is the reason tacit knowledge is localized. Florida’s argument turns the traditional view on its head. Rather than arguing that tacit knowledge is embedded and therefore unlikely to move, he argues that talent is mobile, but is attracted to regions that have cultural and physical attributes that are localized. Lundvall (1988) makes a similar point by highlighting how much learning between firms involves suppliers and customers interacting. Since much of this has to happen face to face, it is localized to the extent that demanding local customers can enable suppliers to innovate because they can interact more and build up the ‘trust’ and shared understanding required for effective learning (see also Maskell and Malmberg, 1999). Again, localization is not the simple result of tacit knowledge being ‘sticky’, but a more complex process of social interaction within networks. Similarly, Lam (2001) has highlighted the importance of these external features and attempted to integrate them into a theory of the socially embedded firm. These works show that while tacit knowledge may play a part in the causes of localization (and regional performance), these causes cannot be usefully reduced to tacit knowledge as many other factors are at work. Looking only at tacit knowledge, and moving from individual, to firm, to regional learning loses sight of all the other more complex contributions to economic change.

17.5 CONCLUSION The ability of tacit knowledge to provide insightful explanations, but also misleading ones, suggests that the concept should be treated with caution. It may be useful, but it can also lead one astray. Part of the problem is that Polanyi (1962) needed a concept that was difficult to operationalize

Tacit knowledge 397 in order to argue against the excessive management of science. We should not, therefore, be surprised that tacit knowledge is so flexible. It is ironic therefore that Polanyi’s concept should be used to support policies and actions he was arguing against. The very fluid nature of the concept, almost by definition ‘something that cannot be defined’, has given it a ‘snake oil’ character whereby it can be used to justify almost any policy. This flexibility has allowed it to be used in a wide range of policy and management areas, but this review suggests that the increased popularity of the concept has more to do with demand factors, particularly the rise of knowledge management, than it does with any great insights that the concept provides. A critical period of reappraisal is therefore probably in order. On the other hand, it is probably impossible to deny the existence of tacit knowledge given the evidence from neurology and biology, but existence and theoretical importance are two different things. This review suggests that the concept does have important implications, particularly about the nature of choice, learning and the development of technological capabilities. First, and most importantly, the existence of tacit knowledge means that knowledge should not be conflated with information. This has important implications for understanding technical change and suggests that many of the theories currently used by policy makers might generate ineffective policy (Pavitt, 1996). Too often, however, tacit knowledge has been used to stabilize, rather than turn over, bad explanations as its fluidity allows it to explain phenomena that traditional theories cannot account for. Tacit knowledge then becomes a universal acid that preserves bad explanations by dissolving anomalies. Second, lessons can be learnt from the way that tacit knowledge has been poorly used. The term ‘tacit knowledge’ is often used as a place-name for a diverse range of causes producing similar effects, many of which have little to do with cognition. Many uses of ‘tacit knowledge’ reduce all kinds of knowledgeable action to the sum of individual behaviour. This ignores how the division of labour allows the economic output of a coordinated group of individuals to be more than the sum of their parts. Third, within the philosophy literature tacit knowledge has been developed further by Searle’s concept of the ‘Background’ (1995). This views tacit knowledge as a social, rather than internal, phenomenon that helps to provide a shared structure that allows individuals to communicate together and coordinate their actions. Searle’s ideas can be used to help to explain the changing structure of innovation processes (Nightingale, 2000a, 2000b). If one explores individual and group learning, this research direction suggests looking at the actual practices at work. Much work on tacit

398

Handbook of knowledge and economics

knowledge focuses on skilled behaviour by individuals (riding a bike, swimming, playing the piano, hammering a nail), while the activity that economists are most interested in involves coordinated groups interacting with technologies (workers in factories, financiers in banks, etc.). These group-based processes of learning (Schön, 1987; Bruner, 1990) are more complicated than a simple trade-off between tacit and codified knowledge (Brown and Duguid, 2000). At the firm level, Tsoukas (1996) has moved away from a codified versus tacit distinction to explore the practices involved in improving firm performance when dealing with socialized dispositions and norm-related, distributed knowledge. Moving outside the firm, Gertler (2001) has argued that a more appropriate direction for research would be to look externally at the social practices that formulate the context and rules of action. Similarly, Martin and Sunley (2001) have criticized excessive reliance on simplistic explanations and emphasized the need for more rigour in the use of concepts and more attention to traditional explanations in economic geography. Together these authors suggest that the concept of tacit knowledge should be used to move explanations outwards from agents, and use it as a foundation for understanding the more complex causal processes, rather than as an ‘explain-everything’ variable. In summary, tacit knowledge is a useful concept when used properly, but its flexibility means that it can be used to explain nearly anything or justify nearly any policy position. Its use should carry a health warning: ‘useful, but handle with care’.

NOTE 1. Almeida et al. (2002) found further evidence that multinational firms were superior to markets and alliances in cross-border knowledge building, and their interviews suggest that this was due to the ability of firms to flexibly use many different forms of knowledge transfer. Similarly, Subramaniam and Venkatraman (2001) found that transnational firms had superior performance in developing new products transnationally.

REFERENCES Alic, J.A. (1993), ‘Technical knowledge and technology diffusion: new issues for US government policy’, Technology Analysis and Strategic Management, 5 (4), 369–83. Almeida, P., Song, J.Y. and Grant, R.M. (2002), ‘Are firms superior to alliances and markets? An empirical test of cross border knowledge building’, Organisational Science, 13 (2), 111–34. Amin, A. and Cohendet, P. (1999), ‘Learning and adaptation in decentralized business networks’, Environment and Planning D: Society and Space, 17, 87–104.

Tacit knowledge 399 Anderson, J.R. (1983), The Architecture of Cognition, Cambridge, MA: Harvard University Press. Arora, A. (1996), ‘Contracting for tacit knowledge: the provision of technical services in technology licensing contracts’, Journal of Development Economics, 50 (2), 233–56. Arrow, K. (1962), ‘Economic welfare and the allocation of resources for invention’, in R.R. Nelson (ed.), The Rate and Direction of Innovative Activity, Princeton, NJ: Princeton University Press, pp. 609–26. Bell, D. (1999), The Coming of the Post Industrial Society, New York: Basic Books. Berry, D.C. (1994), ‘Implicit learning: twenty-five years on: a tutorial’, in M.M. Carlo Umilta and M. Moscovitch (eds), Attention and Performance 15: Conscious and Non-conscious Information Processing, Attention and Performance Series, Cambridge, MA: MIT Press, pp. 755–82. Breschi, S. and Lissoni, F. (2001), ‘Localised knowledge, spillovers vs innovative milieu: knowledge tacitness reconsidered’, Regional Science, 80, 255–73. Brown, J.S. and Duguid, P. (2000), The Social Life of Information, Boston, MA: Harvard Business School Press. Bruner, J. (1990), Acts of Meaning, Cambridge, MA: Harvard University Press. Buckner, R.L. et al. (1995), ‘Functional anatomical studies of explicit and implicit memory’, Journal of Neuroscience, 15, 12–29. Cantwell, J. and Santangelo, G.D. (2000), ‘Capitalism, profits and innovation in the new techno-economic paradigm’, Journal of Evolutionary Economics, 10 (1–2), 131–57. Carlo Umilta, M.M. and Moscovitch, M. (eds) (1994), ‘Preface’, in Attention and Performance 15: Conscious and Non-conscious Information Processing, Attention and Performance Series, Cambridge, MA: MIT Press, pp. xi–xvi. Chandler, A.D. Jr (1977), The Visible Hand, Cambridge, MA: Harvard University Press. Cheeseman, J.M. and Merikle, P.M. (1984), ‘Priming with and without awareness’, Perception and Psychophysics, 36, 387–95. Cohen, W.M. and Levinthal, D.A. (1989), ‘Innovation and learning: the two faces of R&D’, Economic Journal, 99 (3), 569–96. Cohen, W.M. and Levinthal, D.A. (1990), ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35, 128–52. Collins, H.M. (1990), Artificial Experts: Social Knowledge and Intelligent Machines, Cambridge, MA: MIT Press. Cook, S. and Brown, J.S. (1999), ‘Bridging epistemologies: the generative dance between organisational knowledge and organisational knowing’, Organisational Science, 10 (4), 381–400. Damasio, A. (1994), Descartes’ Error: Emotion Reason and the Human Brain, New York: Putnam Books. Damasio, A.R. (1996), ‘The somatic marker hypothesis and the functions of the prefrontal cortex’, Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences, 351, 1413–20. Damasio, A.R. (1997), ‘Deciding advantageously before knowing the advantageous strategy’, Science, 275, 1293–5. Damasio, A.R. (1999), The Feeling of What Happens, Body and Emotion in the Making of Consciousness, London: William Heinemann. Davenport, T. and Prusak, L. (1998), Working Knowledge: How Organizations Manage What They Know, Boston, MA: Harvard Business School Press. David, P.A. (1985), ‘Clio and the economics of QWERTY’, American Economic Review Proceedings, 75, 332–7. Dixon, N.F. (1971), Subliminal Perception: The Nature of a Controversy, New York: McGraw-Hill. Dore, R.P. (2000), Collected Writings, London: Curzon Press. Dosi, G. (1982), ‘Technological paradigms and technological trajectories: a suggested interpretation of the determinants and directions of technical change’, Research Policy, 11 (3), 147–62.

400

Handbook of knowledge and economics

Dosi, G. (1988a), ‘The nature of the innovation process’, in G. Dosi et al. (eds), Technical Change and Economic Theory, London: Pinter, pp. 221–38. Dosi, G. (1988b), ‘Sources, procedures, and microeconomics effects of innovation’, Journal of Economic Literature, XXVI, 1120–71. Dosi, G., Teece, D.J. and Winter, S. (1989), ‘Towards a theory of corporate coherence: preliminary remarks’, unpublished paper, Centre for Research in Management, University of California at Berkeley. Edelman, G.M. (1992), Bright Light, Brilliant Fire: On the Matter of the Mind, New York: Basic Books. Edelman, G.M. (1989), The Remembered Present: A Biological Theory of Consciousness, New York: Basic Books. Edelman, G.M. and Tononi, G. (2000), A Universe of Consciousness: How Matter Becomes Imagination, New York: Basic Books. Eraut, M. (2000), ‘Non-formal learning and tacit knowledge in professional work’, British Journal of Educational Psychology, 70, 113–36. Florida, R. (2002), ‘The economic geography of talent’, Annals of the American Association of Geographers, 92 (4), 743–55. Freeman, C. (1982), The Economics of Industrial Innovation, 2nd edn, London: Pinter. Gertler, M.S. (2001), ‘Tacit knowledge and the economic geography of context or the undefinable tacitness of being (there)’, Paper presented at Nelson and Winter conference, DRUID, Aalborg, Denmark, 12–15 June. Hedlund, G. and Nonaka, I. (1993), ‘Models of knowledge management in the west and Japan’, in P. Lorange et al. (eds), In Implementing Strategic Process: Change, Learning and Cooperation, Oxford: Basil Blackwell, pp. 117–44. Helfat, C.E. (1994), ‘Evolutionary trajectories in petroleum firm research and development’, Management Science, 40 (12), 1720–47. Hicks, D.M. (1995), ‘Published papers, tacit competencies and the corporate management of the public/private character of knowledge’, Industrial and Corporate Change, 4 (2), 401–24. Hicks, D.M., Breitzman, T., Olivastro, D. and Hamilton, K. (2001), ‘The changing composition of innovative activity in the U.S. – a portrait based on patent analysis’, Research Policy, 30 (4), 681–703. Kahneman, D. and Tversky, A. (1974), ‘Judgment under uncertainty: heuristics and biases’, Science, 185, 1124–31. Keeble, D. and Wilkinson, F. (eds) (1999), High Technology Cluster and Collective Learning in Europe, Aldershot: Ashgate. Klevorick, A.K., Levin, R.C., Nelson, R.R. and Winter, S.G. (1987), ‘Appropriating the returns from industrial research and development’, Brookings Papers on Economic Activity, 3, 783–820. Kogut, B. and Zander, U. (1993), ‘Knowledge of the firm and the evolutionary theory of the multinational firm’, Journal of International Business Studies, 24 (4), 625–45. Lam, A. (1997), ‘Embedded firms, embedded knowledge: problems of collaboration and knowledge transfer in global co-operative ventures’, Organization Studies, 18 (6), 976–96. Lam, A. (2001), ‘Tacit knowledge, organizational learning and societal institutions: an integrated framework’, Organization Studies, 21 (3), 487–513. Lawson, C. and Lorenz, E. (1999), ‘Collective learning, tacit knowledge and regions’ innovative capacity’, Regional Studies, 33 (4), 305–17. Lazonick, W. (1991), Business Organisation and the Myth of the Market Economy, Cambridge: Cambridge University Press. Leonard, D. (1995), Wellsprings of Knowledge: Building and Sustaining the Sources of Innovation, Boston, MA: Harvard Business School Press. Lewicki, P. (1986), Non-conscious Social Information Processing, New York: Academic Press. Lewicki, P., Hill, T. and Czyzewska, M. (1992), ‘Non-conscious acquisition of information’, American Psychologist, 47, 796–801. Libet, B. (1993), ‘The neural time factor in conscious and unconscious events’, Experimental and Theoretical Studies of Consciousness, Ciba Foundation Symposium (174), pp. 123–46.

Tacit knowledge 401 Lihlstrom, (1987), ‘The cognitive unconscious’, Science, 237, 1445–52. Locke, J. (1689), An Essay Concerning Human Understanding, Harmondsworth: Penguin (1989 edn). Lundvall, B.-Å. (1988), ‘Innovation as an interactive process: from user–producer interaction to the national system of innovation’, in G. Dosi et al. (eds), Technical Change and Economic Theory, London: Pinter, pp. 349–69. Lundvall, B.-Å. (ed.) (1992), National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter. Markusen, A. (1998), ‘Sticky places in slippery space’, Economic Geography, 72, 293–313. Marshall, A. (1890), Principles of Economics, London: Macmillan. Martin, R. and Sunley, P. (2001), ‘Deconstructing clusters: chaotic concept or policy panacea’, Journal of Economic Geography, 3, 5–35. Maskell, P. and Malmberg, A. (1999), ‘Localised learning and industrial competitiveness’, Cambridge Journal of Economics, 23, 167–86. Maudsley, H. (1876), The Physiology and Pathology of Mind, London: Macmillan. Merikle, P.M. (1992), ‘Perception without awareness: critical issues’, American Psychologist, 47, 792–5. Metcalfe, J.S. (1988), ‘The diffusion of innovation: an interpretive survey’, in G. Dosi et al. (eds), Technical Change and Economic Theory, London: Pinter, pp. 560–89. Nelson, R.R. (1991), ‘Why do firms differ, and how does it matter?’, Strategic Management Journal, 12 (2), 61–74. Nelson, R.R. and Winter, S.G. (1982), An Evolutionary Theory of Technical Change, Cambridge, MA: Belknap Press of Harvard University Press. Nightingale, P. (1998), ‘A cognitive theory of innovation’, Research Policy, 27, 689–709. Nightingale, P. (2000a), ‘Economies of scale in pharmaceutical experimentation’, Industrial and Corporate Change, 9 (2), 315–59. Nightingale, P. (2000b), ‘The product–process–organisation relationship in complex development projects’, Research Policy, 29, 913–30. Nightingale, P. (2003), ‘If Nelson and Winter are only half right about tacit knowledge, which half?’, Industrial and Corporate Change, 12 (2), 149–83. Nonaka, I. and Takeuchi, H. (1995), The Knowledge Creating Company, New York: Oxford University Press. Pavitt, K.L.R. (1984), ‘Sectoral patterns of technological change: towards a taxonomy and a theory’, Research Policy, 13, 343–74. Pavitt, K.L.R. (1987), ‘The objectives of technology policy’, Science and Public Policy, 14, 182–8. Pavitt, K.L.R. (1996), ‘National policies for technical change: where are the increasing returns to economic research?’, Proceedings of the National Academy of Science, 93, 12693–700. Pavitt, K.L.R. and Patel, P. (1991), ‘Large firms in the production of the world’s technology: an important case of non-globalisation’, Journal of International Business Studies, 22 (1), 1–21. Pavitt, K.L.R. and Patel, P. (1997), ‘The technological competencies of the world’s largest firms: complex, and path dependent, but not much variety’, Research Policy, 26, 141–52. Penrose, E.T. (1959), The Theory of the Growth of the Firm, Oxford: Blackwell. Polanyi, M. (1962), Personal Knowledge, Chicago, IL: University of Chicago Press. Polanyi, M. (1967), The Tacit Dimension, London: Routledge. Polanyi, M. (1969), Knowing and Being, Chicago, IL: University of Chicago Press. Porter, M.E. (1990), The Competitive Advantage of Nations, London: Macmillan. Posner, M.I. (1994), ‘Attention: the mechanism of consciousness’, Proceedings of the National Academy of Sciences, 91, 7398–403. Powell, W.W., Koput, K.W., Bowie, J.I. and Smith-Doerr, L. (2002), ‘The spatial clustering of science and capital: accounting for biotech firm–venture capital relationships’, Regional Studies, 36 (3), 291–306. Reber, A.S. (1989a), ‘Implicit learning and tacit knowledge’, Journal of Experimental Psychology General, 118, 219–35.

402

Handbook of knowledge and economics

Reber, A.S. (1989b), Implicit Learning and Tacit Knowledge: An Essay in the Cognitive Unconscious, Oxford Psychological (19), Oxford: Oxford University Press. Reber, A.S. (1992), ‘An evolutionary context for the cognitive unconscious’, Philosophical Psychology, 5, 33–51. Rogers, E.M. (1995), Diffusion of Innovations, 4th edn. New York: The Free Press. Rosenberg, N. (1976), Perspectives on Technology, Cambridge: Cambridge University Press. Rothwell, R., Freeman, C., Horsley, A., Jervis, V.T.P., Robertson, A.B. and Townsend, J. (1974), ‘SAPPHO updated’, Research Policy, 3 (3), 372–87. Scott, A.J. (1998), Regions and the World Economy: The Coming Shape of Global Production, Competition, and Political Order, Oxford: Oxford University Press. Schacter, D.L. (1992), ‘Implicit knowledge: new perspectives on unconscious processes’, Proceedings of the National Academy of Sciences, 89, 11113–17. Schön, D.S. (1987), Educating the Reflective Practitioner, London: Jossey-Bass. Searle, J.R. (1983), Intentionality. An Essay in the Philosophy of Mind, Cambridge: Cambridge University Press. Searle, J.R. (1995), The Construction of Social Reality, New York: Free Press. Searle, J.R. (1998), Mind, Language and Society: Philosophy in the Real World, New York: Basic Books. Senker, J. (1995), ‘Tacit knowledge and models of innovation’, Industrial and Corporate Change, 4 (2), 425–47. Squire, L.R. and Butters, N. (eds) (1984), The Neuropsychology of Memory, Oxford: Oxford University Press. Sternberg, R.J. (1986), Intelligence Applied, New York: Harcourt. Steiner, M. (1998) (ed.), Clusters and Regional Specialisation: On Geography, Technology and Networks, London: Pion. Storper, M. (1997), The Regional World: Territorial Developments in the Global Economy, New York: Guilford Press. Subramaniam, M. and Venkatraman, N. (2001), ‘Determinants of transnational new product development capacity: testing the influence of transferring and deploying tacit overseas knowledge’, Strategic Management Journal, 22 (4), 410–20. Sveiby, K. (1997), New Organizational Wealth: Managing and Measuring Knowledge Based Assets, San Francisco, CA: Berrett-Koehler. Teece, D.J. (1977), ‘Technology transfer by multinational firms: the resource cost of transferring technological know-how’, The Economic Journal, 87, 242–61. Teece, D.J., Pisano, G. and Schuen, A. (1997), ‘Dynamic capabilities and strategic management’, Strategic Management Journal, 18 (7), 509–33. Tidd, J., Pavitt, K.L.R. and Bessant, J. (1997), Managing Innovation, New York: John Wiley and Sons. Tononi, G. and Edelman, G.M. (1999), ‘Consciousness and complexity’, Science, 282, 1846–51. Tsoukas, H. (1996), ‘The firm as a distributed knowledge system: a constructionist approach’, Strategic Management Journal, 17, 11–25. Tsoukas, H. (2002), ‘Do we really understand tacit knowledge?’, in M. Easterby-Smith and M.A. Liles (eds), Handbook of Organisational Learning and Knowledge, Oxford: Blackwell, pp. 411–27. Tulving, R. (1983), Elements of Episodic Memory, Oxford: Oxford University Press. Vincenti, W.G. (1990), What Engineers Know and How They Know It, Baltimore, MD: Johns Hopkins University Press. Whitley, R. (1999), Divergent Capitalism: The Social Structuring and Change of Business Systems, Oxford: Oxford University Press.

18 The firm as a ‘platform of communities’: a contribution to the knowledge-based approach of the firm Ash Amin and Patrick Cohendet

18.1 INTRODUCTION Firms are no longer viewed merely as machines of transactional efficiency, bureaucratic order or labour exploitation. They are seen as repositories of competences, knowledge and creativity, as sites of invention, innovation and learning. In their seminal appeal for an organizational foundation to the theory of the firm, Kogut and Zander (1992, p. 383) claim: In contrast to a perspective based on the failure to align incentives in a market as an explanation for the firm, we began with the view that firms are repositories of capabilities, as determined by the social knowledge embedded in enduring individual relationships structured by organizing principles.

Fransman (1994) has interpreted this critical change as a shift from firms conceived as pure ‘processors of information’ to firms conceived as ‘processors of knowledge’. Viewing the firm as a processor of information has led to an understanding of the constitution and behaviour of the firm as a pure optimal reaction to external signs and factors (market prices) that are detected by the firm. The focus is on the process of allocation of resources needed to cope with such adaptation. Viewing the firm as a processor of knowledge – that is, as a locus of construction, selection, usage and development of knowledge – leads to the recognition that cognitive mechanisms are essential, and that routines play a major role in maintaining the internal coherence of the organization. These theories focus on the processes of knowledge creation. In support of such a vision of the firm, there is growing recognition that the process of production and circulation of knowledge within the firm is a key determinant of the capability of firms to innovate (Kogut and Zander, 1992, 1996; Nonaka and Takeuchi, 1995; von Krogh et al., 1998; Choo and Bontis, 2002). More precisely, there is a growing consensus that the spark of innovation is the interplay of different types of knowledge. For example, Nonaka and Takeuchi’s vision (1995) is based on the idea that knowledge emerges out of a dialogue between people’s tacit and explicit knowledge. 403

404

Handbook of knowledge and economics

They emphasize the interplay of different types of knowledge along two main dimensions: the ‘epistemological dimension’, centred around the critical assumption that human knowledge is created and expanded through social interaction between tacit knowledge and codified or explicit knowledge (‘knowledge conversion’), and the ‘ontological dimension’, which is concerned with the interaction of knowledge held at different levels (individual, group, organizational and inter-organizational). However, besides those strong common principles, one of the main difficulties faced by the knowledge-based approach to the firm is that it is not a homogeneous approach. It can rather be considered as the meeting point of heterogeneous approaches that share a denial of the firm viewed as a processor of information, but that come from different streams of thought. Schematically, we consider that the different perspectives can be grouped into three main theoretical approaches: the ‘strategic management approach’, the ‘evolutionary economics approach’, and the ‘social anthropology of learning approach’. Reducing a multitude of nuances to three schools of thought introduces many risks of simplification and misrepresentation. However, we believe that there is some value in presenting these approaches as ‘flag-bearers’ of common theoretical assumptions. The three approaches can be summarized along the following lines. The strategic management approach is typified by Prahalad and Hamel’s early notion (1990) of core competences in the corporation, but also overlaps with the resource-based view (RBV) of the firm. In this approach, corporate performance is tied to corporate design. The firm’s structure, procedures and environment dictate performance, over and above the behaviour of the individuals in various roles in the organization. This stream of research privileges the governance philosophy of ‘management by design’ – that is, managers deciding on how knowledge should be managed in the organization. They manage, under the constraint of limited attention (the rare resource), the core domain of the corporation, and try to align knowledge activities in the directions they envision. They establish an environment that encourages learning in order to reinforce and stimulate the competences accumulated. The evolutionary economics approach is based on the seminal work of Nelson and Winter (1982). This approach views the firm as a repository of knowledge embodied in the routines of the organization. The evolutionary economics approach is a hybrid theory that includes the fundamental principles of any evolutionary theory – including the principle of heredity played by routines, the principle of generation of variety, and the principle of selection – and an emphasis upon routines as the key collective organizational device for cognition. Organizational routines are considered the building blocks of the core dynamic capability that expresses the firm’s

The firm as a ‘platform of communities’ 405 ability to integrate, build and reconfigure internal and external competences to address rapidly changing environments. In contrast to the strategic management approach, the evolutionary economics approach does not privilege any particular cognitive role for the manager. Cognitive effort is considered to be shared by all members of the firm. The social anthropology of learning approach is inspired by the work of Lave and Wenger (1991) and Brown and Duguid (1991). This approach centres its interest in the actual process of how knowledge is formed and made explicit through social interaction. The emphasis is on the working community as an active entity of knowing, which reveals specific forms of knowledge through its daily practices. Knowing is harnessed to the sociology of interpersonal and collective relations in firms, influenced by factors such as trust and reciprocity, corporate narratives, languages of communication and socialization strategies. This approach favours organizing for learning in doing, in both an experimental and a path-dependent nature, and is based on working with the social dynamics of communities and organizational cultures. These three approaches are rooted in very different understandings of knowledge. They vary in terms of the entity that activates, nurtures, stores and develops knowledge. They differ in their identification of the learning mechanisms at stake. They place different emphases on incentive mechanisms to stimulate learning and innovation. Differences of this scale immediately suggest that any attempt to weave the three strands into a unified perspective on knowledge creation and management is fraught with difficulty, running the risk of mixing epistemological incompatibles. For example, the strategic management approach and the evolutionary economics approach are associated with an ‘epistemology of possession of knowledge’ (Cook and Brown, 1999) in that both tend to see knowledge as something people possess, while, for the social anthropology of learning approach, what matters is the ‘knowing’ that emerges from the pragmatics of individual and group practices. As Cook and Brown (1999: 381, emphasis in original) explain: Much current work on organizational knowledge, intellectual capital, knowledge-creating organizations, knowledge work, and the like rests on a single, traditional understanding of the nature of knowledge. We called this understanding the ‘epistemology of possession’, since it treats knowledge as something people possess. Yet this epistemology cannot account for the knowing found in individual and group practice. Knowing as action calls for an ‘epistemology of practice’ . . . We hold that knowledge is a tool for knowing, that knowing is an aspect of our interaction with the social and physical world, and that the interplay of knowledge and knowing can generate new knowledge and new ways of knowing. We believe this generative dance between knowledge and knowing is a powerful source of organizational innovation.

406

Handbook of knowledge and economics

Thus the three main approaches of the knowledge-based theory of the firm should not be considered as a unified framework, but as complementary theoretical views where each one highlights a specific aspect of knowledge. As Grant (2002, p. 133) underlines in describing the different streams of research on the knowledge-based theory of the firm, ‘the outcome was not so much a new theory of the firm as a number of explorations into aspects of the firm and the organization of production that were unified by their focus on the role of knowledge as a factor of production’. Thus a powerful and relevant theoretical building of the knowledge-based view of the firm requires a clear understanding of how the three main approaches complement themselves. In the economics literature the strategic management approach and the evolutionary economics approach have already received considerable attention. Scholars have recognized the importance of the third stream, the social anthropology of learning approach, in particular to understand local practices in the work environment. However, the theoretical contribution of this approach to the knowledge-based view of the firm is still lacking. The aim of this chapter is thus to fill this analytical gap, considering the immense theoretical potential offered by looking at knowledge as a process and practice, rather than as a possession, in relation to the pragmatics of everyday learning in situated contexts of embodied and encultured practice. In its attempt to highlight the contribution of the social anthropology of learning approach to the knowledge-based theory of the firm, this chapter naturally focuses on community as the unit of analysis and all-important site of knowledge formation – the site where hybrid knowledge inputs meaningfully interact. As Brown and Duguid (1991, p. 53) claim, ‘it is the organization’s communities, at all levels, who are in contact with the environment and involved in interpretative sense making, congruence finding and adapting. It is from any site of such interactions that new insights can be co-produced.’ Accordingly, we assume that the process of generating, accumulating and distributing knowledge – in both sites of informal interaction and informally constituted units such as R&D labs – is achieved through the functioning of informal groups of people, or autonomous ‘communities’, acting under conditions of voluntary exchange and respect for the social norms that are defined within each group. Communities can be considered as key building blocks of the organization and management of corporate innovation and creativity. The chapter is structured as follows. In the first section, the notion of ‘knowing communities’ as the elementary units of knowledge formation in the firm is analysed. The second section investigates the vision of the firm as a ‘platform of communities’ as suggested by Brown and Duguid

The firm as a ‘platform of communities’ 407 (1991). Then in the third section, the contribution of this approach to the knowledge-based views of the firm and the ways it complements the strategic management approach and the evolutionary economics approach is discussed.

18.2 KNOWING COMMUNITIES Many recent works (e.g. Brousseau, 2001; Gensollen, 2001; Cowan and Jonard, 2001) emphasize that in an economy increasingly based on knowledge, a growing part of the process of generation and diffusion of knowledge is ensured by the functioning of ‘knowledge-intensive communities’ or ‘knowing communities’. To a great extent, the current economics literature on communities addresses the issue of how communities of practice (Lave and Wenger, 1991) accumulate knowledge through social interaction, how virtual communities work in relation to the development of the Internet (Lerner and Tirole, 2001), or how scientific communities create new knowledge (Knorr-Cetina, 1999, Cowan et al., 2000). All these communities have in common the fact that they consist of agents who interact frequently by means of a non-hierarchical architecture of communication about a common interest or objective in a given field of knowledge. An essential characteristic that arises from the analysis of these systems of voluntary cooperative exchange is the importance of the norms of behaviour that guide the actions of the members of the community, as well as the intensity of the relations of trust that seems to control the relations. As Bowles and Gintis (2000) underline, the notion of ‘community’ is one of the oldest economic concepts; it pre-dates the modern values of market and planning, but is condemned to history as ‘the anachronistic remnant[s] of a less enlightened epoch that lacked the property rights, markets and states adequate to the task of governance’ (p. 15). In particular, the parochialism of community has been considered antithetical to modern institutions, an old-fashioned idea in the context of market and state institutions. However, communities have survived the emergence of modern social institutions, not least because of their important contribution to governance, when market contracts (in the provision of local public goods, for example) and government fiats have failed. Associations, neighbourhood groups and other forms of grouping offer efficient arrangements that are not plagued by the usual problems of moral hazard and adverse selection, or by the illusion that governments have both the information and the inclination always to offset market failures. Among the many various types of communities, the recent fast-growing importance given to the notion of ‘knowing communities’ can be explained

408

Handbook of knowledge and economics

as follows: as the knowledge base of society dramatically expands and becomes more and more complex, traditional hierarchical structures increasingly face difficulties with integrating and consolidating dispersed parcels of knowledge. These parcels are increasingly generated by and consolidated in informal collective contexts or ‘knowing communities’, suitable to deal with some of the irreversible sunk costs associated with the processes of creation and maintenance of knowledge. Thus knowing communities appear as genuine active units of competences, which are useful to the organization as a whole since they assume a significant part of the processes of production, accumulation and validation of knowledge. These communities can be formed within traditional hierarchical settings (within functional departments, or project teams), but can also cut across the hierarchical structures of the firm by gathering members interested in a particular field of knowledge. 18.2.1

Different Types of Knowing Communities

Among the active knowing communities in the process of knowledge creation, two major types of communities can be distinguished: epistemic communities, which are really geared towards the creation of new knowledge; and communities of practice, which are aimed at the success of an activity, and for which knowledge creation is unintentional. In terms of knowledge, each community is characterized by a principal learning mechanism (circulation of the ‘best practices’ for the communities of practice, publications under control of peers in some epistemic communities). The frequent interactions of members of the community naturally reinforce the cohesion of the learning processes. 18.2.1.1 Epistemic communities Epistemic communities are, according to Cowan et al. (2000, p. 220), ‘small groups of agents working on a commonly acknowledged subset of knowledge issues and who at the very least accept a commonly understood procedural authority as essential to the success of their knowledge activities’. The members of an epistemic community have a common objective of deliberate knowledge creation. To this end, they gradually construct a common structure allowing a shared understanding. These communities are, for example, groups of researchers, a ‘task force’ or a group of designers within a firm, a ‘school’ in painting or music. What binds these communities is the existence of a procedural authority, that is, a set of rules or codes of conduct defining the objectives of the community and the means needed to be implemented in order to reach them. Epistemic communities are thus structured around an objective ‘to reach’ and a procedural

The firm as a ‘platform of communities’ 409 authority founded by themselves (or with which they were founded) in order to carry out this objective. This form of organization generates the creation of knowledge by supporting the synergy of individual varieties. Because of the heterogeneity of the representatives, one of the first tasks of the epistemic communities is to create a ‘codebook’ in the sense of setting out a dictionary and a grammar so that the cognitive work can take place. The validation of the cognitive activity of a representative of the community is made according to the criteria fixed by the procedural authority. What is evaluated is the individual contribution to the effort towards the collective objective to be reached. 18.2.1.2 Communities of practice Communities of practice (Lave and Wenger, 1991) represent groups of people engaged in the same practice communicating regularly between themselves about their activities. The members of a community of practice primarily seek to develop their competences in the considered practice, while spreading and comparing incessantly the ‘best practices’ tested by the members. The communities of practice can be seen as a means of developing individual competences through the continuous improvement of the common practice. This objective is reached through the construction, the exchange and the sharing of a repertory of resources; this repertory is not necessarily formally clarified. Self-organization is thus an essential characteristic of the communities of practice. More precisely, the autonomy and the identity of the communities – two key characteristics of self-organization – authorize the collective acquisition and treatment of the stimuli coming from the environment (Wenger, 1998a; Dibiaggio, 1998). Thus it is the mutual commitment of its members that ensures the cohesion of the community, and it is this same commitment that governs the recruitment of new members. The evaluation of an individual by the community of practice relates at the same time to the values adopted by the individual and the progress made in its practice. Within the communities of practice, knowledge is thus primarily the ‘know-how’ (Brown and Duguid, 1991) that is tacit and socially localized. This collective know-how is constantly enriched by the individual’s practices, so that the collective device generates a process of continuous creation of knowledge useful for the organization (although unintentional, as opposed to the epistemic communities). In the literature that has grown to recognize learning in communities, the distinction between epistemic communities and communities of practice tends to be based on a linear representation of the process of knowledge transformation. This process is viewed as evolving from separate departments in charge of producing new (deliberate) knowledge

410

Handbook of knowledge and economics

or handling and distributing information to the other departments that assimilate and use this new knowledge to improve their current activities. These departments can also produce new knowledge from their routine activities, but this is a non-deliberate form of knowledge production that emerges as a by-product of learning by using or learning by doing. In other words, this vision separates on the one side the units in charge of the ‘exploration’ dimension, and on the other side, the units in charge of the ‘exploitation’ dimension. We argue, in contrast, that, while the separations between deliberate and non-deliberate forms of knowledge and between tacit and codified knowledge are useful analytically as a means of dissecting a complex knowledge process, it would be an error to assume that these separations hold in reality. While epistemic communities may be established explicitly as knowledge communities, the sociology of their knowledge practices is not radically different from that of communities of practice. We would, thus, go further than the currently fashionable claim that the rise of the knowledge economy is encouraging convergence on the grounds that the essence of successful innovation and learning lies in the ways these two types of communities deliberately interact and jointly organize the production and circulation of knowledge. 18.2.2

Properties of Knowing Communities

As informal groups, knowing communities exhibit specific characteristics that distinguish them from the traditional organized entities usually analysed in economics or business science: ●

●

●

Communities have no clear boundaries, and there is no visible or explicit hierarchy that can control the quality of work or the respect of any standard procedure. It has been repeatedly argued that what holds the community together is the passion and commitment of each of its members to a common goal, objective or practice in a given domain of knowledge. Thus the notion of contract is meaningless within the members of the community, and in particular there is a priori no motive to think of any financial or contractual incentive devices to align the behaviour of the members of the community. The recent literature has emphasized that some of the specific motives that guide the behaviours of members of the community could have an economic interpretation. Frequency of interactions within the community considerably reduces opportunistic behaviours. With repeated interactions, delays and moral hazard problems

The firm as a ‘platform of communities’ 411

●

will be attenuated through the creation of norms of cooperation and routines,1 as well as the intensification of reputation mechanisms. Therefore a large part of agency problems will be resolved spontaneously in the knowledge-based economy. The validation of the knowledge takes place in the first analysis within a given community. In the same way, the interpretation of the knowledge provided by the outside (in particular by the hierarchy) is examined, criticized and reprocessed (to lead sometimes to creative adaptations) within communities.

Among these common traits, one is repeatedly underlined by the literature: members of a given community have common values, and their interactions are governed by a type of trust grounded in the respect of the common social norms of the community (Lazaric, 2003). As underlined by Adler (2001), this question of trust is central to the understanding of a community. For Adler, it is precisely because of their extensive reliance on trust that the economic foundations of knowledge-intensive communities deviate from standard institutions. In a socio-economic world turned towards resource creation, trust emerges as a device allowing economic action and mutual transactions. It allows firms and organizations to deal with the radical uncertainty of the environment and the opportunism of agents and to minimize at the same time the costs of strong contract-based incentives. Trust allows agents to initiate and maintain cooperation and to engage in a continuous process of knowledge creation. The existence of trust is thus a prerequisite for knowledge creation, sharing and use. However, the nature of trust we refer to when analysing knowing communities must be clearly defined. It departs from the tradition of ‘calculative’ trust, central to classical economics.2 In our view, when focusing on knowledge-intensive communities, trust is more than a simple lubricant, neither reduced to a merely calculative process, nor to purely interpersonal attachment. It is rather a prerequisite for cooperation and coordination in socio-economic systems characterized by uncertainty and incomplete contracts. In such a perspective, we follow Nooteboom (1999), who departs from Williamson’s approach and argues that non-calculative trust is not unavoidably blind. I argue that it can go beyond calculative self-interest without being blind . . . It is not thereby blind, for two reasons. First, routinization is based on proven past performance and reliability of a co-operation relation, and thus has a rational basis even through it is no longer based on conscious deliberation. Second, trust is indeed not unlimited: it applies only up to some ‘golden opportunity’ of opportunism which goes beyond a partner’s stability to resist temptation, or up

412

Handbook of knowledge and economics

to a crisis which may force a partner to defect in order to survive. (Nooteboom, 1999, pp. 797–8).

According to this vision, trust represents a bilateral psychological state that leads an individual, in a given situation, to suppose that another individual will adopt a behaviour that a priori conforms to his or her own. ‘Trust is the expectation that arrives within a community of regular, honest and cooperative behavior, based on commonly shared norms on the part of other members of that community . . .’ (Fukuyama, 1995, p. 26). Trust is an expectation or belief that the other party will act benevolently. It describes a situation in which the counterpart cannot be forced to fulfil this expectation, with the risk that expectations may not be fulfilled. Therefore trust consists in accepting a certain level of risk (or vulnerability), to be validated by experience and practice (Coriat and Guennif, 1996). Trust allows ‘ambivalent engagements’ to be turned into ‘believable engagements’, through a set of repeated acts that constitute moments of verification.3 In a knowledge-intensive community, the driving force of building trust is that members are aiming at reinforcing a specialized field of knowledge: the more intense the cognitive efforts to build and consolidate a given field of knowledge, the higher the level of cognitive trust between agents. There is a co-evolutionary link between building a piece of specialized knowledge and building cognitive trust.4 Cognitive trust refers to the ability of the individual to perform tasks related to a current practice. It refers to both judgments of competence and reliability about the other members of a team . . . Judgments of competence are based upon verifying instances of predictably professional behavior (i.e., correct task execution), while reliability refers to the congruence between words and actions (i.e., respect for deadlines). (Rocco et al., 2000, p. 12)

The cognitive dimension focuses on the ‘rational’ basis for trust, building on partial knowledge and frequently involving a search for the evidence on which to base one’s trust, focusing on characteristics such as the competence, reliability and credentials of counterparts. Models of trust primarily rooted in the cognitive dimension include intentional trust, competence trust and system trust Intentional trust simply indicates the will of the partner to maintain commitment. Calculation of self-interest includes reputation (Weigelt and Camerer, 1988) and the assessment of future benefits of present cooperativeness. We consider that intentional trust is the dominant type of trust that characterizes knowing communities. Intentional trust within a knowing community can be observed when the behaviour of the participants,

The firm as a ‘platform of communities’ 413 exposed to an unexpected event, are not guided by any form of contractual scheme, but by the respect of the social norm of the group (Arena et al., 2005; Szulanski et al., 2004). Competence trust refers to the expectations that the partner will be able to do a certain thing in a certain way. This type of trust is frequent in hierarchical groups such as project teams, but could also operate in a context where members of different communities are exchanging knowledge (we later examine a case of coordination of different communities of specialists in a hospital). System trust is a form of cognitive trust based on expectations of behaviour in accordance with social roles, as is the concept of trust in trust, whereby one relies upon the trust of others as a basis for extending trust. All the above features that characterize learning communities contribute to make clear the differences between communities and the other forms of coordination units that can be found within the firm: ●

●

●

Knowing communities differ from functional groups. Contrary to communities, these units are under the responsibility of a hierarchy at the top of them, with clear boundaries separating those belonging to the unit and those apart from it. Of course, such functional units can contribute to the process of accumulation of knowledge. However, when compared to communities (where cognitive links are continuously activated and enhanced ‘naturally’ between members), these units require considerable efforts to be made in order to activate the conservation of routines, the power of replication of the routines, and the continuous improvement of the routines between members. Moreover, while communities are loci of active and deliberate learning processes between members in order to create, exchange and accumulate knowledge, functional units are mainly characterized by passive modes of learning, such as ‘learning by doing’, which has been much described by the literature. Knowing communities differ from project teams, or task forces. These teams of employees with heterogeneous skills and qualifications are often coordinated by team leaders and put together to achieve a particular goal in a given period of time. Communities may share some traits with teams: for instance the group interest generally coincides with the interest of the members, as noted by Marschak (1954). However, there is no visible hierarchy in communities, nor any constraint of time in the process of knowledge generation and accumulation. Knowing communities differ from coalitions in the sense that the strategic calculation of agents does not generally determine their

414

●

Handbook of knowledge and economics adherence to a given community. Moreover, a coalition has by definition a clear boundary, contrary to communities. Knowing communities differ from cliques in network theories. If cliques share with communities the characteristic of having no clear boundaries, they differ in the sense that the relationships between agents within a clique do not generally express the cognitive dimension. They do not address a clear objective of creation and accumulation of knowledge.

Knowing communities are repositories of useful knowledge, which is embedded in their daily practices and habits. The local daily interactions constitute an infrastructure that supports an organizationally instituted learning process that drives the generation and accumulation of knowledge by the community. Most of the time, the accumulation of knowledge by a given community is shaped by a dominant mode of learning (such as ‘by circulation of best practices’) adopted by the community, and can circulate through the existence of a local language understandable by the members only. The communal setting provides the context in which the collective beliefs and the representations structuring the individual choice are built. Communities allow the strengthening of individual commitments in an uncertain universe. Individuals remain attentive to the specific contexts and can therefore update the shapes of their cooperative engagements.

18.3 THE MYRIAD OF COMMUNITIES WITHIN THE FIRM In and across organizations, there are many different kinds of formal of informal groups, which vary in remit, organization and membership (Cohendet et al., 2000). Firms have been seen traditionally as constellations of diverse formal groups of learning sites where knowledge is formed, practised and altered (Liedtka, 1999). These groups might be found in traditional work divisions and departments, but they also cut across functional divisions, spill over into after-work or project-based teams, and straddle networks of cross-corporate and professional ties. For example, within firms, classical formal (or hierarchical) groups include functional groups of employees who share a particular specialization corresponding to the hierarchical division of labour (e.g. marketing or accounting). They also include teams of employees with heterogeneous skills and qualifications, often coordinated by team leaders and put together to achieve a particular goal in a given period of time.

The firm as a ‘platform of communities’ 415 However, the main point we wish to underline in this chapter is that the production and diffusion of knowledge within the firm appear also increasingly embedded in those informal contexts and structures that have been defined above: the knowing communities. Some of these communities emerge spontaneously from the hierarchical structures of the firm (some workshop staff may constitute a community of practice overlapping with the functional division of operations in the firm), while some communities may result from an adherence to a common passion of very dispersed individuals within the firm (e.g. a community of practice of people interested in computing in a given organization will not in general overlap with the staff of the computer department, but may comprise agents of the firm working in different positions, departments and even locations of the firm). In such a perspective, one of the major roles of the firm is to bring coherence to the interactions of these various communities. As underlined by Brown and Duguid (1991, p. 53), in this representation, the firm is perceived as a collective of communities, not simply of individuals, in which enacting experiments are legitimate, separate community perspectives can be amplified by inter-changes among communities. Out of this friction of competing ideas can come the sort of improvisational sparks necessary for igniting organisational innovation.

18.3.1

Advantages of Coordination by Communities

In the knowledge-based firm, coordination by informal communities has some advantages compared with the traditional mode of coordination by the hierarchy. In a nutshell, it offers the possibility of ‘economizing’ on hierarchy. One of the major advantages is that in so far as the implementation of knowledge rests on the existence of common language and representations, the accumulation and the treatment of knowledge are done ‘naturally’ within a given community, without a pressing need to resort to powerful incentive mechanisms. The community constitutes a place of trust, in the strong sense, for each one of its members. Thus, in the unforeseen situations, commitments will not be guided by the spirit of the contracts, but by the respect of the social norms specific to the community. As already seen, one of the main characteristics of communities is that they ‘freely’ absorb the sunk costs associated with building the infrastructure needed to produce or accumulate knowledge, either in a completely non-deliberate manner (embedded in their daily practices), or in a more deliberate (but still ‘free’) way corresponding to their willingness to contribute to the cognitive building of knowledge. Viewed from the perspective of a hierarchical organization, the building (or reproduction) of

416

Handbook of knowledge and economics

such an infrastructure to accumulate knowledge (definition of a common language, definition of dominant learning processes etc.) would entail significant sunk costs comparable to the sunk costs required by any process of codification. Members of communities can take charge, through passion and commitment to a common objective or practice, of the sunk costs of the process of generation or accumulation of specialized parcels of knowledge. Communities are developers and repositories of useful knowledge embedded in their daily practices and habits. The local daily interactions constitute an infrastructure that supports an organizationally instituted learning process that drives the generation and accumulation of knowledge by the community. Most of the time the accumulation of knowledge by a given community is shaped by a dominant mode of learning adopted by the community (such as by circulation of best practices), and can circulate through the existence of a local language understandable only by the members. Otherwise, coordination by communities does not require the implementation of heavy (and costly) extrinsic incentives. A key characteristic of communities is the absence of a visible hierarchy and the fact that unlike other institutions, communities do not need ‘alternative bundles of contracts understood as mechanisms for creating and realigning incentives’ (Langlois and Foss, 1996, pp. 10–11). It has been repeatedly argued that what holds the community together is the passion and commitment of each of its members to a common objective or practice. Members of a given community have common values and the interactions between them are governed by a type of trust grounded on the respect of common social norms of the community. Trust within the community can be measured when one can observe that the behaviour of the participants, exposed to an unexpected event, is not guided by any form of contractual scheme, but by the respect of the social norms of the group. An effective community monitors the behaviour of its members, rendering them accountable for their actions. In contrast with hierarchical coordination modes, communities more effectively foster and utilize social incentives that collectives have traditionally deployed to regulate their common activity: trust, solidarity, reciprocity, reputation, personal pride, respect, vengeance and retribution, among others. The development of various communities corresponds to a progressive division of the tasks of knowledge creation where each community specializes in a parcel of new knowing. ‘Such communities, which are situated at intermediate levels of organizational structure, can efficiently set out heuristic and exploratory routines in order to deal with specific problemsolving activities’ (Cohendet and Llerena, 2003, p. 273).

The firm as a ‘platform of communities’ 417 Moreover, communities allow the stabilization of individual commitments in an uncertain universe. Individuals remain attentive to specificities of the situations and can consequently update the forms of their cooperative commitment. Sense construction in a community is essentially a procedural step. ‘Community provides the interpretative support necessary for making sense of its heritage’ (Wenger, 1998b, p. 98). Communities are thus ‘suppliers’ of sense and collective beliefs for the agents and play a central role of coordination in the organization. The community framework provides the context within which are built the collective beliefs and the reference systems that structure individual choice. Adopting the idea that knowledge creation is primarily realized in contexts of action and that the action is always collective, the consideration of the intermediate level of communities is thus necessary to focus on the learning in the processes of action (Dupouët and Laguecir, 2001). 18.3.2

Limitations of Coordination by Communities

In certain circumstances, however, autonomous communities can pose problems as a knowledge governance mechanism. Many risks of failure can be identified. One of the major causes of failure is the risk of parochialism, discrimination or vengeance on other communities, or autism or incompatibility with the hierarchical imperatives of organizations. A second associated problem is the weakness identified in the community, as pointed out by Bowles and Gintis (2000, p. 11): Where group membership is the result of individual choices rather than group decisions, the composition of groups is likely to be more culturally and demographically homogeneous than any of the members would like, thereby depriving people of valued forms of diversity. To see this, imagine that the populations of a large number of residential communities are made up of just two types of people easily identified by appearance or speech, and that everyone strongly prefers to be in an integrated group but not to be in a minority. If individuals sort themselves among the communities there will be a strong tendency for all of the communities to end up perfectly segregated for reasons that Thomas Schelling (1978) pointed out in his analysis of neighbourhood tipping. Integrated communities would make everyone better off, but they will prove unsustainable if individuals are free to move.

A third limitation is the risk of lack of inputs variety: The personal and durable contacts that characterise communities require them to be of relatively small scale, and a preference for dealing with fellow members often limits their capacity to exploit gains from trade on a wider basis. Moreover, the tendency for communities to be relatively homogeneous may make it impossible to reap the benefits of economic diversity associated with

418

Handbook of knowledge and economics

strong complementarities among differing skills and other inputs. (Bowles and Gintis, 2000)

Here perhaps lies the main potential disadvantage of communities when compared to deliberate governance. Most of the time the attention of the members of a given community is focused on a specialized topic but the emergence of diversity requires in general the creative interaction of different communities. Working by community (by definition around a practice or particular cognitive objective) can stifle knowledge fit across the organization (e.g. integration of several heterogeneous professional bodies, cross-fertilization etc.). This is particularly the case when different communities have difficulties in communicating with each other (we shall come back to this point in the next section). On this matter, organized schemes, by offering the possibility of voluntarily bridging disciplines or groups (e.g. team projects, interdisciplinary groups), may offer significant advantages. But in what sense can we say that this myriad of heterogeneous groups interacting with one another constitute a coherent collective, underpinned by certain global norms and a unique organizational culture? Communities are disparate and dispersed, thus another management challenge is the alignment of different types of community within the firm. Nothing guarantees, a priori, the systematic alignment of interests and objectives of the different communities in place. The constituent communities of an organization are not necessarily homogeneous or convergent towards a common objective. The risks of intercommunal conflicts, autism or parochial partitioning are latent. A global vision of the organization conceived as a coherent assembly of communities is therefore necessary. This implies thinking in terms of combining soft and hard structures of learning within a given organization. This perspective raises the issue of the delicate matching between the ‘visible’ hierarchical structure of the firm and the ‘invisible’ process of active knowledge formation that occurs within the myriad of communities in a given firm. A too strict hierarchy enforcing members to follow the rules decreed by the ‘visible structures’ would prevent the firm from benefiting and deriving value from the knowledge accumulated by the ‘invisible communities’. Such decisions certainly would not eliminate the functioning of community (people would continue to talk and exchange ideas about their practices), but the useful knowledge accumulated at the level of the respective communities would be prevented from flowing to the rest of the organization. The hierarchy cannot influence the internal functioning of communities, but can find ways to let the knowledge accumulated by communities flow and bring

The firm as a ‘platform of communities’ 419 value to the firm. On the other hand, leaving to the communities the whole process of knowledge creation and formation will expose the firm to risks of incoherence, inconsistency and anarchy. To a large extent, hierarchy is involved in fine-tuning the cognitive distance between communities (Nooteboom, 2002). Thus the definition of the suitable modes of governance supposes a thorough analysis of the interactions between communities.

18.4 THE FIRM AS A ‘PLATFORM’ OF INTERACTING COMMUNITIES Where the modes of interaction and learning between communities are strongly heterogeneous, the coherence of the firm requires the formation of a common direction and a common vision that guides the heterogeneous actors and reconciles their contradictory interests. This common reference system or this common vision will be produced through the intra- and inter-community interaction in these specific contexts. We shall use the notion of ‘cognitive (or knowledge) platform’ as the generic term for designating a specific architecture of knowledge that articulates dispersed communities holding specialized pieces of (practised) knowledge.5 18.4.1

Different Types of Cognitive Platforms between Communities within the Firm

A ‘cognitive platform’ consists of at least one dimension of common knowledge shared by the different communities. As Grant (2002, p. 139) underlines, there are different types of common knowledge, ‘each of which is likely to fulfil a different role in permitting knowledge integration’: language and other forms of symbolic communication, shared meaning and representation, commonality of specialized knowledge, recognition of individual knowledge domains (extent to which an individual is aware of the know-how and other forms of knowledge possessed by the other individuals). Rules and directives, time-patterned sequencing, routines as ‘grammars of action’, or ‘ba’ as a concept of shared cognitive space are other categories of possible common knowledge. We can add that ‘corporate culture’ is another form of common knowledge that is essential for the coordination of actions and the creation of resources. This genuine ‘collective grammar’ ensures the connection between the various communities in the organization and the homogenization of their objectives.

420

Handbook of knowledge and economics

To fully describe the cognitive platform, in addition to the type of common knowledge that holds communities together, the nature of interactions between communities should also be defined. Are communities tightly or loosely coupled to the platform? In other words, to refer to the language of social network theory, could we speak of ‘strong’ or ‘weak’ interactions between communities? We propose to explore this issue through the following hypotheses: the structure of interactions between communities can be defined by two main factors: the repetitiveness of interactions between communities, and the quality of communication between communities, which could be assimilated to the cognitive distance between them. Indeed, the two phenomena have common features, but distinguishing them is important to clarify their different contexts of interaction. ●

●

The repetitiveness of interactions between communities expresses the ‘quantitative’ dimension of their relationships.6 Some communities may meet frequently (e.g. workers and managers using the same canteen), and this can generate some benefits for the firm (e.g. formation of a certain common knowledge, circulation of news that ‘something is not going well’), even though the intensity of communication between them is low (e.g. minimal common language or grammar to improve the circulation of knowledge between the communities). The repetition of interactions comes with an exchange of information and tacit knowledge, essential elements in the construction of trust (Lorenz, 1993). A high degree of repetition of interactions between knowing communities contributes to stimulating the processes of learning, creating favourable conditions for the resolution of conflicts, and encouraging the realization of economies of scale. Organizational devices, such as group projects or frequent meetings encouraging the socialization of experiences, are regularly introduced by the management to compensate for the lack of spontaneous interaction between heterogeneous communities. This enables us to better understand the importance given to the construction of privileged learning platforms by firms (‘ba’ in the sense of Nonaka and Konno, 1998). Frequent quantitative interactions between communities contribute to lower the cognitive distance between them, but do not guarantee in the long term the existence of a common grammar and codes between heterogeneous units. The quality of communication between communities expresses the ‘qualitative’ dimension of the relationships between them. Some communities can be joined together through a rich texture of

The firm as a ‘platform of communities’ 421 Table 18.1

Different types of platforms of interaction between communities within the firm Low repetitiveness of interaction between communities

High repetitiveness of interaction between communities

Low quality of communication between communities

Platform with weak interactions between communities (weak ties, strong cognitive distance)

Platform with moderate interactions I (Strong ties, strong cognitive distance) between communities

High quality of communication between communities

Platform with moderate interactions II (Weak ties, weak cognitive distance) between communities

Platform with strong interactions between communities (strong ties, weak cognitive distance)

communication, even if the quantitative ‘degree of repetition’ of interaction is low. Mintzberg (1979), for example, quotes the wellknown example of operations in hospitals, where the members of the different communities involved (surgeons, anaesthetists, nurses) meet infrequently, but when they do so, they know exactly what to do and how to work together (thanks to the possibility of communication provided during their respective training).7 Circulation of knowledge in an innovating firm is based essentially on the sharing of codes and languages, allowing various communities to interact.8 Thus it is a question of relational or cognitive proximity (Nooteboom, 2000) between distributed units, requiring attention to syntactic, semantic and pragmatic communication, shared tacit knowledge, flow and interpretation of information, and trust or other conventions of collaboration.9 As a result, the specific combinations of the two dimensions (repetitiveness of interactions and quality of communication) lead to four different organizational contexts corresponding to different types of platform of interactions between communities (Table 18.1). Thus the nature of the cognitive platform that brings heterogeneous communities together is one of the most critical features of the knowledgebased firm. Each firm is characterized by a cognitive platform that expresses the quality of the common knowledge shared by its different communities. The existence of a common knowledge base is essential to informal interaction: ‘Common knowledge not only helps a group coordinate, but also to some extent can create groups, collective identities’

422

Handbook of knowledge and economics

(Chwe, 2001, p. 11). However, at this point, to assess the key issue of the coherence of the firm, we have to investigate the question of the global architecture of the firm that requires a delicate matching between the formal hierarchical modes of coordination of the firm and the cognitive platform of interactions between informal communities. 18.4.2

Hierarchical Structures and Platforms of Communities

The issue of the coherence of the firm raises a critical question concerning the conceptualization of its knowledge governance. The perspectives considered in this chapter clearly suggest two main competing modes of knowledge management. The first mode is management by design, suggested by the strategic management approach to the firm (Prahalad and Hamel, 1990), which tends to focus on appropriate coordination mechanisms and issues of organizational design. This type of management is a renewed version of classical management principles based on managerial intervention, hierarchy, orders and blueprints. The appropriate managerial tools in a knowledge-based context may be different from those used in the classical case of transaction cost economies (where the firm is conceived of as a processor of information), but the intention is the same: the governance of the firm is conceptualized from a top-down vision of the firm, suggesting a ‘hard’ infrastructure of learning where managers use extrinsic incentives mechanisms (such as stock options) to align the knowledge activities of employees to the vision they seek to promote. The second mode is management by communities, suggested by the perspective that emphasizes learning by doing, of both an experimental and a path-dependent nature. This mode takes us in a different management direction, into the realm of how the practices of engagement/ enrolment/translation can be supported. This shift of focus springs from the interest in the process itself of how knowledge is formed and made explicit, something that the knowledge-based perspective tends to take for granted. The governance of the firm relies on a bottom-up vision of the firm, where managers, promoting a ‘soft’ infrastructure of learning, permanently ‘enact’ new forms of organizational devices suggested by the social dynamics of communities. The nature of incentive mechanisms used in such a context is essentially intrinsic, embodied in current practices. In this chapter, we suggest that the second mode of management captures the very essence of the governance of firms in a knowledge perspective. Communities, considered as a mode of coordination, can correct major ‘learning failures’ characteristic of hierarchical organizations, and can provide specific advantages in terms of coordination that cannot be

The firm as a ‘platform of communities’ 423 fulfilled by a hierarchical approach (‘management by design’). As Amin and Cohendet (2004, p. 113) put it: The first and obvious ‘management’ step implicit in a model of learning by doing is clear recognition of the limits of management by design, of the topdown inculcation of creativity. This is not an argument against organizing for innovation, simply an appreciation that the social dynamics of learning in communities cannot be engineered with the tools of hierarchical or transactional governance in the firm.

However, like any coordination problem, coordination through communities also faces certain risks of failure. We suggest that in practice circumventing these risks implies exploration of hybrid forms of management able to find complementarity between ‘management by design’ and ‘management by communities’. Such a complementarity requires that hierarchical structures vary with the nature of the cognitive platform of interactions between communities. In order to extract the potential benefits from intercommunity interactions, the role, the nature and the design of hierarchies should strongly differ according to the different organizational contexts. The implications can be highlighted by the following typology (which extends the above typology of cognitive platforms). 1.

2.

The first category (low repetitiveness of interactions between communities, low intensity of communication between communities), corresponds to the traditional sequential process mode of management (as in a typically Taylorist organization). The strong division of labour relies on specialized units that do not interact on a frequent basis, and do not develop rich modes of communication. The essence of coordination relies on intensive managerial coordination that establishes ex ante top-down rules and procedures to be followed by the entire organization, and centralizes the global vision of the product creation process. Classic incentive and coordination mechanisms such as Taylorist time-and-motion management principles drive decision making. Management by design clearly dominates management by communities, although local mechanisms of learning in communities (e.g. at shop-floor level) can transmit learning-by-doing effects at the global level of the organization. The second category (high repetitiveness of interactions, low intensity of communication between communities) corresponds to the overlapping problem solving mode (as in matrix types of organizations) that aims at bridging and cross-fertilizing through repeated informational exchanges between specialized groups in the organization. The

424

Handbook of knowledge and economics absence of a rich architecture of communication between the groups leads to an expensive search for a cognitive consensus between communities, and calls for active managerial involvement, mostly ex post, to solve disputes and conflicts between communities, but also designated to implement common knowledge and to coordinate beliefs while producing sense. The low intensity of communication between communities, especially noticeable in emergent relations built around many communities, can lead to an expensive search for cognitive alignment between communities. Coordination by leadership (necessarily conscious and intentional) appears to be the ideal solution in instances where the costs of communication or compatibility are onerous or where the resolution of coordination problems is urgent. Therefore a script of leadership emerges, charged with coordinating intricate actions or beliefs while producing sense. Foss (1999) has shown that, in some circumstances, leadership can offer less expensive solutions than complex mental processes or formation of conventions. It can, in particular, facilitate coordination of beliefs: Leadership is designed to co-ordinate the interlocking actions of many people through the creation of common knowledge . . . It is the ability to resolve social dilemmas, by influencing beliefs . . . Leaders may create a common knowledge when none exists previously. They also may solve coordination problems that persist even in the presence of common knowledge. (Foss, 1999, p. 4)

This type of situation also requires a specific coupling between management by design and management by communities (Muller, 2004). Part of the solution might reside in the hands of ‘middle management’, which plays, for authors such as Nonaka and Takeuchi, a decisive role in the innovative quality of the business. The middle managers can be seen as mediators who know the norms and habits of the communities sufficiently well to translate messages of the hierarchy into a language intelligible to different communities, and in turn, to translate the messages coming from communities for the hierarchy. 3.

The third category (low repetitiveness of interactions and high intensity of communication between communities) corresponds to the modular organization (organizational structures with existing cognitive platforms that allow loose-coupled systems to function efficiently). In such contexts, learning at the component level is insulated from disruptions by unexpected changes in product architecture during development projects. The modular platform expresses the

The firm as a ‘platform of communities’ 425 existence of a common cognitive architecture that links communities together (e.g. different communities of work in a hospital – nurses, surgeons, anaesthetists). The existence of such an infrastructure of knowledge (common grammar, common codes, common languages) may be due to very different historical factors (a type of education that has anticipated the cognitive forms of relationships between heterogeneous communities, shared experience that has lasted long enough to permit a common grammar to be built, a decision taken by the hierarchy to build a modular platform of knowledge etc.). But, whatever the reason, the common infrastructure of knowledge has taken time and sunk costs to be built. It not only defines what the communities have in common, but it also implicitly defines what they do not have in common.10 The role of hierarchy is to define ex ante the nature of the platform, and ex post to redefine the platform if radical innovations are unavoidable. Standardized interfaces between each community and the common platform of knowledge allow each community to work independently of others. This implies specific advantages, in particular the fact that, provided that the platform holds, the need for coordination by hierarchy is significantly reduced. In this case, management by communities temporarily dominates management by design. However, if the constraint of the interfaces cannot be respected, then the efficacy of the common platform becomes severely questioned. This could happen, for example, when emergent innovations in one community imply the reformulation of the whole cognitive platform. In such a context, sense-making interventions by the hierarchy may be needed to decide if the novelty produced requires reformulating the common platform. If so, a new cognitive process of definition of a common grammar, codes and language has to be initiated. In summary, the role of the hierarchy is to intervene at critical moments when the need to reformulate a common platform of knowledge between communities is perceived as essential. Category 3 is thus a case where management by design and management by communities sequentially alternate as dominant modes of coordination. 4.

In the fourth category (high repetitiveness of interactions and high intensity of communication between communities), we can envisage governance by community alone, with hierarchy needed only to ‘authorize’ or ‘enact’ the organizational forms produced by the interactive autonomous communities. The organization can be largely left to operate in a self-organizing manner. It is likely that, in such a situation, the unceasing efflorescence of communities allows the

426

Handbook of knowledge and economics

Table 18.2

Different types of focus of management corresponding to different organizational contexts of interaction between communities Low repetitiveness of interaction between communities

High repetitiveness of interaction between communities

Low intensity of communication between communities

Category 1 Focus of management: establishing a full control system (full design and specification decisions, and control decision)

Category 2 Focus of management: solving conflicts (partial design and specification, adjudication of disputes)

High intensity of communication between communities

Category 3 Focus of management: (re)designing the common knowledge platform

Category 4 Focus of management: enacting organizational forms that emerged from the auto-organized process

organization to innovate constantly since it does not disrupt corporate integrity (this dimension can be related to the creative spiral as conceived by Nonaka and Takeuchi, 1995). This mode could be called management by ‘enactment’, echoing the work done by Ciborra, 1996, who has described the knowledge platform at Olivetti in such terms. In such a context, where management by communities seems to clearly dominate management by design, the main role of the hierarchy is to select options (Foss, 2003) and enact the innovative outcomes produced by the constant interactions of communities. However, as Lazaric and Raybaut (2004) have shown, when the political and cognitive dimensions are dissonant and when conflicts persist, there could be a specific need for the reintroduction on a strong hierarchy even in this kind of distributed network to mitigate conflicts and align incentives. The four contexts of matching that have been detailed above are represented in the Table 18.2.

The firm as a ‘platform of communities’ 427

18.5 CONCLUSION: WHAT DOES THE ‘COMMUNITY’ APPROACH BRING TO THE THEORY OF THE FIRM? If we restrict the analysis to what Casson (1998) considers as the main issues to be addressed by a theory of the firm – the boundary of the firm; the formation, growth and diversification of the firm; the internal organization of the firm; and the role of the entrepreneur – it is clear that the ‘community approach’ alone is unable to provide a complete and satisfactory answer to all these critical issues. It is not per se (it has no ambition to be) a separate theory of the firm to be compared to the other well-known theories anchored in the literature. However, if we consider the many ways in which the community approach complements existing theories of the firm, the contribution is considerable. To assess this contribution, we shall first focus on the complementarity of the ‘community approach’ with the two other dominant perspectives on the knowledge based firm: the strategic management approach and the evolutionary economics approach. Then we shall investigate on the nature of complementarity with classical theories of the firm. The community approach complements the strategic management approach in two ways. First, as seen above, it contributes to its account of the governance of the knowledge-based firm by linking the hierarchical architecture of learning, which exists in many visible forms, with the invisible architecture of learning based on communities. This view clearly aims at integrating economic, organizational and sociological theories of governance in the firm in a cognitive perspective, in order to reconceptualize the governance forms ‘as mixes or configurations of simpler and potentially disentangleable components’ (Grandori, 1997, p. 29). What is suggested by the four types of organizational contexts we have identified is that the nature of the governance of the firm, and more precisely the relative intensity of interactions between ‘management by design’ and ‘management by communities’, depends strongly on the interrelational context of the organization. Second, it contributes to adding precision to the central notion of ‘competence’. For the strategic management approaches (Prahalad and Hamel, 1990; Stalk et al., 1992; Doz, 1996), the delimitation of the competence domain is essentially the privilege of the manager, who designs an ex ante vision of the management of knowledge within a firm. Managers are in charge of defining the frontier between the domain of competence and the domain of transaction. They thus endeavour to design specific incentives to align the behaviour of members of the firm to the vision of the firm they wish to promote. While for the community approach, as Wenger (1998)

428

Handbook of knowledge and economics

noted, a community drawing on interaction and participation to act, interpret and innovate, acts ‘as a locally negotiated regime of competence’. Communities could thus be seen as elementary units of competence that offer a microfoundation of this central notion.11 The ‘community’ approach also complements the evolutionary theory of the firm. The central concept of the evolutionary theory is the notion of routine. However, when dealing with this concept, the evolutionary theory proceeds as if the firm possesses the knowledge incorporated into the routines, and as if the competence results finally only from a selection of the best routines stored within the repertoire. It lacks an analysis of ‘intermediate links’ that are the genuine catalyst of the process of knowledge creation in the organization, where the creative ideas emerge or are tested, where the first validation of the innovation is carried out. It is precisely on this point that the taking into account of the concept of community (practised knowledge) can be justified. A related characteristic of communities is indeed that once knowledge is accumulated in the practices of a given community, the degree of inertia of routines and their power of replication of the routines kept by the community is much stronger than the power to replicate routines encapsulated in a given hierarchical organization. Routines ‘stick’ easily to a given community, while organizations must deploy considerable effort to deliberately replicate the knowledge contained in an organizational routine (Cohen et al., 1996). Exploring the interaction between routines and communities is a rich agenda for future research. For instance, one of the limitations of this chapter is that communities within the organization are supposed to be given. The representation of the firm as a set of overlapping heterogeneous communities poses the crucial problem of the evolutionary analysis of the spontaneous or intentional emergence of the cognitive platforms that have been considered.12 Vis-à-vis classical theories of the firm, in particular the transaction-based perspective, the ‘community approach’ highlights an important nuance concerning the design of the boundary of the firm. In the transactional vision that follows the Coasian perspective, the firm is viewed as a device that compensates for market failures. Thinking in terms of communities raises the possibility of designing governance structures that incorporate another fundamental mode of coordination (when market and hierarchies both fail, such as in the case of ‘learning’ failures). We have not addressed these possibilities in this chapter, but it is clearly another rich agenda for future research. In addition, we have generally assumed that communities are confined within the boundaries of the firm. However, this clearly does not hold in practice. Communities are open entities, and the interactions between members of a given community generally extend far beyond the

The firm as a ‘platform of communities’ 429 boundary of the firm to reach distant members, often belonging to different organizations, involving regular exchange of knowledge and best practice. For instance, the development of virtual communities away from the traditional organizations shows that the communities overflow the framework of the organizations. Through such everyday interactions, communities import useful and sometimes strategic knowledge from the outside world.13 This represents immense potential for innovation in the firm and contributes to explain Brown and Duguid’s (1991, p. 53) position when they propose that: large organisations, reflectively structured, are perhaps well positioned to be highly innovative and to deal with discontinuities. If their internal communities have a reasonable degree of autonomy and independence from the dominant worldview, large organisations might actually accelerate innovation.

To conclude, this chapter has argued that the concept of community can clarify a major aspect of the everyday coordination of knowledge creation. It has responded to the appeal by Crémer (1998), who has argued for an advanced theoretical analysis of the networks of non-hierarchical communication within the firm: A considerable amount of work is yet to be done on non-hierarchical communities in firms. In contrast with the theory of hierarchies, the research in this perspective should aim at a better understanding of the advantages and drawbacks of the different networks of communication. It should also aim at exploring their organizational consequences.

NOTES 1. 2.

3.

The ‘truce’ hypothesis in Nelson and Winter’s (1982) routine analysis. Economic and managerial literature offers multiple and sometimes conflicting definitions of ‘trust’. Initially, trust was considered by Arrow (1974) as a ‘lubricant of a social system’. Then, Kreps (1990) offered a game-theoretic model (where trust and reputation are confounded) that anchored the ‘calculative’ trust vision of economics. Kreps’s vision has been severely questioned by Williamson (1993). The latter offers a more subtle perspective where networks’ relations are based on convergent reciprocal interests. Williamson aimed mainly to defend the unity of the neoclassical model by eliminating the concept of ‘trust’ that threatens the fundamental duality of transactional economics in terms of market and organization. For Williamson, trust must be limited to personal and loving relations. In economic analysis, rather than a blind trust, it is the agents’ calculativeness that allows to determine costs and benefits of cooperation. ‘I submit that calculativeness is determinative throughout and that invoking trust merely muddies the (clear) waters of calculativeness’ (Williamson, 1993, p. 471). In situations of uncertainty, it is the institutional setting that strengthens mutual engagements (ibid., p. 469). According to him, the recourse to the concept ‘trust’ is then superfluous and misleading. The process of construction of trust is progressive and cumulative. Trust cannot be understood but as an incremental construction in a long-term perspective. Thus the

430

4. 5.

6.

7. 8.

9.

10.

Handbook of knowledge and economics dynamic of trust is evolutionary by nature. Dynamics of trust transform the zerosum game (in the Darwinian or Malthusian sense) into a non-zero-sum game (in the Lamarckian sense). As Szulanski et al. (2004), or Lazaric (2003) advocate, trust may be a prerequisite for transferring knowledge or initiating a codification process. Purvis et al. (2001) refer to knowledge platforms as ‘organisational architectures for integrating and combining in a dynamic way specialised knowledge bases dispersed across firms and generated in the course of practice and reflection by individuals or firms working on specific projects, products, processes and disciplines in different locations at different times’. We have simplified the analysis of the quantitative dimension of the interaction between communities by focusing only on one characteristic (the frequency of interaction). Bogenrieder and Nooteboom (2004, p. 291) have suggested a richer definition based on four characteristics (one of them being the frequency of interaction): ‘In our view, the “strength of ties” has four aspects. One aspect is intensity, which refers to the effort and commitment of resources involved, and to the scope of activities taken up in the tie (share of total activities). The resources that are committed are not necessarily only resources of money, time, or effort, and may also include psychological resources (commitment, loyalty, fairness, and empathy). A second aspect is frequency of interaction, a third is openness of communication, and a fourth is duration of ties. Strong ties yield shared experience, which reduces cognitive distance. Durable ties enable the development of empathy and identification (McAllister, 1995; Lewicki and Bunker, 1996; Hansen, 1999) as a basis for trust.’ It is worth noticing that, in this case, the type of relationship between members of the different communities involved in a given operation is based on competence trust. Nohria and Ghoshal (1997, p. 87) argue, for example, that in a decentred organization ‘the real leverage lies in creating a shared context and common purpose and in enhancing the communication densities within and across the organization’s internal and external boundaries’. Interestingly, though, here too it is the soft infrastructure that Nohria and Ghoshal highlight, choosing to emphasize the role of socialization (e.g. via corporate encounters, conferences, recreational clubs), normative integration (e.g. via incentives such as access to health care or travel concessions, company rituals, inculcation of corporate or brand standards), and effective communication between self-governing units (e.g. via both Internet and relational or cognitive proximity. Different people have a greater or lesser ‘cognitive distance’ between them (Nooteboom, 1992, 1999a). A large cognitive distance has the merit of novelty, but the problem of incomprehensibility. In view of this, organizations need to reduce cognitive distance, that is, achieve a sufficient alignment of mental categories to understand each other, utilize complementary capabilities and achieve a common goal. This yields the notion of organization as a ‘focusing device’. ‘Since mental categories have developed on the basis of interaction with others, in a sequence of contexts that make up experience, there will be “cognitive distance” between people with different experiences, and cognitive similarity to the extent that people have interacted within a shared experience (though we do not wish to imply that “cognitive distance” allows for any simple, one-dimensional scale). Cognitive distance yields both a problem and an opportunity. The opportunity is that we learn from others only when they see and know things differently. In the absence of claims of objective knowledge, interaction with others is the only path we have to correct our errors. The problem is that people may not understand each other and have to invest in understanding’ (Bogenrieder and Nooteboom, 2004, p. 298). This can be related to well-known developments suggested by Bourdieu on the fact that more important than the notion of common knowledge is the notion of acceptance by one community of what we ‘do not want to know’ (about what the other community is doing). For economists, this suggests a radical reconsideration of the way to perceive the notion of asymmetries of information.

The firm as a ‘platform of communities’ 431 11.

As an example of the need to integrate the different approaches of the knowledge-based firm, one may consider the decision of a big firm to acquire a small start-up company for reinforcing its core competence. The interpretation of the decision as essentially motivated by the reinforcement of the strategic cognitive base of the firm has certainly to do with the representation of top managers of the big firm, and belongs to the strategic management approach. However, on practical grounds, the target of the decision to acquire could be the fact that the start-up company owns a specific patent in the core domain of the big company. But, it may happen that, once the acquisition is made, the absorption of knowledge from the small start-up is more than difficult and that the patent may not be operating because the active communities that constituted the nucleus of the start-up have decided to leave and not to join the big company. In that case, the need to think in terms of communities to understand the formation of competences in a dynamic context is critical. The different approaches are thus essentially complementary. 12. Such a representation encounters a major difficulty if we adopt a dynamic perspective. Indeed, communities can change deeply over time. Some communities may emerge or disappear; others change or become institutionalized. In our perspective, to simplify the analysis, we shall suppose that communities are given (their norms, their process of learning, their mechanisms of recruitment etc. will be assumed to be constant). This will allow us to focus on the mechanisms of intra-community coherence. In other terms, this assumption means that the speed of evolution of each  community is slow. It means also that the community becomes the unit of analysis. 13. Of course, a related problem is that the functioning of communities is exposed to the risk of disclosing to the outside world elements of knowledge that the hierarchy does not intend to disclose. This will certainly be a growing phenomenon. We have always supposed in this chapter that the considered communities were not only inside the organization, but also ‘remained faithful’ to the organization.

REFERENCES Adler, Paul S. (2001), ‘Market, hierarchy and trust: the knowledge economy and the future of capitalism’, Organization Science, 12 (2), 214–34. Amin, A. and Cohendet, P. (2004), Architectures of Knowledge: Firms, Capabilities and Communities, Oxford: Oxford University Press. Arena, R., Lazaric, N. and Lorenz, E. (2005), ‘Trust, codification and epistemic communities: implementing an expert system in the French steel industry’, Introduction to the Handbook on Trust, mimeo CNRS UNSA, France: Valbonne. Arrow, K. (1974), The Limits of Organizations, London and New York: Norton and Company. Bogenrieder, I. and Nooteboom, B. (2004), ‘Learning groups: what types are there? A theoretical analysis and an empirical study in a consultant firm’, Organization Studies, 25, 287–313. Bourdieu, P. (1977), Outline of a Theory of Practice, Cambridge: Cambridge University Press. Bowles, S. and Gintis, H. (2000), ‘Social capital and community governance’, Working Paper 01-01-003, Santa Fe Institute. www.santafe.edu/sfi/publications/workingpapers/01-01-003. pdf. Brousseau, E. (2001), ‘E-économie: qu’y a-t-il de nouveau?’, Annuaire des relations internationales, Bruxelles: Émile Bruylant. Brown, J.S. and Duguid, P. (1991), ‘Organizational learning and communities of practice: toward a unified view of working, learning and innovation’, Organization Science, 2 (1), 40–57.

432

Handbook of knowledge and economics

Casson, Mark C. and Wadeson, Nigel S. (1998), ‘Communication costs and the boundaries of the firm’, International Journal of the Economics of Business, 5 (1), 5–28. Choo, C.W. and Bontis, Nick (2002), The Strategic Management of Intellectual Capital and Organizational Knowledge, New York: Oxford University Press. Chwe, M.S.Y. (2001), Rational Ritual: Culture, Coordination, and Common Knowledge, Princeton, NJ: Princeton University Press. Ciborra, C.U. (1996), ‘The platform organization: recombining strategies, structures, and surprises’, Organization Science, 7 (2), 103–18. Cohen, M.D., Burkhart, R., Dosi, G., Egidi, M., Marengo, L., Warglien, M. and Winter, S. (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5 (3), 653–98. Cohendet, P. and Llerena, P. (2003), ‘Routines and communities in the theory of the firm’, Industrial and Corporate Change, 112 (3), 271–97. Cohendet, P., Créplet, F. and Dupouët, O. (2000), ‘Organisational innovation, communities of practice and epistemic communities: the case of Linux’, in Alan Kirman and Jean-Benoît Zimmerman (eds), Economics with Hetrogeneous Interacting Agents, Berlin: Springer, pp. 303–26. Cook, S.D.N. and Brown, J.S. (1999), ‘Bridging epistemologies: the generative dance between organizational knowledge and organizational knowing’, Organization Science, 10 (4 ), 381–400. Coriat, B. and Guennif, S. (1996), ‘Incertitude, confiance et institution, Communication’, Colloque ‘La confiance en question’, 22–23 March, Aix-en-Provence. Cowan, R. and Jonard, N. (2001), ‘The workings of scientific communities’, Research Memorandum 030, MERIT, Maastricht Economic Research. Cowan, R., David, P.A. and Foray, D. (2000), ‘The explicit economics of knowledge codification and tacitness’, Industrial and Corporate Change, 9 (2), 211–53. Crémer, J. (1998), ‘Information dans la théorie des organisations’, Working Paper, Institut d’Économie industrielle, Université de Toulouse. Dibiaggio, L. (1998), Information, Connaissance et Organisation, PhD dissertation, Université de Nice–Sophia Antipolis. Doz, Y. (1996), ‘The evolution of cooperation in strategic alliances: initial conditions or learning processes?’, Strategic Management Journal, 17, 55–83. Dupouët, O. and Laguecir, A. (2001), ‘Elements for a new approach of knowledge codification’, ETIC final conference, Strasbourg. Foss, N. (1999), ‘Understanding leadership: a coordination theory’, Working Paper 99-3, DRUID. Foss, N. (2003), ‘Selective intervention and internal hybrids: interpretating and learning from the rise and decline of the Oticon Spaghetti Organization’, Organization Science, 14 (3), 331–49. Fransman, M. (1994), ‘Information, knowledge, vision and theories of the firm’, Industrial and Corporate Change, 3 (2), 1–45. Fukuyama, F. (1995), Trust: The Social Virtues and the Creation of Prosperity, New York: Free Press. Gensollen, M. (2001), ‘Internet: Marchés électroniques ou réseaux commerciaux?’, Revue Économique, 52, 137–64. Grandori, A. (1997), ‘Governance structures, coordination mechanisms and cognitive models’, Journal of Management and Governance, 1 (1), 29–47. Grant, R.M. (2002), ‘The knowledge-based views of the firm’, in C.W. Choo and N. Bontis (eds), The Strategic Management of Intellectual Capital and Organizational Knowledge, Oxford: Oxford University Press, pp. 133–48. Knorr-Cetina, K. (1999), Epistemic Cultures, Boston, MA: Harvard University Press. Kogut, B. and Zander, U. (1992), ‘Knowledge of the firm, combinative capabilities, and the replication of technology’, Organization Science, 3, 383–97. Kogut, B. and Zander, U. (1996), ‘What firms do: coordination, identity, and learning’, Organization Science, 7 (5), 502–18.

The firm as a ‘platform of communities’ 433 Kreps, David M. (1990), ‘Corporate culture and economic theory’, in James E. Alt and Kenneth A. Shepsle (eds), Perspectives on Positive Political Economy, New York: Cambridge University Press, pp. 90–143. Langlois, R. and Foss, N. (1996), ‘Capabilities and governance: the rebirth of production in the theory of economic organization’, Kyklos, 52 (2), 201–18. Lave, J. and Wenger, E. (1991), Situated Learning: Legitimate Peripheral Participation, Cambridge: Cambridge University Press. Lazaric, N. (2003), ‘Trust building inside the “epistemic community”: an investigation with an empirical case study’, in F. Six and B. Nooteboom (eds), The Trust Process in Organizations, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 147–67. Lazaric, N. and Raybaut, A. (2004), ‘Knowledge creation facing hierarchy: the dynamics of groups inside the firm’, Journal of Artificial Societies and Social Simulation, www://jasss. soc.surrey.ac.uk/7/2/3/html. Lerner, J. and Tirole, J. (2001), ‘The open source movement: key questions’, European Economic Review, 45, 816–26. Liedtka, J. (1999), ‘Linking competitive advantage with communities of practice’, Journal of Management Inquiry, 8, 5–16. Lorenz, E.H. (1993), ‘Flexible production systems and the construction of trust’, Politics & Society, 21 (3), 304–21. Marschak, J. (1954), ‘Towards an economic theory of organization and information’, in R.M. Thrall, C.H. Coombs and R.L. Davis (eds), Decision Processes, New York: John Wiley & Sons, pp. 187–220. Mintzberg, H. (1979), The Structuring of Organizations, Englewood Cliffs, NJ: Prentice-Hall. Muller, P. (2004), ‘Autorité et gouvernance des communautés intensives en connaissance: une application au développement du logiciel libre’, Revue d’Economie Industrielle, 106, 49–68. Nelson, R.R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Nohria, N. and Ghoshal, S. (1997), The Differentiated Network: Organizations’ Knowledge Flows in Multinational Corporations, San Francisco, CA: Jossey-Bass. Nonaka, I. and Konno, N. (1998), ‘The concept of “Ba”: building a foundation for knowledge creation’, California Management Review, 40 (3), 40–54. Nonaka, I. and Takeuchi, H. (1995), The Knowledge-Creating Company: How the Japanese Companies Create the Dynamic of Innovation, Oxford: Oxford University Press. Nooteboom, B. (1992), ‘Towards a dynamic theory of transactions’, Open Access publications from Tilburg University, no. 12-372855. Nooteboom, B. (1999a), ‘Innovation and inter-firm linkages: new implications for policy’, Research Policy, 28, 793–805. Nooteboom, B. (1999b), Inter-firm alliances; Analysis and design, London: Routledge. Nooteboom, B. (2000), ‘Learning by interaction: absorptive capacity, cognitive distance and governance’, Journal of Management and Governance, (1–2), 69–92. Nooteboom, B. (2002), Trust: Forms, Foundations, Failures and Figures, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Prahalad, C.K. and Hamel, G. (1990), ‘The core competence of the corporation’, Harvard Business Review, 68 (3), 79–91. Purvis, R., Sambamurthy, V. and Zmud, R. (2001), ‘The assimilation of knowledge platforms in organizations: an empirical investigation’, Organization Science, 12 (2), 117–35. Rocco, E., Finholt, T.A. and Hofer, Erik C. (2000), ‘Out of sight, short of trust’, Working Paper, School of Information, University of Michigan, www.crew.umich.edu/ publications/01-10.pdf. Stalk, G., Evans, P. and Schulman, L.E. (1992), ‘Competing on capabilities: the new rules of corporate strategy’, Harvard Business Review, 70 (March–April), 57–69. Szulanski, G., Cappetta, R. and Jensen, J.R. (2004), ‘When and how trustworthiness matters:

434

Handbook of knowledge and economics

knowledge transfer and the moderating effect of causal ambiguity’, Organization Science, 15 (5), 600–613. Von Krogh, G., Ichigo, K. and Nonaka, I. (1998), ‘Knowledge enablers’, in G. von Krogh, J. Roos and D. Kleine (eds), Knowing in Firms: Understanding, Managing and Measuring Knowledge, London: Sage. Weigelt, K. and Camerer, C. (1988), ‘Reputation and corporate strategy: a review of recent theory and applications’, Strategic Management Journal, 9 (5), 443–54. Wenger, E. (1998a), ‘Communities of practice: learning as a social system’, Systems Thinker, 9 (5), 2–3. Wenger, E. (1998b), Communities of Practice: Learning, Meaning and Identity, Cambridge: Cambridge University Press. Williamson, O. (1993), ‘Calculativeness, trust, and economic organization’, Journal of Law and Economics, 36 (1), 453–86.

19 The architecture and management of knowledge in organizations Mie Augier and Thorbjørn Knudsen

19.1 INTRODUCTION In recent years, the rise of the knowledge economy has created new challenges for strategic management and made managing intellectual capital an integral part of firm strategy (Teece, 1998), thus making the creation, development and capturing of value from knowledge and competencies a critical issue. This development has led to a burst of attention to knowledge assets in the management, organization and strategy literatures. The rise of competencies and capabilities approaches to firm organization the last decade has been linked to the knowledge economy and the increasing importance of innovation, rapid technological change and knowledge assets, among other things (Teece, 1998; Eisenhardt and Martin, 2000; Teece, 2000). It has been emphasized that the knowledge assets that are significant for firm innovation and growth are the individual knowledge, skills and expertise that are embedded in the firm’s physical and social structure; this helps knowledge to be shaped into competencies (Teece, 1998). As a result, organizational competencies can be seen as reflecting a need to coordinate patterns of complex behavior with interactions on many levels of the firm.1 Among other things, the dynamic nature of organizational competencies has been used to address questions relating to the design of internal organization and the boundaries of firms (Dosi, 1988). In particular, the boundaries of business firms can be understood in terms of learning, path dependencies, technological opportunities, selection environment and the firm’s position of complementary assets (Dosi, 1988). Because firms face strategic decisions on the basis of their past history (Chandler, 1962; Simon, 1993), it is natural to view questions relating to the development of strategy and competences as issues of adaptation that take place in an evolutionary setting. Several contributors have recently offered insights into and evidence on how firms can develop their capability to adapt and capitalize on rapidly changing environments (Prahalad and Hamel, 1990; Chandler, 1990; Teece, 1993). Successful companies must be able to respond quickly 435

436

Handbook of knowledge and economics

and adapt to rapid (technological as well as social) changes in the face of individual as well as organizational ambiguity (March, 1978). Survival and innovation in a rapidly changing environment require that firms are able to coordinate and deploy external and internal competencies and are able to learn from the experiences of their past successes (and failures) as well as the successes (and failures) of other firms (Levinthal and March, 1993). Business strategy, in this set-up, involves skills in anticipating the shape of an uncertain future; skills in generating strategic alternatives for the firm; and skills in implementing new projects and plans fast and effectively (Simon, 1993). This involves both an entrepreneurial and a strategic element, combining the search for new strategic opportunities and the ability to better exploit them (March, 1991; Teece, 1998). Yet little systematic thinking has been offered about the ways in which organizations choose and adapt their structure in order to better meet the need to evaluate information within the context of lower communication costs and increased connectivity of various media, phenomena that are commonly associated with the knowledge economy (Collins and Chow, 1998). The new challenge that must be addressed is that both good and bad alternatives travel faster in the knowledge economy (Collins and Chow, 1998). Little advice has been offered in the way of designing organizations that meet this challenge. A possible source of this omission is that previous research bearing on knowledge management seems to have run into conceptual difficulties (Romanelli, 1991; Child and McGrath, 2001). We believe that the lack of useful models of knowledge organizations lies at the heart of these difficulties. In order to fill this gap we draw on previous research in economics, management studies and strategy to think about the organization of knowledge. The purpose of this chapter is to meet the challenge of modeling knowledge organization by introducing a new, unifying, way of thinking about the organization of knowledge. The organization of knowledge is here thought of as an architecture that may help boundedly rational agents make better choices. This logic was introduced by Simon (Simon, 1947) and later elaborated, especially in March and Simon (1958). ‘Human rationality’, he wrote, ‘gets its higher goals and integrations from the institutional settings in which it operates and by which it is molded . . . [Therefore] . . . [t]he rational individual is, and must be, an organized and institutionalized individual’ (Simon, 1947, pp. 101–2). Simon (and March) argued that organizations make it possible to make decisions by virtue of the fact that they constrain the set of alternatives to be considered and the considerations that are to be treated as relevant. Organizations can be improved by improving the ways in which those limits are defined and imposed.

The architecture and management of knowledge in organizations 437 While other dimensions of knowledge deserve attention, we believe this is a useful place to begin modeling knowledge organizations. We are here concerned with modeling the business organization as an entity whose knowledge is an emergent property of the individual members’ cognitive skills. In order to focus on this issue we do not consider other important aspects of knowledge organization such as incentives, power, rules, and the possible tacit and codified aspects of knowledge. We do not imply that these and other additional issues are unimportant, but we believe that they are best considered in separate works. The chapter is organized as follows: Section 19.2 sets the stage for our argument by introducing insights from the behavioral theory of the firm (bounded rationality, imperfect environmental matching and organizational architecture). Section 19.3 introduces the modeling framework. The central idea is to represent the organization of knowledge as an architecture, which is a structure that defines the flow of information among members with limited levels of cognitive skill. Section 19.4 draws on the behavioral theory of the firm in order to consider the decomposability of architectures as a way to enhance adaptation. Section 19.5 concludes.

19.2 THE CONTEXT: A BEHAVIORAL THEORY OF THE FIRM A Behavioral Theory of the Firm is built around a political conception of organizational goals, a bounded rationality conception of expectations, an adaptive conception of rules and aspirations, and a set of ideas about how the interactions among these factors affect decisions in a firm (Cyert and March, 1992, ch. 9). 19.2.1 A Political Conception of Organizational Goals Whereas goals in rational choice theory are pictured as given alternatives each with a set of consequences attached, and the problem of choice consisting in the selection of the best alternative, goals within the behavioral theory of the firm are pictured as reflecting the demands of a political coalition, changing as the composition of that coalition changes. Goals reflect several dimensions (such as the goals of the organization and the presence of particular problems) and aspirations with respect to each dimension of the goals are pictured as changing in response to the experiences of the organization and its components as well as the experiences of others to whom they compare themselves. Thus it is the dynamic nature of aspirations that enables the generation of new decision alternatives.

438

Handbook of knowledge and economics

‘Alternatives are not given but must be searched for’, Simon (1955, p. 33) wrote. The firm, therefore, must engage in active search and imagination to create sustainable strategic opportunities (Winter, 2000). 19.2.2 A Bounded Rationality Conception of Expectations In the behavioral view, agents have only limited rationality, meaning that behavior in organizations is intendedly rational – neither emotive nor aimless. ‘Organizations are formed with the intention and design of accomplishing goals; and the people who work in organizations believe, at least part of the time, that they are striving towards these same goals’ (Simon, 1955, p. 30). Since firms are seen as heterogeneous, boundedly rational entities that have to search for relevant information, expectations in the behavioral theory of the firm are the result of making inferences from available information, involving both the processes by which information is made available and the processes of drawing inferences. Much information is gathered by search activity. The intensity of search depends on the performance of the organization relative to aspirations and the amount of organizational slack (March and Simon, 1958, pp. 47–52). The direction of search is affected by the location (in the organization) of search activity and the definition of the problem stimulating the activity. Thus the search activity of the organization both furthers the generation of new alternative strategies and facilitates the anticipation of uncertain futures. 19.2.3 An Adaptive Conception of Rules and Aspirations ‘Decision making’ in the behavioral theory is assumed to take place in response to a problem, through the use of standard operating procedures and other routines, as also through search for an alternative that is acceptable from the point of view of current aspiration levels for evoked goals. Decision making is affected, therefore, by the definition of the problem, by existing rules (which reflect past learning by the organization), by the order in which alternatives are considered (which reflects the location of decision making in the organization and past experience), and by anything that affects aspirations and attention (Cyert and March, 1992, ch. 9). Within this framework, four concepts were developed (Cyert and March, 1963). The first is the ‘quasi-resolution of conflict’, the idea that firms function with considerable latent conflict of interests but do not necessarily resolve that conflict explicitly. The second concept is ‘uncertainty avoidance’. Although firms try to anticipate an unpredictable future in so far as they can, they also try to restructure their worlds in order to minimize their dependence on anticipation of the highly uncertain future. The third

The architecture and management of knowledge in organizations 439 concept is ‘problemistic search’ – the idea that search within a firm is stimulated primarily by problems and directed to solving those problems. The fourth concept is ‘organizational learning’. The theory assumes that firms learn from their own experiences and the experiences of others (Levinthal and March, 1993). Because of the focus on informational inefficiencies, rules and bounded rationality, the behavioral theory of the firm is well suited as a basis for accommodating issues of knowledge and change in organizations. It also provides a conceptual framework for thinking about how to design an organization. As emphasized by Simon, people make mistakes, and they end up with inefficient governance structures and bad entrepreneurial judgments. Simon’s particular view of organizational knowledge and design is part of his broader view of social science in general, and models in the social sciences in particular. Simon was a man of science who dreamed of a better world with scientific models, and he first entered the domain of social science with the ambition of spreading the use of mathematics, thinking that the fields of social science needed a little ‘stiffening up’. At the same time, he thought, those models should correspond to the empirical realities of the real world. For instance, to Simon, mathematics was a language that could add considerably to the social sciences if it was empirically sound, but if it was not empirically sound, it didn’t matter that it was good mathematics, for it was not enough that it was logically consistent. His commitment to models was present throughout his work, but he wanted models that could provide concrete expressions about human behavior. For example, in much of his work from the early 1950s, collected in the papers in Models of Man (Simon, 1957), Simon used different mathematical models such as matrix theory and partial differential equations. In the course of using these mathematical tools, Simon found that they were somewhat limited because they were not flexible enough to accommodate the richness and the complexity that he saw as inherent in the decision-making process. And it was this richness that he wanted to model. He turned to computer language because it proved useful for expressing the information-processing activity that he saw as inherent in decision making. In the remainder of this chapter, we build on the core of the above insights from the behavioral theory of the firm in order to develop a model of knowledge organization. We assume that decision makers are boundedly rational. Mistakes get made even if individuals are well intentioned. Because organizations are built of boundedly rational agents, there is an imperfect matching between the business environment and the organization’s conception of it. For this reason, mistakes get made at the organizational level even in the absence of agency problems. Within a context defined by these assumptions, Section 19.3 considers the design

440

Handbook of knowledge and economics

of organizations that can help their members make better choices and Section 19.4 is concerned with the possibility of adapting organizations by enhancing the individual members’ cognitive skills.

19.3 THE CONTENT: ORGANIZATIONS AS ARCHITECTURES We draw on the behavioral theory of the firm to represent the knowledge organization in a simple and stylized way as an architecture. Such an architecture defines the cognitive skills of organization members as a screening function. The individual agents are boundedly rational, a condition reflected in their ability to screen alternatives. Because rationality is bounded, mistakes get made even if individuals are well intentioned, as mentioned above. We model the organization as a knowledge structure, an architecture, reflecting the way information flows among organization members. Our modeling approach is based on Christensen and Knudsen’s (2003a, 2003b) extension of Sah and Stiglitz (1986, 1988).2 This extension also provides a revision of Moore and Shannon’s (1956) work on electrical circuits that served as an inspiration for Sah and Stiglitz (1986, 1988). The main revision necessary to fit a realistic description of social organizations is that decision rights can be delegated in social organizations. In our model, social agents are endowed with powers that are much greater than the powers of relays or switches in an electrical circuit. Electrical circuits are often made of simple unsophisticated agents (Huberman and Hogg, 1995), such as relays, that quickly evaluate a binary distribution of zeros and ones. Social systems, by contrast, are built out of more sophisticated agents, such as product managers, that evaluate complicated issues such as the distribution of product quality in a new market. The degree of sophistication of agents can be described in terms of the decision rights that are delegated to the agent. More sophisticated agents can make more far-reaching choices than simple, unsophisticated agents. Our agents are sophisticated because it is within their power to make decisions on behalf of the entire organization. For this reason, decision rights can be delegated. But this possibility comes with a cost. Our agents are likely to make errors, both of commission and of omission. New product designs that reduce income are accepted and implemented whereas product designs that increase income are dumped. Even researchers that are supposedly very sophisticated evaluators make these errors. For example, editors and reviewers of academic journals accept mediocre articles, whereas very good articles are commonly rejected, at least in economics (Gans and Shepherd, 1994).

The architecture and management of knowledge in organizations 441 In our model, the agent’s competence in evaluating alternatives is modeled by imputing a screening function to this agent. An organizationlevel screening function represents the knowledge structure of the organization, the organization’s beliefs about the alternatives it is facing. That is, we think of a knowledge structure as an evaluation structure. This representation of a knowledge structure only captures a subset of the many meanings that can be associated with the concept of knowledge. In particular, our representation of the evaluation aspect of knowledge structures apparently omits consideration of alternative generation and search. This omission is only apparent. We develop a very flexible modeling framework that uses the previous literature on economic architecture as our point of departure. We could, in principle, model both alternative search and alternative generation, but leave this topic for subsequent work.3 It is further important to note that our model excludes issues relating to incentive structures. We find that incentive structures and knowledge structures are best treated as distinct topics even if they are intimately related. We use Sah and Stiglitz (1986, 1988) as our particular starting point in representing the knowledge structure of an organization. In their work, all members of a social organization have the same level of competence in evaluation of alternatives. This implies that the agents have a common cognitive framework. This assumption is useful when we want to assess the properties of alternative evaluation structures (by controlling for differences in competence), but it is in no way essential for the modeling framework we propose here. The very flexible modeling framework we develop can readily be used to capture organizations in which the cognitive framework is heterogeneous and the competence distributed. As emphasized by Cyert and March (1963), there is an imperfect matching between real organizations and their business environment. Also at the organizational level mistakes get made even in the absence of agency problems. But sometimes the organization can help individual agents make better decisions (Simon, 1947; March and Simon, 1958). From a modeling perspective, the challenge is to derive the organization-level screening function on the basis of the individual members’ screening function and the way a particular architecture structures information flows. When this challenge is met, useful insights can be gained regarding the knowledge properties of alternative forms of organizing. Yet further consideration is necessary to address issues regarding the dynamics of knowledge organization. 19.3.1

The Architecture of Organizational Forms

We begin by introducing the building blocks that allow us to represent the organization of knowledge as an architecture. The starting point is to

442

Handbook of knowledge and economics

represent the agent as a fallible decision maker. Mistakes get made when agents evaluate the quality of alternatives. The source of these mistakes is the uncertainty experienced by an agent. Uncertainty is added because of the finite processing capabilities of the agent. This is an expression of limits as a characteristic of the decision maker, to be understood in any possible way, ranging from the failure to obtain, assimilate, interpret and evaluate information relevant to an alternative, to a state of profound uncertainty (Simon, 1955). The agent can be engaged in decision making individually or as a member of an organization. This organization is a decision-making structure, referred to as an architecture. Following the ideas and terminology of Sah and Stiglitz (1986), the basic process of making a decision is loosely as follows. A decision-making structure will (repeatedly) be confronted with an alternative drawn from an initial portfolio I, representing the currently available alternatives (and thus the current state of the business environment).4 The alternative enters the structure through one of its agents and traverses the structure until it is ultimately either rejected or accepted. Rejection means that nothing is altered; the alternative is terminated and dumped in a waste bin T. In this case, there is no direct economic consequence for the organization. Acceptance means that the organization realizes the alternative, which, according to its quality, creates economic value. In this case, there is a direct economic consequence. This is symbolized by storing the alternative in a final portfolio F. In both cases, a cost is paid for making the decision. If an alternative capable of producing any income is rejected, then the organization made an error, denoted a Type-I error. If, on the other hand, the organization accepts an alternative producing a negative income, it is said to have made a Type-II error. In both of these cases, the decision was a failure, and in all other cases it was a success. The ultimate fate of the alternative depends on how the agents are interconnected, thereby motivating the study of different organizational architectures. The two basic architectures, the two-member hierarchy and polyarchy, are the simplest structures. Figure 19.1 provides a static overview of a graph representing the hierarchy of Sah and Stiglitz, and its environment, the initial portfolio I and the final portfolio F in which alternatives accepted by the hierarchy are stored. The termination node is denoted T, the waste bin where the rejected alternatives are dumped. The hierarchy represents a serial processing as an alternative traverses the structure. It is straightforward to generalize it to n-member hierarchies simply by adding nodes to the sequence between I and F. In contrast to the simple hierarchy, the polyarchy, shown in Figure 19.2, has a more specialized behavior and is more difficult to generalize in an unambiguous way (Christensen and Knudsen, 2003a). Additional rules

The architecture and management of knowledge in organizations 443

T

I

F

Notes: In order for the architecture to accept an alternative, both agents must accept the alternative individually. Or, the other way around, each individual has the ability to terminate the alternative. The solid lines indicate acceptance; the dashed lines indicate rejection.

Figure 19.1

The two-member hierarchy

F

I

T

Notes: Here, the alternative is sent to any of the agents at random. The individual can accept the alternative on behalf of the organization, but in order to get an overall rejection, both agents must reject. This behavior is partly implemented in the dynamic rules that constrain the traversal. The solid lines indicate acceptance; the dashed lines indicate rejection.

Figure 19.2

The two-member polyarchy of Sah and Stiglitz

of traversal, the dynamics of the architecture, must be supplied to ensure the polyarchy behavior. The basic modeling choice here is how polyarchy members dispatch alternatives to other members, and whether they evaluate an alternative more than once. A straightforward choice is a structure where polyarchy members dispatch alternatives to the nearest neighbor and evaluate each alternative only once. In this case, the polyarchy models a market that resembles a sandwich stall. Other markets can be modeled by altering this specification.

444

Handbook of knowledge and economics

Sah and Stiglitz further defined a unifying substructure called a committee of n members and consensus k. This substructure was constructed by picking a polyarchical structure and supplying a dynamic rule according to which the organization should accept only if k or more agents evaluate the project positively (e.g. k > n/2 represents a simple majority rule). The invention of the n-member committee of consensus k was both a generalization of the two basic structures and a unification of the two into a common framework. We define an organizational form as an n-member architecture: a collection of organizational members, a collection of channels through which the members can pass information or control to each other, and a set of dynamic rules that help define the flow of information or control and thus help define their decision rights. The purpose of an organization as considered here is to make decisions regarding whether to accept or reject alternatives, and these decisions have economic consequences. Decision rights are introduced as the right of an individual agent to ultimately accept an alternative on behalf of the architecture. Note that an agent can be an individual organization member or a collection of organization members. We can now outline a typology of organizational forms. Beginning from the n-member hierarchy, note that only one member has the right to individually accept an alternative. By contrast, in the n-member polyarchy every member has the right to individually accept an alternative. Thus the minimal delegation of decision rights occurs in the hierarchy, and the maximal delegation occurs in the polyarchy (market). The hierarchy and the polyarchy (market) are the limiting structures, as the delegation of decision rights in hybrids lies in between these extremes. A hybrid is a combination of a hierarchy and polyarchy. An external hybrid is a polyarchy that includes at least one hierarchical element. As more hierarchical elements are included, there is less delegation of decision rights and in the limit the hierarchy is obtained. What about the internal hybrid then? In order to maintain a hierarchical organization and yet delegate the right to accept projects on behalf of the organization, the internal hybrids are best viewed as nested structures. In this case, the components of the hierarchy, the nodes, are polyarchical substructures including more than one agent. The most straightforward internal hybrid is therefore a hierarchy where the last node, leading to ultimate acceptance, is an n-member committee of consensus k. In this case, decision rights are delegated; acceptance is possible even if some committee members reject. Possibly, the other nodes in the hierarchy are substructures that include more than one organization member, but this is of little importance because only the highest level in the hierarchy has the right to ultimately accept alternatives on behalf of the organization.

The architecture and management of knowledge in organizations 445 In summary, a typology of organizational forms has been defined. This typology, shown in the appendix, includes the four classes of organizational form, ranked according to the delegation of decision rights. The hierarchy has the minimal delegation, and then follows the internal hybrid, the external hybrid, and finally the polyarchy (market). It is further possible to rank the wealth of external and internal hybrids according to the delegation of decision rights. This typology should be useful for a number of purposes. In the context of the present chapter, our immediate concern is to provide a useful way of thinking about knowledge organizations and, as shown below, it turns out that hybrids have important properties. We have proposed that an organization-level screening function represents the knowledge structure of the organization. We have also defined a typology of architectures that includes the broad classes of organization mentioned in the recent literature on new organizational forms and the knowledge economy. Supplementing the hierarchical form and the market form are the internal hybrids and the external hybrids that have emerged in response to the challenges of the knowledge economy. What remains in order to provide a systematic way of thinking about organizations as knowledge structures is to derive the organization-level screening function. This function represents the organization-level knowledge, which emerges on the basis of the individual members’ screening function. 19.3.2

Modeling the Knowledge Structure of Architectures

Following Christensen and Knudsen (2003a), a single agent has a reduced screening function of a = f(x). The agent’s screening function represents the agent’s cognitive ability within a particular environment.5 Uncertainty is added because of the finite processing capabilities of the agent, which, depending on the business environment, has more or less severe implications. In turbulent or complex environments, the agent will make more mistakes. The agent-screening function takes this into account as an expression of the limits that characterize the decision maker. In a turbulent and complex environment, the agent-screening function may not be monotonous, it can be very ugly and it can even be a hard-coded function representing choice by the flipping of a coin. The organization as an entity has a screening function of F(a) = q(f(x)). In order to derive the organization-level screening function, the organization is modeled as a graph.6 The graph-screening function F(a) represents the level of knowledge as a function of the individual members’ cognitive skills and the choice of architecture, the organization structure that defines the flow of information among the organization members. Methods to derive the organization-level screening function F(a) have been provided

446

Handbook of knowledge and economics 1 Sketch of graph

0.9

Acceptance probability

0.8 0.7 0.6

8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8

Agent screening Graph screening

0.5 0.4 0.3 0.2

Agent profit: 0.1803 Graph profit: 0.0945

0.1 0 –1

–0.8

–0.6 –0.4 –0.2 0 0.2 0.4 0.6 Project value (uniform pdf, μ = 0, ␴ = 0.57735)

0.8

1

Notes: The dashed line represents the agent-screening function f(x) and the solid line represents the screening function of the hierarchy q(f(x)). We use a uniform project distribution, normalized in the interval [−1,1]. The analytical expression of the graph screening function, F(a) = q(f(x)), was derived by a recursive procedure provided by Christensen and Knudsen (2003a). The revenue was estimated by numerical integration, assuming zero costs of maintaining the shown organization. The agent-screening function in this example was f(x) = 1/2tanh(3/2x) + 1/2.

Figure 19.3

Agent versus hierarchy

in Christensen and Knudsen (2003a).7 In the following, we draw on these methods to illustrate how the knowledge structure of architectures can be represented. In all examples, the agents have reasonable, but far from perfect, cognitive skills. As shown in Figure 19.3, with a probability of about 0.05 an agent accepts the worst offer that is available, and with a probability of about 0.95 an agent accepts the best available offer. The choice of an agent-screening function reflects the ability of the agent to pass judgment within a particular context. In better circumstances our agents might have a screening function closer to the step-function of the omniscient agent. But also within the current environment, investment in an increased skill level might improve the agent’s screening function. Depending on the architecture, the individual agents’ cognitive skills may be enhanced as

The architecture and management of knowledge in organizations 447 well as reduced. Some architectures have structures that will enhance the individual agents’ cognitive skills. This is not the case in the hierarchy shown in Figure 19.3. As can be seen from Figure 19.3, the hierarchy further reduces the cognitive abilities of its members. Even if the hierarchy correctly puts less weight on bad projects, it also rejects too many projects with positive value. We estimated the performance of the agent and of the organization under the assumption that the evaluation costs are zero. Depending on the particular level of agent costs (wage), and the particular cost model of the organization (large-scale costs, unit costs, or perhaps freelance costs), an additional cost would have to be assigned to the organization before a more realistic comparison with the individual agent were possible. In this example, the hierarchy of eight members is not a very good architecture from the perspective of knowledge management. There might well be other reasons for choosing the hierarchy (e.g. issues of contracting), but from a knowledge perspective, the hierarchy is a bad choice given a symmetric uniform project distribution and a screening function similar to that shown. Obviously, if the project distribution only included projects with negative value, the deepest possible hierarchy would outperform any other structure. More generally, this illustrates that the properties of the organization should be evaluated with respect to a particular skill level of the employees, and with respect to a particular business environment. What about the market structure? As shown in Figure 19.4, the market structure is also a bad choice. The individual agent clearly outperforms the market structure. It is the reverse case of the hierarchy. Even if the market structure correctly puts more weight on good projects, it also accepts too many projects with negative value. If the project distribution only included projects with positive value, however, the deepest possible market structure would outperform any other structure. From the perspective of knowledge management, both the hierarchy and the market structure (polyarchy) are not good choices when the business environment offers a distribution of projects with both negative and positive value. Finally, in the example shown in Figure 19.5, we turn to one of the many possible hybrid forms of organization. As can be seen, this particular (external) hybrid outperforms the individual agent. Depending on the (unspecified) evaluation costs, this is an example of a structure one might see in response to the challenges of the knowledge economy. The architecture shown in Figure 19.5 is an example of a hybrid composed of four hierarchies that cooperate with respect to the evaluation of alternative projects. There are four independent routes to acceptance of a project. In three of the hierarchies, the project must be approved at eight

448

Handbook of knowledge and economics 1 Sketch of graph

0.9

Acceptance probability

0.8 0.7 0.6

8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8

Agent screening Graph screening

0.5 0.4 0.3 0.2 Agent profit: 0.1803 Graph profit: 0.0945

0.1 0 –1

–0.8

–0.6 –0.4 –0.2 0 0.2 0.4 0.6 Project value (uniform pdf, μ = 0, ␴ = 0.57735)

0.8

1

Note: Apart from the architecture, the same specification is used as in the example shown in Figure 19.3.

Figure 19.4

Agent versus market (polyarchy)

levels before it is ultimately accepted, but in one of the hierarchies only three levels need to approve the project before an ultimate acceptance occurs. As the value of knowledge increases and as the costs of organizing information exchange decrease, we would expect to see the emergence of various forms of hybrids. In summary, we have introduced a new way of thinking about the organization of knowledge, as an architecture that may help boundedly rational agents make better choices. By way of example, we have further illustrated that the emergence of hybrid forms of organizing is consistent with their properties as knowledge structures. In a business environment that offers a distribution of projects with both positive and negative value, we would expect hybrid forms of organizing to emerge. In particular this would be the case if the cost of information exchange decreased, as widely reported during the last decade. Our modeling framework also points to a number of well-known trade-offs that may be considered from a design perspective. The organization’s ability to evaluate new projects can be enhanced in three ways: (1) the agent’s ability to screen projects may be

The architecture and management of knowledge in organizations 449 1 Sketch of graph

0.9

Acceptance probability

0.8 0.7 0.6

8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8

Agent screening Graph screening

0.5 0.4 0.3 0.2 Agent profit: 0.1803 Graph profit: 0.2306

0.1 0 –1

–0.8

–0.6 –0.4 –0.2 0 0.2 0.4 0.6 Project value (uniform pdf, μ = 0, ␴ = 0.57735)

0.8

1

Notes: In this hybrid, projects enter at the agent shown in position (1,1) of the grid. Apart from the architecture, the same specification is used as in the example shown in Figure 19.3.

Figure 19.5

Agent versus (external) hybrid

improved through training programs and education; (2) the organization might adapt its architecture to obtain better fit with its business environment; and (3) the organization might choose a different business environment. These issues introduce considerations regarding the search for better cognitive ability (screening functions), adaptation of the architecture to accommodate a fluctuating business environment and the effect of adapting aspirations on search processes and risk-taking. Issues of adaptation are addressed in the following section, which considers the dynamics of knowledge organization.

19.4 DECOMPOSABILITY OF THE KNOWLEDGE ORGANIZATION In order to address issues regarding organizational adaptation, we use the framework of the behavioral theory of the firm to introduce the property of near decomposability (ND). ND is an architectural feature

450

Handbook of knowledge and economics

that increases the rate at which organizations, and other evolving entities, can adapt to achieve a better fit (in the sense of a match) with the environment (Simon and Ando, 1961; Simon, 1996). In architectures that are ND, (1) the short-term (high-frequency) behavior of each subsystem is approximately independent of the other subsystems at its level, and (2) in the long run, the (low-frequency) behavior of a subsystem depends on that of the other components only in an (approximately) aggregate way. That is, in ND systems each subsystem can adapt to achieve a better fit as if it were independent of the other subsystems. ND confers evolutionary advantages on the organization because, in nearly decomposable systems, each component can evolve toward greater fit, with little dependence upon the changes taking place in the details of other components. Turning to the modeling framework introduced above, the improvement of the agent’s ability to screen projects might be viewed as a straightforward matter, but a little consideration tempers this optimism. In many real settings, the improvement of the agent’s cognitive ability is a matter of adaptation rather than choice. Within the framework introduced above, this implies that the agent would have to engage in adaptive search for better screening functions rather than being able to choose the desired improvement in screening. If the available screening functions were distributed in random order, this would make the improvement of cognitive ability a matter of search in a rugged landscape with a local maximum as the most likely outcome. This individual-level problem of search for better cognitive ability is further complicated by the interaction of organizational members. When one member changes the screening function, the organization-level screening function will change, but the organization-level effects of such changes are very hard, if not impossible, to predict for boundedly rational agents. For this reason, the adaptation to improve organization-level screening ability through search for improved individual screening functions is best supported by a structure where subsystems can adapt independently, that is, systems with the ND property. Previous research has shed some light on the strategies that can help individuals to improve their mental models (Levinthal, 1997; Gavetti and Levinthal, 2000). How organization-level properties can be derived from individual-level mental models and the architectures that support them is an issue that has not yet been addressed. As indicated by the examples in the previous section, this is an important issue that deserves further attention. The present chapter contributes by identifying a fundamental trade-off between improving individual-level screening functions and improving the way architectures support individuals. Because an organization can achieve better fit through improved

The architecture and management of knowledge in organizations 451 individual-level screening, through improved architecture, or through an effort on both dimensions, it is important to identify the most promising strategy. Turning to the issue of architectural adaptation, and holding the individual-level screening function constant, we could envision a business environment that fluctuated between offering projects with negative and positive value. In this case, the optimal architecture would fluctuate between the hierarchy and the market. One strategy would be to choose architectures that on average performed well (external hybrids), possibly with minor adjustments that could accommodate the fluctuations of the business environment (flatter organizations in good times). In a turbulent business environment, flexibility is an important trait, but it is valuable only if the costs of adapting from one structure to the next are not prohibitive (either in terms of switching costs or, indirectly, in terms of the costs of a reduced fit between architecture and environment). One way to reduce the costs of adapting the architecture is through the use of ad hoc project organizations, an increasingly attractive option as reduced communication costs allow employment of knowledge workers on a project basis. Another way to reduce adaptation costs is to take advantage of the ND property. As previously mentioned, ND confers evolutionary advantages on the organization because, in nearly decomposable systems, each subcomponent can evolve toward greater fitness with little dependence upon the changes taking place in the other subcomponents. In a turbulent environment, we would therefore expect the emergence of architectures constructed such that there were limited interface between its subsystems. In this way, each subsystem can adapt without much interference in the other subsystems. One question that presents itself is, therefore, whether there is a trade-off between the ND property and the organization-level screening function. Figure 19.6 shows an example of an architecture that both has the ND property and achieves a better fit than the individual agent. This structure even outperforms the hybrid shown in Figure 19.5.8 The architecture in Figure 19.6 is composed of two independent blocks – nearly independent subsystems that are connected through two liaisons. One of the subsystems is an external hybrid that includes three cooperating hierarchies with five layers. The second subsystem is an external hybrid that includes three cooperating hierarchies, each with three layers. This second subsystem is used as a control system that is activated only in case the first subsystem rejects an alternative at layer five. The control system could, for example, be thought of as an independent organization that offered executive-level consultancy services.

452

Handbook of knowledge and economics 1 Sketch of graph

0.9

Acceptance probability

0.8 0.7 0.6

8 7 6 5 4 3 2 1

Agent screening Graph screening

1 2 3 4 5 6 7 8

0.5 0.4 0.3 0.2 Agent profit: 0.1803 Graph profit: 0.2356

0.1 0 –1

–0.8

–0.6 –0.4 –0.2 0 0.2 0.4 0.6 Project value (uniform pdf, μ = 0, ␴ = 0.57735)

0.8

1

Notes: This hybrid is designed as a nearly decomposable system. There is limited interface between two relatively independent substructures. The two substructures are connected through two agents that act as liaisons. Apart from the architecture, the same specification is used as in the example shown in Figure 19.3.

Figure 19.6

Agent versus nearly decomposable hybrid

Not only does the architecture shown in Figure 19.6 improve performance; it also has adaptive advantages because of its ND property. In order to compare the performance of architectures across examples, we have held constant the screening function of the agents. If the architecture shown in Figure 19.6 engaged in search for better screening functions, each subsystem could, independent of the other, adapt towards better organization-level screening. The question of trade-off between the ND property and adaptation toward a better organization-level screening function is an important issue that deserves attention in future research. In the present chapter, we have outlined a modeling framework that points to the evolution of architectures and the evolution of screening functions as intimately related issues. In order to better understand the emergence of knowledge organizations, these issues must be considered with respect to the nature of the business environment (the expected value and fluctuations of opportunities and

The architecture and management of knowledge in organizations 453 threats) and the costs of improving organization-level screening through experimentation with new organizational forms (architectures) or through search for improved individual-level screening.

19.5 CONCLUSION The purpose of this chapter was to introduce a new, unifying, way of thinking about the organization of knowledge as an architecture that may help boundedly rational agents make better choices. We have outlined a modeling framework that can be used to develop models of knowledge organizations. We have primarily focused on the architecture of knowledge organizations and, through examples, illustrated a way to think about hybrid forms of organization. Within a context defined by the behavioral theory of the firm, decision makers are boundedly rational and the possibility of designing architectures that can support their members is a critical issue. Within the context of lower communication costs and increased connectivity of various media, both good and bad alternatives travel faster. The need to design architectures that help their members reduce error by rejecting bad alternatives and accepting good ones is therefore an important but largely overlooked issue in knowledge management. The present chapter offers an approach to think about this issue in a systematic way. The design of an architecture requires consideration of the cognitive skills of possible employees, the distribution of alternatives that are available in the business environment, and the costs associated with alternative modes of employment. A basic lesson for knowledge management is that an architecture should be chosen to achieve a match on all three dimensions. As emphasized by Herbert Simon (1997, p. 37), this is not a simple issue: The design of products (and not just the choice of products) is often a central concern, and marketing procedures, manufacturing procedures, pricing policies, the central organization structure, even long-term strategies are designed, and not just chosen. Design calls for initiative, focus of attention on major problems, search for alternatives. One cannot choose the best, one cannot even satisfice, until one has alternatives to choose from.

In this chapter, we have pointed out how architectures can be designed to improve the search for alternatives that happen to be included in the choice set. One possibility that is commonly aired is that such issues of design are ultimately unimportant because both markets and organizations work as

454

Handbook of knowledge and economics

self-adjusting mechanisms. But others find that the notion of a ‘market’ as a self-adjusting mechanism is not enough to explain how real-world markets or organizations actually function: ‘Nor does this self-adjusting system have much to do with the information that the firm must gather in order to carry out the numerous search and design activities mentioned earlier’ (Simon, 1997, p. 36). According to Simon, ‘a study of the allocation of management time would almost certainly show that it is the latter that account for most of the managers’ days . . .’ (Simon, 1997, p. 37). We agree and, therefore, point to the design of knowledge organizations as one of the critical unsolved issues in knowledge management. As a first step, we offer a framework that can help managers and researchers to think about the ways in which organizations may be designed to reduce error in various search activities. We hope that further steps will soon be taken to develop the architectural approach to designing knowledge organizations.

NOTES 1. Specifically, Winter (2003, p. 991) defines the concept of organizational capability in relation to the broader idea of organizational routines: ‘An organizational capability is a high level routine (or collection of routines) that, together with its implementing input flows, confers upon an organization’s management a set of decision options for producing significant outputs of a particular type.’ 2. Sah (1991) provides an overview of the literature and Stiglitz (2002) assesses its current status within information economics. Koh (1992, 1994) and Christensen and Knudsen (2003a, 2003b) add in a number of ways to the basic models of Sah and Stiglitz (1986, 1988). 3. In order to do this, we would model alternative search as a graph (a sampling path) and alternative generation as a process that changed the distribution of alternatives that an organization can choose from. 4. Sah and Stiglitz (1986) refer to project evaluation. Here we refer to alternative evaluation. 5. Cognitive ability refers to both innate capacity and learned mental models. 6. Christensen and Knudsen (2003a) pin down a number of requirements (invariants) that must be fulfilled in order to make the graph representative of realistic organizations. The most important thing is that the graph must be directed, finite and connected. 7. F(a) is a polynomial dictated by the geometry and topology of the graph representing the organization. The highest order is equal to the number of agents on the longest path to overall acceptance. The lowest order (possible) is equal to the minimal number of acceptances by agents required to get an overall acceptance. Christensen and Knudsen (2003a) showed how to derive the graph-screening function. 8. The architecture in Figure 19.6 even contains one agent fewer than the architecture shown in Figure 19.5.

The architecture and management of knowledge in organizations 455

REFERENCES Chandler, A. (1962), Strategy and Structure: Chapters in the History of Industrial Enterprise, Cambridge, MA: MIT Press. Chandler, A. (1990), Scale and Scope, Boston, MA: Harvard University Press. Child, J. and McGrath, R.G. (2001), ‘Organizations unfettered: organizational form in an information intensive economy’, Academy of Management Journal, 44 (6), 1135–48. Christensen, M. and Knudsen, T. (2003a), ‘The architecture of economic organization: toward a general framework’, mimeo, Department of Marketing, University of Southern Denmark, Odense. Christensen, M. and Knudsen, T. (2003b), ‘Extremal screening, cost and performance of economic architectures’, mimeo, Department of Marketing, University of Southern Denmark, Odense. Collins, J.J. and Chow, C.C. (1998), ‘It’s a small world’, Nature, 393, 409–10. Cyert, R.M. and March, J.G. (1963/1992), A Behavioral Theory of the Firm, Englewood Cliffs, NJ: Prentice-Hall. Dosi, G. (1988), ‘Sources, procedures and microeconomic effects of innovation’, Journal of Economic Literatures, XXVI, 1120–71. Eisenhardt, K.M. and Martin, J.A. (2000), ‘Dynamic capabilities: what are they?’, Strategic Management Journal, 21 (10/11), 1105–21. Gans, J.S. and Shepherd, G.B. (1994), ‘How are the mighty fallen: rejected classic articles by leading economists’, Journal of Economic Perspectives, 8 (1), 165–79. Gavetti, G.M. and Levinthal, D. (2000), ‘Looking forward and looking backward: cognitive and experimental search’, Administrative Science Quarterly, 45, 113–37. Huberman, B.A. and Hogg, T. (1995), ‘Distributed computation as an economic system’, Journal of Economic Perspectives, 9 (1), 141–52. Koh, W.T.H. (1992), ‘Variable evaluation costs and the design of fallible hierarchies and polyarchies’, Economics Letters, 38, 313–18. Koh, W.T.H. (1994), ‘Making decisions in committees: a human fallibility approach’, Journal of Economic Behavior and Organization, 23, 195–214. Levinthal, D. (1997), ‘Adaptation on rugged landscapes’, Management Science, 43, 934–50. Levinthal, D. and March, J.G. (1993), ‘The myopia of learning’, Strategic Management Journal, 14, 95–112. March, J.G. (1978), ‘Bounded rationality, ambiguity, and the engineering of choice’, The Bell Journal of Economics, 9 (2), 587–608. March, J.G. (1991), ‘Exploration and exploitation in organizational learning’, Organization Science, 2 (1), 71–87. March, J.G. and Simon, H.A. (1958), Organizations, New York: Wiley. Moore, E.F. and Shannon, C.E. (1956), ‘Reliable circuits using less reliable relays (Part I)’, Journal of the Franklin Institute, 262, 191–208. Prahalad, C.K. and Hamal, G. (1990), ‘The core competence of the corporation’, Harvard Business Review, 68 (3), 79–91. Romanelli, E. (1991), ‘The evolution of new organizational forms’, Annual Review of Sociology, 17, 79–103. Sah, R.K. (1991), ‘Fallibility in human organizations and political systems’, Journal of Economic Perspectives, 5, 67–88. Sah, R.K. and Stiglitz, J.E. (1986), ‘The architecture of economic systems: hierarchies and polyarchies’, American Economic Review, 76, 716–27. Sah, R.K. and Stiglitz J.E. (1988), ‘Committees, hierarchies and polyarchies’, Economic Journal, 98, 451–70. Simon, H.A. (1947), Administrative Behavior, Boston, MA: MIT Press. Simon, H.A. (1955), ‘Recent advances in organization theory’, in S.K. Bailey et al. (eds), Research Frontiers in Politics and Government, Washington, DC: Brookings Institution, pp. 23–44. Simon, H.A. (1957), ‘A behavioral model of rational choice’, in Models of Man, Social and

456

Handbook of knowledge and economics

Rational: Mathematical Essays on Rational Human Behavior in a Social Setting, New York: Wiley. Simon, H.A. (1993), ‘Altruism and economics’, The American Economic Review, 83 (2), 156–61. Simon, H.A. (1996), The Sciences of the Artificial, 3rd edn, Boston, MA: MIT Press. Simon, H.A. (1997), An Empirically Based Microeconomics, Cambridge: Cambridge University Press. Simon, H.A. and Ando, A. (1961), ‘Aggregation of variables in dynamic systems’, Econometrica, 29, 111–38. Stiglitz, J.E. (2002), ‘Information and change in the paradigm in economics’, American Economic Review, 92, 460–501. Teece, D.J. (1993), ‘The dynamics of industrial capitalism: perspectives on Alfred Chandler’s scale and scope’, Journal of Economic Literature, 31 (1), 199–255. Teece, D.J. (1998), ‘Capturing value from knowledge assets: the new economy, markets for know-how, and intangible assets’, California Management Review, 40 (3), 55–79. Teece, D.J. (2000), ‘Strategies for managing knowledge assets: the role of firm structure and industrial context’, Long Range Planning, 33 (1), 35–54. Winter, S.G. (2000), ‘The satisficing principle in capability learning’, Strategic Management Journal, 21, 981–96. Winter, S.G. (2003) ‘Dynamic capabilities’, Strategic Management Journal, 24 (10), 991–5.

The architecture and management of knowledge in organizations 457

APPENDIX

1. Firm

2. Internal

T

I

F

3a. External

T

3b. External

T

I

F

T

3c. External

F

F

F

T

I

I

I

4. Market

T

F I

Figure 19A.1

Examples of organizational forms: from firms to market organization

20 Distributed knowledge and its coordination* Markus C. Becker

20.1 WHAT IS DISTRIBUTED KNOWLEDGE? Consider a small corner-shop bakery. It is owned and operated by one baker, who produces bread in the early morning, and sells it later in the day. In such a case of crafts production, the baker knows everything necessary to produce and market bread. Compare this bakery with a five-person bakery (maybe the same one after considerable growth). The question relevant for understanding distributed knowledge is: how much does the knowledge that each person has differ? In principle, there is a continuum of possibilities, with two end points. (a) All five persons in the bakery hold the same knowledge. For instance, all five know how to bake bread, how to sell bread, and they also know how to make and sell all the different types of bread offered. (b) They all hold different knowledge. One knows how to bake bread, but has never bothered to learn about how to be good at selling it; one is good at selling but does not know how to bake bread and so on. (c) In between these two extremes, there are intermediate forms. Employee A knows very well how to bake bread, but also knows a little bit about selling; employee B knows very well how to sell bread, but also knows a little bit about baking. Focusing on the knowledge held by the actors, the three situations in our examples have the following characteristics. At extreme (a), a multi-person firm is a simple multiplication of a oneperson crafts-shop operation. Every employee has the same knowledge.1 No division of labor takes place, no specialization effects accrue. The knowledge held by any two employees in the firm completely overlaps – they know the same. The knowledge held by the different employees is essentially a clone, a perfect replica (think of McDonald’s franchisees and their knowledge of operating a McDonald’s restaurant). Because of this virtue, the knowledge that each of the actors holds is shared knowledge – everyone knows exactly what everyone else knows.2 458

Distributed knowledge and its coordination 459 At this point, it is important to note the difference between shared knowledge and common knowledge. In the example above, we have focused only on task-related knowledge (how to bake and sell bread). What has remained outside the picture is the knowledge about whether other employees know that employee A knows how to bake bread. The term ‘shared knowledge’ refers to the task-related knowledge held by the different actors. The term ‘common knowledge’ refers to knowledge about whether the other actors are informed about what knowledge a particular actor holds. In our example, in bakery (a), the bakers have shared knowledge, as well as common knowledge (they have the information about what it is that the others know).3 In a bakery of type (c), such as a large commercial bakery, each employee might not have information about what it is that the other employees know how to do (well) – shared knowledge without common knowledge. At the other extreme (b), each actor is specialized to an extreme degree – in the limit, to such a degree that she focuses completely on, for instance, one particular step in the production process (preparing dough, baking, selling). This will be the case when division of labor is taken to the extreme (in the vein of Taylor and Gilbreth). Here, there is no overlap in the task-related knowledge held by the different actors. The implication is that while (in the case of division of labor according to steps in the value chain) there are specialization effects, actors lack (deeper) understanding of the previous and the following steps, that is, of the operations the other actors carry out. This situation is one of distributed, or dispersed, knowledge (Hayek, 1937, 1945; Minkler, 1993; Tsoukas, 1996; Rulke and Galaskiewicz, 2001).4 The implication of such a situation is that because of the lack of overlap in task-related knowledge, a group of agents knows something no single agent (completely) knows. The agents that make up a system, such as a firm (Tsoukas, 1996), an industrial district (Foss and Eriksen, 1995), or an economy (Hayek, 1945) collectively possess knowledge that no single agent possesses (Foss and Foss, 2003). Consider the example given by Foss and Foss (2003, p. 6): ‘Jack knows that p is the case and Jill knows that p implies y, but neither knows that y is the case. However, if Jack and Jill’s information states are “added” there is a sense, which is more than metaphorical, in which they may know that y is the case.’ While the knowledge is still held in the individuals’ minds, what makes the collectively held knowledge more than a simple ‘addition’ of individually held knowledge is the architecture of the organization, that is, the way in which agents are linked.5 In the example, the information that y is the case is present in the system comprising Jack and Jill, but in a distributed form (Foss and Foss, 2003). The architecture linking both agents will decide

460

Handbook of knowledge and economics

whether this information will be integrated when solving a task requires it. Architecture, that is, the structure of the linkages between the elements, matters. Modular systems are the prime illustration of why architecture matters. At the same time, modular systems also represent that form of architecture that attempts to the most extreme degree to make use of the dispersion of knowledge. The whole idea of modularity is that the modules can be ‘black-boxed’, their production and design outsourced, while the system can be coordinated by focusing on the design of an architecture, interfaces and standards (Baldwin and Clark, 1997; Langlois, 2002).6 The motivation for using modular design is that the system’s components can be separated and recombined (Schilling, 2000), thereby leading to benefits such as shortened lead times, leveraging product variations, enabling common components to be used within and across product lines and so on (Sanchez, 1997). In order to achieve maximum benefits from modularization, however, the interactions and interdependencies between modules need to be well specified (e.g. knowing the ‘law’ that specifies the interaction). The reason is that modularity of a system is one way to organize the division of labor. For division of labor to work, however, two conditions are required. The original problem has to be represented in an additive form, and each part into which the original problem is decomposed must not interfere with the other partial solutions (Munari et al., 2003). Only in this case can the partial solutions be simply ‘summed up’ to give the solution to the complex problem, can evolve each at their own speed, and can be freely recombined (cf. Munari et al., 2003). In the case of modular products, the principle of one-to-one mapping of functions and components (Ulrich, 1995) is a particular clear illustration (such as having one component, the screen, fulfilling one function, optical output of a computer). Only where a clear one-to-one mapping of functions and components is entirely possible will it be possible to attain a certain functionality of the product by assembling the modular components, which have been designed (and produced) only by focusing on standardized interfaces, not taking into account interactions and interdependencies between modules other than those captured by the standardized interfaces. A counter-example to such an ideal case of modularization would be the NVH (noise–vibration–harshness) characteristics of a car. No one-to-one mapping of this function (having good NVH characteristics) to any component in the car is possible, as is possible for instance in the case of the interior illumination. The reason is that noise, vibration and harshness of the ride are the result of many interactions of many components of the car, which we have difficulties specifying precisely.

Distributed knowledge and its coordination 461 The last pattern of knowledge dispersion we need to look at in more detail is the intermediate situation (c): agents have some degree of overlap and some degree of specialization. Where the overlap in the knowledge held by the different actors is not perfect, as in (a), the knowledge they hold is asymmetric. In this first section, we have defined distributed knowledge, and have analyzed the spectrum of different patterns of dispersion of knowledge. Two issues are important to keep in mind when analyzing the coordination of distributed knowledge: the distinction between shared knowledge and common knowledge; and the importance of the structure of links between agents (architecture). In the next section, we turn to the effects of distributed knowledge on organizations in order subsequently to be able to analyze responses to situations of distributed knowledge.

20.2 EFFECTS OF DISTRIBUTED KNOWLEDGE 20.2.1

Economies of Scale

Economies of scale refer to decreasing average unit costs of producing a (single or composite) output under a given technology with rising output (Silvestre, 1987, p. 80). There are two drivers of economies of scale: indivisibility in input factors, and set-up cost. Indivisibility in input factors means that once one (indivisible) unit of the input factor is acquired, its factor price is a fixed cost. Within the maximum capacity of one unit of input factor (e.g. one machine), producing one additional output unit will reduce average unit cost. Set-up costs have the same implication. They represent fixed costs. Once the set-up (e.g. of one machine) is complete, any additional unit of output will reduce the average unit cost of output produced using the machine. Set-up costs arise due to (i) concentrating and getting psychologically ready for the task, (ii) learning how to do it, and (iii) preparing the tools needed (ibid., p. 81). The drivers of economies of scale are therefore either the usage ratio of the capacity of indivisible units of input factors, or set-up costs. Famously, Adam Smith has linked economies of scale and the division of labor (which is linked to the distribution of knowledge).7 On closer inspection, the dispersion of knowledge is not directly causally implied in bringing about economies of scale. Economies of specialization and economies of scale, however, often appear coupled. We therefore consider economies of specialization separately, as it is a different causal mechanism that feeds into the same effect (higher productivity) (see also Martens, 1999).

462

Handbook of knowledge and economics

20.2.2

Economies of Specialization

Economies of specialization refer to productivity improvements. They are often coupled with economies of scale (namely, where either indivisibilities of factor inputs or set-up costs are involved). Economies of specialization, however, are brought about by a different causal mechanism. This causal mechanism is learning, accruing experience with one task, and becoming better at carrying out that task. Productivity gains result, as expressed in the learning curve (Argote, 1999). Productivity gains are enabled by the fact that when specializing, one focuses on less and/ or more narrow tasks. By definition, if work time and the time needed to carry out the task are held constant, agents produce fewer activities each more often per period (Ippolito, 1977). Agents therefore proceed faster along their learning functions (Ippolito, 1977). Accordingly, specialization can thus be operationalized by considering (a) the number of activities performed per labor input, (b) the time distribution associated with the performance of these activities, (c) the level of productivity of the inputs (i.e., criterion (a) and (b) are sensitive to this factor), and (d) the degree of ‘similarity’ among the activities within job allocations performed by labor (Ippolito, 1977, p. 471). Higher factor productivity, of course, feeds into decreasing average cost (economies of scale). This is why economies of scale and economies of specialization are sometimes conflated, even by Smith himself (Martens, 1999). Distinguishing the two is important because the causal mechanism leading to productivity improvements is different. That additional analytical precision becomes important, for instance, when the limits to productivity improvements matter. Famously, Smith’s answer is that ‘the division of labour is limited by the extent of the market’ (Smith, 1776) – and because the division of labor is the main driver of productivity improvements, productivity improvements are also limited by the extent of the market. That makes perfect sense, as economies of scale are essentially average unit costs, for which output is crucial. On the other hand, productivity gains due to specialization are also subject to other limits. While it is true that higher output will increase frequency and therefore learning effects, it is for instance possible to simulate the process – in which case the extent of the market would not be a ‘hard’ constraint with regard to economies of specialization. There are also other forms of learning that might lead to a different limit to productivity gains than the extent of the market. To sum up, the pattern of dispersion of knowledge in an organization influences how difficult the transfer of (task-related) knowledge in the firm will be, due to asymmetries in the underlying knowledge.

Distributed knowledge and its coordination 463 The balance between commonality and diversity of knowledge held by the agents involved seems important. Where the knowledge bases of the parties are homogeneous, there will be few coordination problems (at least as far as its cognitive aspect is concerned), but there will also be few opportunities for mutual learning, as each party’s knowledge is already similar to the other’s. On the other hand, if the knowledge bases are highly diverse, opportunities for mutual learning are high, but communication and coordination between parties are likely to be more problematic (Lazaric and Marengo, 2000, p. 14). It is interesting to note that for many decades, specialization was considered an efficient way to capture and use knowledge. More recently, the importance of shared knowledge and redundancy in cases such as organizing for new product success has been increasingly emphasized, however, pointing to an important balance between specialization and shared knowledge (Moorman and Miner, 1997, p. 102). 20.2.3

Economies of Speed/Time to Market

As mentioned above, the extreme form of knowledge dispersion is represented by modularization. Modules have no overlap and the standardization of interfaces means that they can be combined freely. To the extent that this is the case, one of the main benefits of modularization to be reaped is economies of speed/time to market, due to the fact that the development of different components can take place in parallel (Sanchez, 1997). Importantly, thanks to modular design, the development of components can begin before the design of the other components, linked to the component in question, is finished. Maybe the most basic insight in this section is that among the effects of the dispersion of knowledge, there are both economies and diseconomies, immediately pointing to limits to the dispersion of knowledge. Maximum modularization, therefore, is not always a good idea. Furthermore, which of the economies and diseconomies will be dominant is also influenced a great deal by the perspective taken. In organizations, functional perspectives are often strong. From a production perspective, for instance, economies of scale will be in the foreground, while from a product development perspective, economies of speed will attract most attention. It is thus predictable that decisions on the division of labor and the dispersion of knowledge (often made in order to benefit from economies) can easily be imbalanced, with respect to the full spectrum of economies and diseconomies they imply. Finally, the distinction of two levels of knowledge, task-related and underlying knowledge, is important because overlaps on the underlying knowledge level are crucial for dealing with dispersion on the level of task-related knowledge.

464

Handbook of knowledge and economics

20.3 WHAT PROBLEMS DOES DISTRIBUTED KNOWLEDGE POSE? 20.3.1

Increasing Coordination Costs

In organization theory, coordination refers to aligned actions (Heath and Staudenmeyer, 2000). In order to align actions, knowledge about the interdependencies is crucial (Thompson, 1967). For bringing about coordination, two processes are required: intelligence generation (information gathering), and information processing (including interpretation). Intelligence generation refers to the seeking, identification and acquisition of information. Information processing refers to what happens with the acquired information. Typically, it needs to be processed, for instance applied in calculation, and interpreted in order to be become meaningful for the coordination task (Fransman, 1998). Coordination activity is costly. Coordination costs derive from the two steps in the coordination process identified above. In both cases, there are opportunity costs for engaging in these processes. And in both cases, the costs increase with increasing dispersion of knowledge. Intelligence generation (information-gathering) cost increases with increasing dispersion of knowledge because, with increasing specialization, it is less and less possible to solve problems by deducing a solution to a concrete problem from general knowledge. Specialist knowledge has to be acquired (Hayek, 1937). Information-processing (and interpretation) costs increase with increasing dispersion of knowledge for two reasons: first, increasing specialization leads to increasing complexity of the knowledge in question. Increasing complexity in turn means that the quantity of information required to understand a specialized domain increases (intelligence generation) (Martens, 2001). The opportunity cost of obtaining information (outside the domain of specialization) increases with the degree of specialization (Martens, 2001). Second, due to increased complexity (e.g. larger diversity between different bits of information), interpreting and making sense of the information also becomes more difficult, and timeand resource-consuming (information processing). The result is that with increasing dispersion of labor (specialization), coordination costs increase. Probably the most noted aim of the division of labor is to lead to productivity gains, and thus to reducing unit cost (economies of scale).8 Because one ‘side-effect’ of the division of labor is to give rise to coordination costs, those also need to be factored into the total cost. Coordination costs provide the answer to the question ‘Why does the firm not exploit [increasing returns] further and in the process become a monopoly?’ (Stigler, 1951, p. 188).9 As Stigler emphasized, some functions do not show

Distributed knowledge and its coordination 465 increasing returns with increasing division of labor, for two reasons. First, the wider the range of functions the firm undertakes, the greater the tasks of coordination (ibid., p. 189). Because coordination is costly, coordination costs also have to be considered. Second, ‘there are other functions subject to diminishing returns, and these are, on balance, at least so costly that average cost of the final product does not diminish with output’ (ibid., p. 188). The reason why the firm cannot outsource those functions that are subject to decreasing returns is that the demand for the products or services produced by these functions might be too small for any firm to offer them.10 20.3.2

Difficulties with Knowledge Transfer

The dispersion of knowledge also has an impact on the difficulty with which that knowledge can be transferred from one agent to another. The role of shared knowledge for transferring specialized knowledge is central here (Demsetz, 1988; Grant, 1996a, 1996b; Okhuysen and Bechky, 2009). As Demsetz (1988) argues, shared knowledge (e.g. of mathematical notation) helps transmit specialized knowledge (such as solutions to mathematical problems). Two levels of knowledge are important here: the specific knowledge relating to the task at hand (e.g. designing an automotive component), and more general knowledge required as background to make sense of the task-specific knowledge and apply it (such as cognitive categories required for interpretation, e.g. general mathematical knowledge of how to interpret a mathematical formula). When transferring specialist knowledge from, say, a design specialist to a specialist in material science, the difficulty of this transfer of task-related knowledge will be strongly influenced by the degree of common knowledge the two specialists have. A second difficulty for knowledge transfer arises from different conceptualizations of the expertise of others. Empirical findings (Cramton, 2001; Rulke and Galaskiewicz, 2001; see also Cohendet and Llerena, 2003) indicate that in groups made up of generalists (i.e. larger overlap, more shared knowledge) group members are more likely to share conceptualizations of one another’s expertise. Furthermore, shared conceptualizations are prerequisites for knowledge transfer (Okhuysen and Bechky, 2009). Very different conceptualizations will represent barriers to knowledge transfer (Dougherty, 1992). The information transferred will not be interpreted in such a way that the knowledge produced by the receiver’s interpretation will be similar to the knowledge possessed by the sender. Obviously, this is of great importance for absorptive capacity, that is, being able to receive, interpret and apply knowledge (Cohen and Levinthal, 1990). Much of such underlying knowledge, however, is largely unconscious (taken-for-granted assumptions in a profession are a good example). The way such knowledge is acquired, and

466

Handbook of knowledge and economics

in which cognitive schemata are formed, is largely unconscious (Witt, 1998), and has become known as ‘implicit learning’ (Reber, 1993). Implicit learning involves the transfer of uncodifiable (tacit) knowledge. Such knowledge can only be transferred through a time-consuming process involving faceto-face interactions among individuals). The more different the underlying knowledge held by the two agents, the longer the time needed. While taskrelated knowledge might or might not be tacit, the underlying knowledge required in order to apply task-related knowledge is usually highly tacit (Reason, 1990; Reber, 1993; Witt, 1998; Knudsen and Foss, 1999).11 The implication is that, because of dispersed knowledge, agents have less common overlap, fewer commonalities on an underlying level such as cognitive frames. Such commonalities are a prerequisite for transferring knowledge (Dougherty, 1992; Demsetz, 1988; Grant, 1996a, 1996b; Argote and Ingram, 2000; Lapré and van Wassenhove, 2001; Lane and Lubatkin, 1998; Larsson et al., 1998). Specialization therefore creates a barrier to knowledge transfer, presenting an endogenous limit to the division of labor. 20.3.3

Limits to Authority

In the economics literature, the most prominent mechanisms for achieving coordination are hierarchies and markets, the former achieving coordination by authority and the latter by the price mechanism (Coase, 1937; Williamson, 1975, 1985). In between these two coordination mechanisms, there are intermediate forms such as clan-based control (Ouchi, 1980), or networks. In this section, we focus on authority, and on the impact of an increase in distributed knowledge on the power of authority as a coordination mechanism. Authority means the right to decide on what other people should do and how. Authority means ‘deliberate direction’ (Hayek, 1937, p. 49). The question which of these two coordination mechanisms – firm or market – is more appropriate has been famously debated in the so-called ‘Socialist Planning Debate’ in the 1930s. It was Hayek who argued that authority as a coordination mechanism must run into limits. Starting from the observation that ‘the knowledge of the circumstances of which we must make use never exists in concentrated or integrated form, but solely as the dispersed bits of incomplete and frequently contradictory knowledge which all the separate individuals possess’ (Hayek, 1945, p. 519), his arguments were the following. (1) In order to address the main economic problem of society, to adapt to changes in the particular circumstances of time and place, knowledge of the particular circumstances of time and place is required (ibid., pp. 522–4); (2) such knowledge can

Distributed knowledge and its coordination 467 never be transferred completely to one central decision maker, simply because of the limited span of control, as Hayek (ibid., p. 525) called it. Simon added the reason for limits to the span of control, limited cognitive resources (Simon, 1955). (3) The person invested with the authority needs to acquire the knowledge (intelligence dissemination, knowledge transfer), and therefore the problems/barriers identified in (2) apply. Where such barriers apply, the consequence is that the person invested with authority has difficulties acquiring knowledge about what the subordinates are doing. To the extent that a superior can only direct a subordinate regarding the activities he or she knows about, ‘[i]f the subordinate is knowledgeable about activities that the superior is not, the superior could not direct the subordinate to engage in those activities’ (Minkler, 1993, p. 576).12 (4) Because increasing complexity and increasing requirement of information also apply to the decision maker, intelligence generation (informationgathering), and information-processing (and interpretation) cost increase with increasing specialization/division of labor, as identified in (1).13 An important consequence of these four drivers of problems for bringing about coordination by authority is that [d]istributed knowledge causes authority (as a centralized decision-making system) to fail in all its forms (Radner 1997), not because processes are costly, but because too much information is lost in a perhaps efficient system; simply, no central agent can possess the relevant knowledge to produce high-quality decisions, and it is technically necessary to shift to a (more costly) decentralized team-line scheme. (Grandori, 2002, p. 257)

As a consequence, the allocation of decision rights (authority) needs to take into account limits presented by distributed knowledge. Demsetz (1988, p. 157) argues that ‘[f]irms and industries must form a pattern of economic organization that takes account of the need for acquiring knowledge in a more specialized fashion than the manner in which it will be used’. Because of this asymmetry in the cost of acquiring and using knowledge, it makes sense to economize on the cost of producing, maintaining and using knowledge by allocating decision rights to a few individuals who have invested in acquiring specialized knowledge, and who provide directions to other individuals who do not need to have the same level of specialized knowledge in order to understand why following the directions will lead to achieving a particular objective. 20.3.4

Increasing Imbalance between Exploration and Exploitation

As described above, probably the most noted motivation for the division of labor is economies of scale and of specialization. The reason is simple:

468

Handbook of knowledge and economics

both economies of scale (decreasing unit costs due to higher throughput with indivisible production factors) and economies of specialization increase efficiency. An organization characterized by high division of labor and specialization is characterized by being geared towards efficiency. Efficiency is only one important parameter, however, at least in the context of a changing environment in which organizations need to adapt to such changes. In such a context, the ‘balance between efficiency and adaptiveness, between exploiting what is known and exploring what might come to be known’ is important (March, 1994, p. 2). With increasing division of labor, one risks developing an imbalance: the organization is caught in a ‘success trap’ or ‘competency trap’, where mutual positive feedback between experience and competence leads to a situation where competence in an activity leads to success, which leads to more experience with the activity, in turn leading to greater competence (March, 1994). A pathology of knowledge generation results where the division of labor is so fragmented that individual tasks are too narrow to allow useful knowledge to be generated (Klein, 2000; Chesbrough and Kusunoki, 2001). 20.3.5

Pervasive Uncertainty

Yet another problem arising from the dispersion of knowledge is that it leads to pervasive uncertainty, a strong form of uncertainty. In a decisionmaking context, uncertainty refers to a situation where a decision can lead to more than one possible consequence. In order to be more precise, different kinds of uncertainty need to be distinguished.14 In the realm of risk, all possible consequences and the likelihood of each consequence are known. In a situation of uncertainty there is a set of possible specific outcomes, but the likelihood of each outcome is initially unknown. The general strategy in such a situation is therefore to increase the amount of information, improving the basis for estimation of probabilities and their accuracy. As opposed to this conceptualization of uncertainty, some authors argue that also a much stronger, pervasive form of uncertainty must be considered (Knight, 1921; March and Simon, 1958). For those authors, uncertainty corresponds to an absence of measurable probabilistic knowledge. Such situations are so ill structured that the possible outcomes will remain unknown despite any attempt to remedy the situation (‘pervasive uncertainty’). One possible reason is that the underlying meaning of some information is not clear, or leads to ambiguity in interpretation, a condition known as ‘equivocality’ (Daft and Lengel, 1986). Another characteristic of situations of pervasive uncertainty is that decision makers cannot ex ante specify all relevant alternatives or outcomes (Minkler, 1993). Whereas under risk, the possible outcomes and their

Distributed knowledge and its coordination 469 probabilities are known, and under uncertainty, the possible outcomes are in principle known but their probabilities can only be subjectively estimated, under pervasive uncertainty, neither the possible outcomes nor their probabilities are known. Therefore the basis for taking decisions is not clear. In this way, the dispersion of knowledge increases the difficulty of taking decisions. Furthermore, it also leads to contract and monitoring problems (Minkler, 1993). Obviously, in order to be able to deal with contingencies, contracts need to possess clauses that in some way describe them. But in order for that to happen, one needs to know (all) the potential alternatives, at least their type (e.g. what kind of technological development can be expected, even if the precise application to a product might not yet be clear). This is severely impeded by pervasive uncertainty, as it prohibits the principal from formulating statistical distributions over agent activities – even agents themselves are initially ignorant of their own alternatives. In such a case, ‘no incentive contract, based on the principal’s knowledge of agent alternative activities, could be written’ (Minkler, 1993, p. 579). Pervasive uncertainty increases with increasing dispersion of knowledge. Three drivers contribute to this correlation. First, dispersion of knowledge leads to costs of intelligence gathering. The more distributed knowledge is, the higher this cost will be. Second, complexity increases information-processing cost (it becomes more difficult and resourceintensive to identify, interpret and understand information). Complexity increases with increasing specialization (which is coupled to the dispersion of knowledge). Third, a relatively high degree of specialization means that there will be few overlaps at the level of task-related knowledge. The effect on agents is that of lack of overview (imagine a scientific field to which you are an outsider), increasing with increasing specialization (linked to the dispersion of knowledge). All three drivers bring about the situation that agents are not capable of identifying all alternatives for solving a problem, nor to attach probabilities to the performance results of choosing the different alternatives – the situation of pervasive uncertainty. In summary, this section has cast light on the following effects of the dispersion of knowledge on organizations. (a) The dispersion of knowledge causes coordination costs. Because coordination costs are increasing with increasing division of labor (and dispersion of knowledge), they work to offset decreasing average unit costs (economies of scale) (Stigler, 1951). (b) There are endogenous limits to the dispersion of knowledge, not just because of rising coordination cost, but also because of a more profound problem: beyond a certain threshold, specialization means that the overlap between the underlying knowledge held by two agents is so small that transferring task-related knowledge is impaired. Transferring such

470

Handbook of knowledge and economics

knowledge, however, is crucial for dealing with the dispersion of knowledge, as explained in more detail in the next section. (c) Another causal mechanism leading to endogenous limits to the dispersion of knowledge is that the more distributed knowledge is, the more knowledge gets lost in applying authority as coordination mechanism (Grandori, 2002). The argument, as well as its critical extension in Foss and Foss (2003), provides an important structure for investigating further how this limit could be attenuated. (d) Dispersion of knowledge risks developing an imbalance towards exploitation (efficiency), to the detriment of exploitation. (e) Dispersion of knowledge gives rise to a condition of pervasive uncertainty. Pervasive uncertainty is a strong form of uncertainty (very different from risk), which means that neither the possible outcomes nor their probabilities are known. It is important to recognize that dispersion of knowledge gives rise to this strong form of uncertainty, because the adequate strategies for dealing with this form of uncertainty differ from those for dealing with other forms of uncertainty (see Becker and Knudsen, 2005).

20.4 COORDINATING DISTRIBUTED KNOWLEDGE Hayek (1937, 1945) was the first to explicitly point to the coordination of distributed knowledge. As is well known, his argument was that there are limits to coordinating such knowledge, which required the decentralization of decision making in an economy. The present section approaches the question how distributed knowledge can be coordinated. In the literature focused on knowledge, the coordination of distributed knowledge is often discussed under the label ‘knowledge integration’. While many authors refer to ‘knowledge integration’, it is, however, not entirely clear what precisely the term means. In the knowledge-based approach to the theory of the firm,15 to integrate knowledge means to combine (Kogut and Zander, 1992; Almeida et al., 2002), mesh (Demsetz, 1988, 1991; Almeida et al., 2002) or synthesize (Kogut and Zander, 1992) distributed knowledge. It is illuminating to note that Grant himself, who has popularized the notion of knowledge integration with his two 1996 articles (Grant, 1996a, 1996b), has more recently used the term ‘knowledge building’ (Almeida et al., 2002). As he explains, the idea behind the term ‘knowledge building’ is that the challenge of managing knowledge consists not only in transferring knowledge, but also in ‘its development through the combination of the transferred knowledge with the recipient’s existing knowledge’ (ibid., p. 148). The notion of knowledge integration has thus been extended to include not just a simple ‘addition’ of ‘knowledge

Distributed knowledge and its coordination 471 elements’, but also the aspect of further developing the knowledge that has been combined. The picture painted by this literature is consistent with that painted by the new product development literature (a literature concerned with a practical instance of knowledge integration). In an article focused on the effectiveness of different mechanisms for integrating marketing and R&D, Leenders and Wierenga (2002, p. 306), for instance, note that there are two extreme views of integration: At one side of the spectrum, there are studies that define integration simply as communication frequency or interaction frequency. These studies do not take into account what type of information is shared and how the information is used. At the other side of the range, we have the composite view of integration, which implies that integration refers to a multiple of factors, of which communication and cooperation are the key elements.

Considering an even wider literature, the term ‘knowledge integration’ is used not only in the management literature, but also in the education, psychology and information systems literatures. In the education literature, for instance, knowledge integration is defined as the establishment of durable conceptual and associative links between knowledge stocks and levels of representation that were previously isolated from each other. Through such integration a new system emerges, which allows a more or less flexible access to the individual knowledge elements, dependent on its internal network structures and coherence. (Krist, 1999, p. 197)

Key terms in this definition are ‘establishing links’, ‘previously isolated’ and ‘access’. The links that are established form a system, a whole. Some authors also use the term ‘net’ or ‘network’ for such a system (cf. Schmalhofer et al., 1997). Cognitive effort is one of the factors that determine how easy or difficult it is to access knowledge elements in the system. Cognitive effort, in turn, depends on the internal network structure and the coherence of the network (Krist, 1999). The psychology and education literature has thus confirmed, and sharpened, the conception of knowledge integration in the management literature. Now, our understanding of knowledge integration in the management literature is quite limited when it comes to the question how precisely integration takes place. What happens when knowledge integration takes place? An important insight is that in the case of knowledge integration by students, some active input, notably reflection, is required: Knowledge integration requires students to expand their repertoire of ideas, but unless those ideas are reflected upon, they cannot be linked to and reconciled

472

Handbook of knowledge and economics

with current ideas. Students are capable of doing this kind of reflection, but may need scaffolding. Scaffolding – here in the form of reflection prompts – can help students be autonomous integrators of their knowledge. (Davis and Linn, 2000, p. 819)

Such reflection can be elicited, and supported, by means such as prompts or scaffolding. Providing such triggers and support reflection has been shown to increase knowledge integration (Davis and Linn, 2000). A more fine-grained and comprehensive model of knowledge integration would thus consist of two steps: in a first step, reflecting on the nature of ideas held, and in a second step, linking, connecting and organizing all ideas into a coherent perspective (Linn, 1996). The latter step also includes reorganizing links among ideas (or knowledge) (Davis and Linn, 2000). Summing up, an insight that carries through several literatures is that reflection is an important component of knowledge integration.16 With this background, we now turn to an overview of knowledge integration mechanisms. 20.4.1

Knowledge Integration Mechanisms – Organizational Level

The knowledge integration mechanisms to be found in the literature can be located on four different levels: the level of the organization, the individual, processes and artefacts. 20.4.1.1 Authority The integration mechanism that has attracted most attention is organization structure (Griffin and Hauser, 1996). In the literature known as the ‘knowledge-based approach’ to the theory of the firm, firms are seen as providing the integration of specialist knowledge (Grant, 1996a). From the classic writings of organization theory onwards, organization structure has been seen to address, among others, the (re-)integration of the division of labor (Lawrence and Lorsch, 1967; Galbraith, 1973). For the knowledge-based approach, firms create the conditions for knowledge integration, for instance by providing incentives designed to foster coordination between individual specialists. To provide the conditions for knowledge integration is not the same as providing knowledge integration, however. What are the mechanisms underlying the integration of knowledge in firms? The main underlying mechanism behind organization structure, which effects knowledge integration, is authority. Knowledge is integrated by putting specialists under the authority of a manager who directs their action. Integration results because such direction is unitary, that is, exercised by one person, but spans a number of specialists. Note that direction of action does not need to be exercised by a person that gives

Distributed knowledge and its coordination 473 orders, but can also be effected by rules, norms and so on that are binding for the persons who are supposed to follow them (this is why Grant, 1996a cites ‘rules and directives’ as an integration mechanism). 20.4.1.2 Structure as a coordination mechanism Structure traditionally plays a prominent role in explanations of coordination in economics, see for instance the structure–conduct–performance paradigm (SCP) in industrial organization, or the emphasis on governance structures in transaction cost economics (TCE) (Williamson, 1975, 1985). While structure plays a prominent role in explaining coordination, structures have been interpreted as having very different ‘contents’. In the literature, one can find structures of transactions, interactions, firms, individuals, property rights, expectations about behavior and so forth. Let us briefly consider some of these possibilities. 20.4.1.2.1 Routines as structures of interaction (recurrent interaction patterns) As Pentland and Rueter have written, routines ‘occupy the crucial nexus between structure and action, between the organization as an object and organizing as a process’ (Pentland and Rueter, 1994, p. 484). Due to the stability involved, patterns are recognizable – this is the static aspect of structures. At the same time, they are structures of (inter-) actions, and thus dynamic.17 Organizational routines have long been credited as contributing to coordination (Stene, 1940; Nelson and Winter, 1982; March and Olsen, 1989; Gersick and Hackman, 1990; Coriat, 1995; Dosi et al., 2000). In fact, Stene (1940, p. 1129) proposed that the ‘[c]oördination of activities within an organization tends to vary directly with the degree to which essential and recurring functions have become part of the organization routine’. The coordinative power of routines derives from several sources: from their capacity to support a high level of simultaneity (Grant, 1996a); from giving regularity, unity and systematicity to practices of a group (Bourdieu, 1992); from making many simultaneous activities mutually consistent (March and Olsen, 1989); from providing each of the actors with knowledge of the behavior of the others on which to base his own decisions (Simon, 1947; cf. Stene, 1940); from providing instructions in the form of programs, and from establishing a truce (see Section 20.3.2) (Nelson and Winter, 1982). It has been claimed that, as coordinating devices, routines can be more efficient than contracts, so that they could even substitute for contracts and make them increasingly unnecessary as relationships mature (Langlois and Robertson, 1995). Empirical research has started to shed some light on the effect of routines on coordination. A study of the investment manuals of a sample of major Swedish firms noted

474

Handbook of knowledge and economics

that standards (and standardized routines) are especially influential for exerting control (Segelod, 1997), one way to bring about coordination. A possible reason is that routine behavior is easier to monitor and measure than non-routine behavior (cf. Langlois, 1992). The more standardized, the easier to compare. The easier to compare, the easier to control. Knott and McKelvey (1999) conducted an empirical test that compares the relative value of residual claims and routines in generating firm efficiency, using a large sample of firms from the US quick printing industry. The study concludes that routines can be more efficient for coordination and control than residual claims, contrary to principal–agent theory, which has propagated residual claims as the most efficient solution to the monitoring problem. Focusing on the question ‘What is the best way for organizations to achieve coordination?’, Gittell (2002) analyzes the performance effects of routines (and other coordination mechanisms such as boundary spanners and team meetings). The study analyzes questionnaire data on care provider groups in acute-care hospitals. It finds that the performance effects of routines were mediated by relational coordination: routines work by enhancing interactions among participants, which was found to have positive performance effects. 20.4.1.2.2 Economic architecture In one of the core articles on distributed knowledge, Tsoukas (1996) identified ‘developing ways of interrelating and connecting the knowledge each individual has’ as one key to achieving coordinated action in the face of distributed knowledge. Likewise, Demsetz (1998, p. 157) also argues that the ‘[e]conomic organization, including the firm, must reflect the fact that knowledge is costly to produce, maintain, and use’. Both the pattern of dispersion of knowledge and the structure of the linkages between the distributed knowledge elements matter for bringing about coordination of distributed knowledge. As Foss and Foss (2003, p. 1) point out, little systematic analysis of how distributed knowledge and economic organization relates exits. The little systematic analysis that has been carried out indicates that it is worthwhile to explore this link further. In their article, Rulke and Galaskiewicz (2001) found that for understanding performance implications of different dispersion patterns of knowledge, it was necessary to take the architecture of communication linkages and decision rights into account. For example, they found that ‘group performance depends on not only the information resources available to the group, but also the processes or structures which groups use to utilize their resources’ (Rulke and Galaskiewicz, 2001, p. 612). This leads them to propose ‘that group structure can modify the effect of knowledge distribution on group performance’ (ibid., p. 613). Indeed, the two-layered analytical framework of the dispersion pattern of

Distributed knowledge and its coordination 475 knowledge on the one hand, and the structure of the organization of links between the actors in an organization on the other, seems to be a fruitful one. Various ways to operationalize structure exist. Rulke and Galaskiewicz (2001), for instance, used the intensity of ties among group members and the degree to which networks of relations were hierarchical or flat. Social network analysis views network structure, for instance, with different types of centrality measures (Wasserman and Faust, 1994). A different approach is to model organizations as economic architectures. Building on the contributions of Sah and Stiglitz (1986, 1988), Christensen and Knudsen (2008, 2010) model n-member architectures as ‘a collection of members, a collection of channels through which the members can pass information or control to each other, and a set of dynamic rules that help define the flow of information or control and thus help define their decision rights’ (Knudsen and Eriksen, 2002, p. 10). This approach also takes into account the rules that decide on how organization members pass on information to each other, who has the right to take decisions on accepting or discarding projects, and so on. The basic idea in this modeling approach is to consider an initial portfolio of projects, of which the organization has to choose some to invest in. The projects enter the organization at particular points of the organization, that is, particular agents, who can have the right to decide that the project should be ultimately accepted (without consulting other agents), ultimately dumped (without consulting other agents), or to give one of the two recommendations and pass the project on to another agent (Christensen and Knudsen, 2008, 2010). The point is that the way information traverses an economic system will influence the errors of judgment are expressed at the system level (Christensen and Knudsen, 2010). Because agents experience limits to rationality, they make errors of judgment. They reject projects with positive value (Type-I error) and accept projects with negative value (Type-II error) (ibid.). This approach nicely fits the problem posed by distributed knowledge (although the model does not capture the dispersion pattern of knowledge at this point of time): it can be argued that one important reason why errors of both types occur is that even though the knowledge to minimize both types of error is present in the organization, its dispersion blocks its use in the individual decisions.18 The other way round, the question then becomes: keeping the knowledge held by the individual agents constant, what impact does the architecture have on error rates of the collective decision taken by the architecture? More practically, if the structure of a decision-making system, its architecture, affects the emergence of error at the collective level, what impact do different architectures have? There are two basic, and a host

476

Handbook of knowledge and economics

of intermediate, architectures. The two basic structures are hierarchy and polyarchy. Hierarchy represents a serial processing in which a project can be rejected by each agent and is only accepted by the organization if all agents accept it. Polyarchy represents parallel processing. Each agent can accept a project of behalf of the organization, but cannot reject a project by herself, so that an ultimate rejection requires that each member rejects a project (Christensen and Knudsen, 2010). The economic architecture approach is a powerful one, the main reason underlying its strength being that a general framework for n-member architectures has been provided (Christensen and Knudsen, 2008, 2010). It provides analytical solutions to questions such as the implications of any n-member architecture on error rates. In the context of the present chapter, this allows getting a firm grip not only on the description of the pattern of linkages between actors, but also on the performance implications of such patterns. The fact that performance is expressed as error rates lends itself well to analyzing the problem in the focus in the present chapter. According to the distributed knowledge argument reviewed above, distributed knowledge means precisely that knowledge gaps occur, which in turn will imply that when decisions have to be taken, not all the knowledge that is present in the organization can necessarily be drawn on in the decision. A different perspective to take would be that of knowledge integration: focusing on structure as a knowledge integration mechanism, what are the performance implications of the different structures? No other analytical framework seems to be able to give answers to these questions with the same precision. 20.4.1.2.3 Transactive memory systems In the organization science literature, the perspective of transactive memory systems has recently been developed for understanding how expertise is coordinated in groups (Wegner, 1986; Liang et al., 1995; Hollingshead, 1998; Lewis, 2003, 2004). A transactive memory system is a shared system for encoding, storing, retrieving and communicating information that develops naturally in relationships and in groups (Majchrzak et al., 2007, p. 151). One important part of a transactive memory system is an indexing system that tells members who has what expertise. The indexing system is created jointly by team members as they gain experience together an develop similar understandings of who knows what. The level of development of a transactive memory system can be captured by (i) memory (or expertise) specialization, (ii) credibility, that is, beliefs about the reliability of members’ expertise; and (iii) task (or expertise) coordination: the ability of team members to coordinate their work efficiently based on their knowledge of who knows what in the group (Lewis, 2003; Majchrzak et al., 2007).

Distributed knowledge and its coordination 477 20.4.2

Knowledge Integration Mechanisms – Individual Level

A completely different way to achieve knowledge integration is by means of individual skills, for instance the skill to fill in knowledge gaps created by the dispersion of knowledge. Rather than by transferring the ‘missing’ knowledge, knowledge gaps can also be dealt with by filling in, such as in ‘inventing around’. Egidi suggests that ‘in reality, individuals . . . have “incomplete” knowledge, and they are able to complete it by recreating its missing components’ (Egidi, 1996, p. 307). Collins and Kusch (1998) argue that this strategy – at least to a certain extent – is applied by everyone in everyday life. Think about your own capacity to ‘repair’ misspellings by still recognizing the meaning. In a similar way, one can assume that software users typically have certain competences (for instance, understanding the meaning of a word even despite a missing letter) – an example of the competence to fill in knowledge gaps. The higher this competence, the less it is necessary to possess the knowledge required in order to fulfill the task. This skill is closely related to learning. Rather than integrating two bodies of specialist knowledge, knowledge is integrated by one specialist learning what the other knows. Because in many concrete problems, only a particular part of the body of specialist knowledge is needed, that might often be feasible. 20.4.3

Knowledge Integration Mechanisms – Artifacts

Artifacts are another mechanism that fosters knowledge integration. First, artifacts represent knowledge. This is why reverse engineering can work.19 Second, artifacts serve as reference for a group, each member of which may hold different knowledge. The artifact provides a common structure that the individuals can use to assemble their inputs into a problemsolving process (Baba and Nobeoka, 1998). For instance, even though a heterogeneous team of specialists (engineers, controllers, marketing experts) has a very small overlap of shared knowledge, and completely different approaches to the same problem (technical feasibility, efficiency, saleability as objectives), a physical prototype of the product can help structure the discussion, and thus aid the integration of the different knowledge inputs. Third, due to the fact that artifacts serve as reference and focal points, they improve communication and coordination among the holders of distributed knowledge, thus facilitating knowledge integration (D’Adderio, 2001; Baba and Nobeoka, 1998). Interestingly, not only real objects have this feature – virtual objects (objects created in computer simulations) have a very similar capacity (D’Adderio, 2001; Baba and Nobeoka, 1998).

478

Handbook of knowledge and economics

20.4.4

Knowledge Integration Mechanisms – Processes

While structures and links are important, it also matters what happens within these structures. Processes are therefore another type of knowledge integration mechanisms. Processes can be influenced and patterned by various means, among which are: ● ● ● ● ● ● ● ● ●

physical facilities design (Griffin and Hauser, 1996) personnel movement (ibid.) informal social systems (ibid.) incentives and rewards (ibid.) formal integrative management processes (ibid.) staff relocation (ibid.; Ghoshal and Gratton, 2002) structured career paths (ibid.)20 social integration through collective bonds of performance, such as peer assistance and joint responsibility (ibid.) emotional integration through shared identity and meaning (ibid.).

Yet another knowledge integration mechanism identified in the literature is organizational routines. Routines are credited with storing the knowledge of an organization. For Nelson and Winter (1982, p. 99), ‘the routinization of activity in an organization constitutes the most important form of storage of the organization’s specific operational knowledge’ (see also Winter, 1987, 1995; Dosi et al., 1992; Levitt and March, 1988; Miner, 1994; Teece and Pisano, 1994; Hodgson, 1998; Phillips, 2002; Zollo and Winter, 2002; Lillrank, 2003). What sets routines as knowledge repository apart from other kinds of knowledge repositories such as databases and documents is that routines are widely credited with being able to store tacit knowledge (Winter, 1987, 1994; Teece et al., 1994; Hodgson, 1998, 1999; Cohendet et al., 1999; Knott, 2003). In conclusion, the section has been characterized by two main insights. First, knowledge integration is not just an ‘addition’ of distributed knowledge elements. It also encompasses reflection, and welding the elements into some kind of whole (e.g. a ‘tool’ or method for solving a problem). Individual knowledge integration mechanisms therefore need to contribute to that. Second, knowledge integration mechanisms are located on multiple levels (organizational, individual, artifacts, processes). While knowledge integration mechanisms on each level contribute to give rise to knowledge integration, it is likely that in order to understand how knowledge integration is effected, it is necessary to take the whole set of knowledge integration mechanisms into consideration. Viewing the set of knowledge integration mechanisms as a nested hierarchy might be

Distributed knowledge and its coordination 479 useful. Mapping such sets in concrete organizations then becomes a task for further research.

20.5 PRACTICAL AND THEORETICAL IMPLICATIONS Distributed knowledge is a topic that lies at the very heart of both economics and management, as the works of Adam Smith (1776) and Frederick Taylor (1912) and their legacy testify. Distributed knowledge represents the driver of some of the most fundamental issues in both fields, such as the industry structure, firm boundaries, the use of price, authority or other coordination mechanisms, the motivation of employees, the retrieval and creation of knowledge in firms or the organization of innovation efforts. Despite its fundamental importance, the topic of distributed knowledge and its coordination was basically assumed away until Hayek (1937, 1945) brought the topic up, and until economics took the issue of asymmetric information (and information cost) more seriously. Even then, however, the problem was that the division of labor and the dispersion of knowledge are coupled, but are neither always closely coupled, nor can be reduced to one dimension. This issue has not yet been completely resolved (see Martens, 1999), and has hampered progress in understanding the coordination of distributed knowledge. One important topic for further research is to precisely define the relationship between the division of labor and the dispersion of knowledge (see Brusoni et al., 2002). This is the first source of complexity in the analysis of distributed knowledge. A second complexity concerns the effects on organizations. Not only is the dispersion of knowledge (necessarily) coupled to some extent with the division of labor, and thus the effects of both arise ‘at the same time’. That makes disentangling the sources of effects on organizations (such as decreasing average unit costs etc.) both more important – in order to understand the causal drivers – and more difficult. The problem is aggravated further by the fact that, as shown in Section 20.2, the dispersion of knowledge itself also has a whole set of effects on organizations, both desirable and less desirable ones. The two challenges for theory are therefore (i) to identify the effects of the dispersion of knowledge, and disentangle them from the effects of the division of labor, on the level of the causal drivers, and (ii) to develop an overarching framework that encompasses the whole set of effects, in order to consider also counteracting effects. An example of the latter problem brings us to the field of management, and to implications for theory. Even though the knowledge level (as

480

Handbook of knowledge and economics

opposed to the level of labor) has received increasing attention over the last years, it still has not received the attention it deserves. For instance, in arguing for the advantages of modularization, clearly, some of the effects of the dispersion of knowledge have not been afforded the weight they deserve. The implication is that some of the advice management theory offers on modularization leads to short-term gains, but to detrimental long-term effects in the knowledge dimension (such as difficulties in assessing new technological developments; see Zirpoli and Becker, 2011a, 2011b). The first ‘result’ of increased attention to the knowledge dimension in the management literature was increased interest in knowledge integration. In reviewing the literature on the various knowledge integration mechanisms identified, it becomes evident that the importance of the organization (structure of connections) between the distributed knowledge elements and their holders is the most important dimension for knowledge integration. More than, for instance personal characteristics, the structure that links the members of an organization has a substantial impact on how well they can integrate the distributed elements of knowledge that are held within the firm when a decision has to be made or a problem to be solved. Clearly, it is not the fact that the knowledge to solve the problem is ‘present’ in the firm, but the fact that it can be mobilized in such a way that ‘all’ the knowledge required to solve the problem is actually applied to the problem that decides on the quality of the decision or problem solving. It therefore seems a very fruitful approach to use the extant methods for analyzing structures of links, such as network analysis, or the economic architecture approach. The latter has the advantage that analytical results on the performance of different architectures are available, and that the performance measure (error rates) nicely maps on the situation of decision making or problem solving in a situation of distributed knowledge. While that is a huge research area, and a very fruitful one, the next issue to be tackled is to investigate the link between the patterns of dispersion of knowledge (structure of the knowledge ‘elements’) and the architectures of connections between those that are applied for integrating such knowledge. What different types of relationship can be observed? What different effects on the innovativeness of firms, or the ability to mobilize and integrate distributed knowledge do different types of relationship have? What are the implications of different types of relationships on industry structure, and on the evolution of firms in an industry? Recent results of research that applies the two-pronged distinction of patterns of knowledge dispersion and architecture of links, for instance, show that in biotechnology, firms tend to have similar modular knowledge over time (the pattern of knowledge dispersion) but exhibit an increasing differentiation in their architecture (Nesta and Dibiaggio, 2002). This result is a great example of

Distributed knowledge and its coordination 481 the potential of the kind of research that can greatly illuminate the coordination of distributed knowledge in the close future.

NOTES * 1. 2. 3.

4.

5. 6.

7. 8. 9. 10. 11. 12. 13.

14. 15. 16. 17.

Many thanks to Patrick Cohendet and Francesco Zirpoli for discussion and helpful comments. The usual disclaimer applies. For another example, imagine a cooperative of plumbers. All of them have to be certified plumbers (in most countries at least in order to be able to install gas, plumbers need to have an official certification) and therefore the same task-related knowledge. See also Moorman and Miner’s (1997, p. 103) definition of dispersion as ‘the degree of consensus or shared knowledge among new product team participants’. Similarly, in the case of franchisees such as the operators of McDonald’s restaurants and the cooperative of plumbers. Note that standardized ‘knowledge packages’, whose acquisition is being certified by a central authority, are crucial for creating common (as well as shared) knowledge. Throughout the chapter, I use the terms ‘distributed knowledge’ and ‘knowledge dispersion’. As defined in the text, these terms do not differ. The term ‘knowledge dispersion’ has been preferred to the term ‘knowledge distribution’, however, because the latter could also be misunderstood as the process of distributing (transferring) knowledge. For a definition of architecture see Section 20.5. Because the systems designer/coordinator has to design architecture, interfaces and standards without having much knowledge about the individual modules, this means in practice that the design of the architecture, interfaces and standards is very much a trialand-error process (Foss and Foss, 2003, p. 7). It also means that the knowledge of the systems designer/coordinator poses limits to the integrating power of the architecture. The question, however, is how tightly the division of labor and the division – or dispersion – of knowledge are linked. For more on this question see Brusoni et al. (2002), Becker and Zirpoli (2003, 2005) and Zirpoli and Becker (2011b). With some qualifications (see Vassilakis, 1987), economies of scale are equivalent to increasing returns to scale. In Coase’s terms: ‘Why is not the whole economy organized as one big firm?’ A different argument pointing in the same direction was supplied by Kaldor and Robinson – they pointed to increased cost and difficulty of coordination in the firm, an argument elaborated by Penrose. This has also been called ‘cognitive friction’: a specific type of information cost, namely the time costs of acquiring the appropriate (largely tacit) cognitive categories for relatively undistorted reception of signals in a given context (Knudsen and Foss, 1999, p. 2). At least, this holds for the case where controlling whether the agent has complied with the directions requires specialist knowledge about the agent’s operations. Problems for authority as a coordination mechanism are further aggravated by the fact that much of the knowledge that matters is tacit and difficult to transfer, adding to problems in transferring the knowledge required to the person invested with authority. Further problems, which, however, are not immediately influenced by increase in division of labor, are that not all elements and interdependencies are observable (in particular the latter) (Grandori, 2002), and that authority needs to be accepted by subordinates in order to be acted upon (Barnard, 1938). This section draws heavily on Becker and Knudsen (2005). See for instance Grant (1996a); Kogut and Zander (1992, 1993); Spender (1996). Consequently, it matters whether structures foster such reflection. Note that there is a debate about whether routines refer to behavior (as in the definition

482

18. 19.

20.

Handbook of knowledge and economics given in the text), or to what generates behavior, for example dispositions (see Hodgson and Knudsen, 2010). In this argument, the comparison, or ‘zero-error level’, is not with a state of omniscience, but rather with that proportion of the error rates due to dispersion and linkages of knowledge. For instance, a particular table represents a solution to the problem of sustaining objects at a certain height against gravitation. As we know, this can be done by four vertical legs, by two legs with a broader base, by one leg ending in a very broad base and so forth. An observer with the appropriate background knowledge (such as of gravitational forces) can infer that knowledge from looking at the table. The example becomes less evident when one thinks of the proportions of the different parts of a table that are required in order to render it stable. In his study of a Japanese auto manufacturer, Takeishi (2002) has cast much light on one of these knowledge integration mechanisms, that is, personnel rotation policies. His findings indicate that the knowledge integration capacity of personnel rotation depends on keeping a tight balance. On the one hand, rotating individual engineers across many components helps build architectural knowledge through hands-on experiences. On the other hand, rotating engineers across components quickly may impede their accumulation of component-specific knowledge (Takeishi, 2002, p. 16). For this reason, the firm also established other mechanisms to improve and maintain component-specific knowledge: the range of rotation was limited, so that componentspecific knowledge obtained in previous assignments remained somewhat relevant for a newly assigned component. Furthermore, a new career path was created in which individual engineers could stay with the same component over a very long period of time as a specialist (ibid., p. 17).

REFERENCES Almeida, Paul, Song, Jaeyong and Grant, Robert M. (2002), ‘Are firms superior to alliances and markets? An empirical test of cross-border knowledge building’, Organization Science, 13 (2), 147–61. Argote, Linda (1999), Organizational Learning: Creating, Retaining, and Transferring Knowledge, Boston, MA: Kluwer. Argote, Linda and Ingram, Paul (2000), ‘Knowledge transfer: a basis for competitive advantage in firms’, Organizational Behavior and Human Decision Processes, 82 (1), 150–69. Baba, Yasunori and Nobeoka, Kentaro (1998), ‘Towards knowledge-based product development: the 3-D CAD model of knowledge creation’, Research Policy, 26 (6), 643–59. Barnard, Chester I. (1938/1968), The Functions of the Executive, Cambridge, MA: Harvard University Press. Baldwin, Calriss Y. and Clark, Kim B. (1997), ‘Managing in an age of modularity’, Harvard Business Review (September–October), 84–93. Becker, Markus C. and Knudsen, Thorbjørn (2005), ‘The role of routines in reducing pervasive uncertainty’, Journal of Business Research, 58 (6), 746–57. Becker, Markus C. and Zirpoli, Francesco (2003), ‘Organising new product development: knowledge hollowing-out and knowledge integration. The Fiat Auto case’, International Journal of Operations and Production Management, Special Issue, 23 (9), 1033–61. Becker, Markus C. and Zirpoli, Francesco (2005), ‘Knowledge- and task-partitioning in the automotive industry: coordination, governance and learning in new product development’, International Journal of Automotive Technology Management, 5 (2), 137–45. Bourdieu, Pierre (1992), Outline of a Theory of Practice, Cambridge: Cambridge University Press. Brusoni, Stefano, Prencipe, Andrea and Pavitt, Keith (2002), ‘Knowledge specialization,

Distributed knowledge and its coordination 483 organization coupling, and the boundaries of the firm: why do firms know more than they make?’, Administrative Science Quarterly, 46 (4), 597–625. Chesbrough, Henry W. and Kusunoki, Ken (2001), ‘The modularity trap: innovation, technology phase shifts and the resulting limits of virtual organizations’, in Ikujiro Nonaka and David Teece (eds), Managing Industrial Knowledge: Creation, Transfer and Utilization, London: Sage, pp. 202–30. Christensen, Michael and Knudsen, Thorbjørn (2008), ‘Entry and exit decisions in flexible teams’, Journal of International Business Studies, 39 (8), 1278–92. Christensen, Michael and Knudsen, Thorbjørn (2010), ‘Design of decision-making organizations’, Management Science, 56 (1), 71–89. Coase, Ronald H. (1937), ‘The nature of the firm’, Economica, 4 (16), 386–405. Cohen, Wesley M. and Levinthal, Daniel A. (1990), ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35 (1), 128–52. Cohendet, Patrick and Llerena, Patrick (2003), ‘Routines and incentives: the role of communities in the firm’, Industrial and Corporate Change, 12 (2), 271–97. Cohendet, Patrick, Kern, Francis, Mehmanpazir, Babak and Munier, Francis (1999), ‘Knowledge coordination, competence creation and integrated networks in globalised firms’, Cambridge Journal of Economics, 23 (2), 225–41. Collins, Harry and Kusch, Martin (1998), The Shape of Actions – What Humans and Machines Can Do, Cambridge, MA: MIT Press. Coriat, Benjamin (1995), ‘Variety, routines and networks: the metamorphosis of Fordist firms’, Industrial and Corporate Change, 4 (1), 205–27. Cramton, Catherine Durnell (2001), ‘The mutual knowledge problem and its consequences in geographically dispersed teams’, Organization Science, 12 (3), 346–71. D’Adderio, Luciana (2001), ‘Crafting the virtual prototype: how firms integrate knowledge and capabilities across organisational boundaries’, Research Policy, 30 (9), 1409–24. Daft, Richard L. and Lengel, Robert H. (1986), ‘Organizational information requirements, media richness and structural design’, Management Science, 32 (5), 554–71. Davis, E.A. and Linn, M.C. (2000), ‘Scaffolding students’ knowledge integration: prompts for reflection in KIE International’, Journal of Science Education, 22 (8), 819–37. Demsetz, Harold (1988), ‘The theory of the firm revisited’, Journal of Law, Economics and Organization, 4 (1), 141–61. Dosi, Giovanni, Nelson, Richard and Winter, Sidney G. (2000), ‘Introduction: the nature and dynamics of organizational capabilities’, in Giovanni Dosi, Richard Nelson and Sidney G. Winter (eds), The Nature and Dynamics of Organizational Capabilities, Oxford: Oxford University Press, pp. 1–22. Dosi, Giovanni, Teece, David J. and Winter, Sidney (1992), ‘Toward a theory of corporate coherence: preliminary remarks’, in Giovanni Dosi, Renato Giannetti and Pier Angelo Toninelli (eds), Technology and Enterprise in a Historical Perspective, Oxford: Clarendon Press, pp. 185–211. Dougherty, Deborah (1992), ‘Interpretive barriers to successful product innovation in large firms’, Organization Science, 3 (2), 179–202. Egidi, Massimo (1996), ‘Routines, hierarchies of problems, procedural behaviour: some evidence from experiments’, in Kenneth Arrow (ed.), The Rational Foundations of Economic Behaviour, London: Macmillan, pp. 303–33. Foss, Kirsten and Foss, Nicolai J. (2003), ‘Authority in the context of distributed knowledge’, DRUID Working Paper No. 03-08, http://www3.druid.dk/wp/20030008.pdf. Foss, Nicolai and Eriksen, Bo (1995), ‘Competitive advantage and industry capabilities’, in Cynthia Montgomery (ed.), Resource-based and Evolutionary Theories of the Firm – Towards a Synthesis, Dordrecht: Kluwer, pp. 43–69. Fransman, Martin (1998), ‘Information, knowledge, vision and theories of the firm’, in Giovanni Dosi, David Teece and Josef Chytry (eds), Technology, Organization, and Competitiveness – Perspectives on Industrial and Corporate Change, Oxford: Oxford University Press, pp. 147–92. Galbraith, Jay R. (1973), Designing Complex Organizations, Reading, MA: Addison-Wesley.

484

Handbook of knowledge and economics

Gersick, Connie J.G. and Hackman, J. Richard (1990), ‘Habitual routines in task-performing groups’, Organizational Behaviour and Human Decision Processes, 47 (1), 65–97. Ghoshal, Sumantra and Gratton, Lynda (2002), ‘Integrating the enterprise’, MIT Sloan Management Review, Fall, 31–8. Gittell, Jody Hoffer (2002), ‘Coordinating mechanisms in care provider groups: relational coordination as a mediator and input uncertainty as moderator of performance effects’, Management Science, 48 (11), 1408–26. Grandori, Anna (2002), ‘Cognitive failures and combinative governance’, Journal of Management & Governance, 5 (3–4), 252–60. Grant, Robert M. (1996a), ‘Toward a knowledge-based theory of the firm’, Strategic Management Journal, 17 (Winter Special Issue), 109–22. Grant, Robert M. (1996b), ‘Prospering in dynamically-competitive environments: organizational capability as knowledge integration’, Organization Science, 7 (4), 375–87. Griffin, Abbie and Hauser, John R. (1996), ‘Integrating R&D and marketing: a review and analysis of the literature’, Journal of Product Innovation Management, 13 (3), 191–215. Hayek, Friedrich August von (1937), ‘Economics and knowledge’, Economica, 4 (13), 33–54. Hayek, Friedrich A. von (1945), ‘The use of knowledge in society’, American Economic Review, 35 (4), 519–30. Heath, Chip and Staudenmeyer, N. (2000), ‘Coordination neglect: how lay theories of organizing complicate coordination in organizations’, Research in Organizational Behaviour, 22, 155–93. Hodgson, Geoffrey (1998), ‘Evolutionary and competence-based theories of the firm’, Journal of Economic Studies, 25 (1), 25–56. Hodgson, Geoffrey M. (1999), Evolution and Institutions – On Evolutionary Economics and the Evolution of Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Hodgson, Geoffrey and Knudsen, Thorbjørn (2010), Darwin’s Conjecture: The Search for General Principles of Social and Economic Evolution, Chicago, IL: University of Chicago Press. Hollingshead, Andrea B. (1998), ‘Retrieval processes in transactive memory systems’, Journal of Personality and Social Psychology, 74, 659–71. Ippolito, Richard A. (1977), ‘The division of labor in the firm’, Economic Inquiry, 15 (4), 469–92. Klein, Jeremy (2000), ‘Knowledge management and institutional failure’, paper presented at ‘Knowledge Management – Concepts and Controversies’, Warwick University, 10–11 February. Knight, Frank (1921), Risk, Uncertainty and Profit, Boston, MA: Houghton. Knott, Anne Marie (2003), ‘The organizational routines factor market paradox’, Strategic Management Journal, 24 (10), 929–43. Knott, Anne Marie and McKelvey, Bill (1999), ‘Nirvana efficiency: a comparative test of residual claims and routines’, Journal of Economic Behavior and Organization, 38 (4), 365–83. Knudsen, Thorbjørn and Eriksen, Bo (2002), ‘The architecture of new organizational forms’, mimeo, University of Southern Denmark, Odense, http://openarchive.cbs.dk/ handle/10398/6876. Knudsen, Thorbjørn and Foss, Nicolai J. (1999), ‘Dispersed knowledge and firm organization: fragments of an Austro-evolutionary approach’, mimeo, University of Southern Denmark, Odense. Kogut, Bruce and Zander, Udo (1992), ‘Knowledge of the firm, combinative capabilities, and the replication of technology’, Organization Science, 3 (3), 383–97. Kogut, Bruce and Zander, Udo (1993), ‘Knowledge of the firm and the evolutionary theory of the multinational corporation’, Journal of International Business Studies, 24 (4), 625–45. Krist, H. (1999), ‘Integration of intuitive knowledge in learning at school’, Zeitschrift für pädagogische Psychologie, 13 (4), 191–206.

Distributed knowledge and its coordination 485 Lane, Peter J. and Lubatkin, Michael (1998), ‘Relative absorptive capacity and interorganizational learning’, Strategic Management Journal, 19, 461–77. Langlois, Richard N. (1992), ‘Transaction-cost economics in real time’, Industrial and Corporate Change, 1 (1), 99–127. Langlois, Richard N. (2002), ‘Modularity in technology and organization’, Journal of Economic Behavior and Organization, 49, 19–37. Langlois, Richard N. and Robertson, Paul L. (1995), Firms, Markets and Economic Change, London: Routledge. Lapré, Michael A. and Van Wassenhove, Luk N. (2001), ‘Creating and transferring knowledge for productivity improvement in factories’, Management Science, 47 (10), 1311–25. Larsson, Rikard, Bengtsson, Lars, Henriksson, Kristina and Sparks, Judith (1998), ‘The interorganizational learning dilemma: collective knowledge development in strategic alliances’, Organization Science, 9 (3), 285–305. Lawrence, Paul and Lorsch, Jay William (1967), Organization and Environment: Managing Differentiation and Integration, Harvard University: Division of Research, Graduate School of Business Administration. Lazaric, Nathalie and Marengo, Luigi (2000), ‘Towards a characterisation of assets and knowledge created in technological agreements: some evidence from the automobile– robotics sector’, Industrial and Corporate Change, 9 (1), 53–86. Leenders, Mark A.A.M. and Wierenga, Berend (2002), ‘The effectiveness of different mechanisms for integrating marketing and R&D’, Journal of Product Innovation Management, 19, 305–17. Levitt, Barbara and March, James (1988), ‘Organizational learning’, Annual Review of Sociology, 14, 319–40. Lewis, Kyle (2003), ‘Measuring transactive memory systems in the field: scale development and validation’, Journal of Applied Psychology, 88 (4), 587–604. Lewis, Kyle (2004), ‘Knowledge and performance in knowledge-worker teams: a longitudinal study of transactive memory systems’, Management Science, 50 (11), 1519–33. Liang, D.W., Moreland, Richard L. and Argote, Linda (1995), ‘Group versus individual training and group performance: the mediating role of transactive memory’, Personality and Social Psychology Bulletin, 21, 384–93. Lillrank, Paul (2003), ‘The quality of standard, routine and nonroutine processes’, Organization Studies, 24 (2), 215–33. Linn, Marcia C. (1996), ‘Cognition and distance learning’, Journal of the American Society for Information Science, 47 (11), 826–42. Majchrzak, Ann, Jarvenpaa, Sirkka L. and Hollingshead, Andrea B. (2007), ‘Coordinating expertise among emergent groups responding to disasters’, Organization Science, 18 (1), 147–61. March, James G. (1994), Three Lectures on Efficiency and Adaptiveness in Organizations, Helsinki: Swedish School of Economics and Business Administration. March, James G. and Olsen, Johan P. (1989), Rediscovering Institutions – The Organizational Basis of Politics, New York: The Free Press. March, James and Simon, Herbert (1958/1993), Organizations, Oxford: Blackwell. Martens, Bertin (1999), ‘The evolution of the concept of division of labour in economics’, Discussion Paper 18-99, Max Planck Institut für die Erforschung von Wirtschaftssystemen. Martens, Bertin (2001), ‘The role of conflict and production technologies in the endogenous evolution of institutions’, paper prepared for the 3rd Knexus Symposium, 8–10 August, Stanford, CA. Miner, Anne S. (1994), ‘Seeking adaptive advantage: evolutionary theory and managerial action’, in Joel Baum and Jitendra Singh (eds), Evolutionary Dynamics of Organizations, Oxford: Oxford University Press, pp. 76–89. Minkler, Alanson P. (1993), ‘The problem with dispersed knowledge: firms in theory and practice’, Kyklos, 46 (4), 569–87. Moorman, Christine and Miner, Anne S. (1997), ‘The impact of organizational memory on new product performance and creativity’, Journal of Marketing Research, 34, 91–106.

486

Handbook of knowledge and economics

Munari, Federico, Sobrero, Maurizio and Zamarian, Marco (2003), ‘Analyzing the relationships between product and organizational modularity. Evidence from the Italian Packaging industry’, mimeo, University of Bologna. Nelson, Richard and Winter, Sidney (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Belknap Press of Harvard University Press. Nesta, Lionel and Dibiaggio, Ludovic (2002), ‘Knowledge organization and firms specialisation in biotechnology’, paper presented at the 2002 EAEPE conference. Nooteboom, Bart (1999), Inter-Firm Alliances – Analysis and Design, London: Routledge. Okhuysen, Gerardo A. and Bechley, Beth A. (2009), ‘Coordination in organizations: an integrative perspective’, Academy of Management Annals, 3 (1), 463–502. Ouchi, William G. (1980), ‘Markets, bureaucracies, and clans’, Administrative Science Quarterly, 25 (1), 129–41. Pentland, Brian and Rueter, Henry (1994), ‘Organizational routines as grammars of action’, Administrative Science Quarterly, 39, 484–510. Phillips, Damon J. (2002), ‘A genealogical approach to organizational life chances: the parent–progeny transfer among Silicon Valley law firms, 1946–1996’, Administrative Science Quarterly, 47, 474–506. Radner, Roy (1997), ‘Bounded rationality, indeterminacy and the managerial theory of the firm’, in Zur Shapira (ed.), Organizational Decision Making, Cambridge: Cambridge University Press, pp. 324–53. Reason, James (1990), Human Error, Cambridge: Cambridge University Press. Reber, Arthur S. (1993), Implicit Learning and Tacit Knowledge. An Essay on the Cognitive Unconscious, Oxford: Oxford University Press. Rulke, Diane L. and Galaskiewicz, Joseph (2001), ‘Distribution of knowledge, group network structure, and group performance’, Management Science, 46 (5), 612–25. Sah, R.K. and Stiglitz, J.E. (1986), ‘The architecture of economic systems: hierarchies and polyarchies’, American Economic Review, 76, 716–27. Sah, R.K. and Stiglitz, J.E. (1988), ‘Committees, hierarchies, and polyarchies’, Economic Journal, 98, 451–70. Sanchez, Ron (1997), ‘Modular architectures in the marketing process’, Journal of Marketing, 63 (Special Issue), 92–111. Schilling, Melissa (2000), ‘Towards a general modular systems theory and its application to interfirm product modularity’, Academy of Management Review, 25 (2), 312–34. Schmalhofer, F., Franken, L. and Schwerdtner, J. (1997), ‘A computer tool for constructing and integrating inferences into text representations’, Behavior Research Methods Instruments and Computers, 29 (2), 204–9. Segelod, Esbjörn (1997), ‘The content and role of the investment manual – a research note’, Management Accounting Research, 8, 221–31. Silvestre, Joaquim (1987), ‘Economies and diseconomies of scale’, in John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave: A Dictionary of Economics, London: Macmillan, vol. 2, pp. 80–83. Simon, Herbert A. (1955), ‘A behavioral model of rational choice’, Quarterly Journal of Economics, 69, 99–118. Reprinted in Herbert A. Simon (1982), Models of Bounded Rationality and Other Topics in Economics, Volume 2: Behavioral Economics and Business Organization, Cambridge, MA: MIT Press, pp. 239–58. Simon, Herbert A. (1947/1997), Administrative Behavior, 4th edn, New York: Free Press. Smith, Adam (1776/1926), The Wealth of Nations, London: Macmillan. Spender, J.-C. (1996), ‘Making knowledge the basis of a dynamic theory of the firm’, Strategic Management Journal, 17 (Winter Special Issue), 45–62. Stene, Edwin O. (1940), ‘Public administration – an approach to a science of administration’, American Political Science Review, 34 (6), 1124–37. Stigler, George (1951), ‘The division of labor is limited by the extent of the market’, Journal of Political Economy, 59 (3), 185–93. Takeishi, Akira (2002), ‘Knowledge partitioning in the inter-firm division of labor: the case of automotive product development’, Organization Science, 13 (3), 321–38.

Distributed knowledge and its coordination 487 Taylor, Frederick Winslow (1912/1993), The Principles of Scientific Management, London: Routledge and Thoemmes Press. Teece, David and Pisano, Gary (1994), ‘The dynamic capabilities of firms: an introduction’, Industrial and Corporate Change, 3 (3), 537–56. Teece, David J., Rumelt, Richard, Dosi, Giovanni and Winter, Sidney G. (1994), ‘Understanding corporate coherence: theory and evidence’, Journal of Economic Behavior and Organization, 23, 1–30. Thompson, James D. (1967), Organizations in Action: Social Science Bases of Administrative Theory, New York: McGraw-Hill. Tsoukas, Haridimos (1996), ‘The firm as a distributed knowledge system: a constructionist approach’, Strategic Management Journal, 17 (Winter Special Issue), 11–25. Ulrich, Karl (1995), ‘The role of product architecture in the manufacturing firm’, Research Policy, 24, 419–40. Vassilakis, Spyros (1987), ‘Increasing returns to scale’, in John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave: A Dictionary of Economics, Volume 2, London: Macmillan, pp. 761–5. Wasserman, Stanley and Faust, Katherine (1994), Social Network Analysis: Methods and Applications, Cambridge: Cambridge University Press. Wegner, D.M. (1986), ‘Transactive memory: a contemporary analysis of the group mind’, in B. Mullen and G.R. Goethals (eds), Theories of Group Behavior, New York: Springer– Verlag, pp. 185–205. Williamson, Oliver E. (1975), Markets and Hierarchies: Analysis and Antitrust Implications – A Study in the Economics of Internal Organization, New York: Free Press. Williamson, Oliver E. (1985), The Economic Institutions of Capitalism, New York: Free Press. Winter, Sidney (1987), ‘Knowledge and competence as strategic assets’, in David Teece (ed.), The Competitive Challenge – Strategies for Industrial Innovation and Renewal, Cambridge, MA: Ballinger, pp. 159–84. Winter, Sidney G. (1995), ‘Four Rs of profitability: rents, resources, routines, and replication’, in Cynthia Montgomery (ed.), Resource-based and Evolutionary Theories of the Firm – Towards a Synthesis, Dordrecht: Kluwer, pp. 147–78. Witt, Ulrich (1998), ‘Imagination and leadership – the neglected dimension of an evolutionary theory of the firm’, Journal of Economic Behavior and Organization, 35 (2), 161–77. Zirpoli, Francesco and Becker, Markus C. (2011a), ‘What happens when you outsource too much’, MIT Sloan Management Review, 52 (2), 59–64. Zirpoli, Francesco and Becker, Markus C. (2011b), ‘The limits of design and engineering outsourcing: performance integration and the unfulfilled promises of modularity’, R&D Management, 41 (1), 21–43. Zollo, Maurizio and Winter, Sidney G. (2002), ‘Deliberative learning and the evolution of dynamic capabilities’, Organization Science, 13 (3), 339–51.

21 Evolution of individual and organizational knowledge: exploring some motivational triggers enabling change Nathalie Lazaric

21.1 INTRODUCTION The relation between institution and individual behaviour has been widely debated in the institutionalist theory, according to which collective learning rests on individual habits, routines and other types of more or less formalized practices (Commons, 1934; Veblen, 1914). New interest in the notion of routine has recently arisen, particularly following Nelson and Winter’s work (1982), which highlighted the relative permanency of firms’ behaviours but also their capacity to innovate. Using a Schumpeterian framework, these authors free themselves from the traditional institutionalist framework and consider that processes of routine selection respond essentially to external regularities. Yet a careful re-examination of the concepts of habits and routines shows the similarity of the notions, in terms of their properties and of their ability to integrate changes (Hodgson, 1993; Lazaric, 2000; Lorenz, 2000). Habits and their potential routinization should be understood through the evolution of knowledge. In fact, the procedural and declarative knowledge held by individuals is not inert but memorized in a certain social and political context, which leads individuals to reconfigure it actively in order to either question or modify it in the course of action according to their understanding of a situation. This issue, induced by the idea that free will is essential, remains critical for Bargh (1997; Lazaric, 2011). In order to understand what drives change, the heart of cognitive mechanisms has to be explored for understanding the development and evolution of declarative and procedural knowledge. Some cognitive processes in human beings remain a mystery but cognitive science research is gradually shedding light on some of their aspects. Furthermore, new dimensions must be taken into account to understand the creation of procedural knowledge: individuals’ emotions and free will. Within organizations, the memorization mechanisms are at once similar and diverse. Indeed, organizations use their own filters and mechanisms to generate organizational coordination. This memory feeds on individual knowledge but also has its own dimension as it does not merely consist 488

Evolution of individual and organizational knowledge 489 of the sum of individual knowledge and must be able to survive when individuals leave. Memorization processes are distributed, that is, they are affected by the interactions of individuals with other people. The routines rest on the organizational memory implemented and on the procedural knowledge and representations of this knowledge (individual and collective representations). The organizational context, where memorization processes and the creation of routines emerge, can, in some measure, influence the creation of procedural and declarative form of knowledge. Motivation also plays a critical role in this process by enabling knowledge absorption and by facilitating knowledge regeneration. In this chapter, I shall first enter the cognitive debate in order to describe and define diverse forms of knowledge at the individual level. For this, I shall use the classical dichotomy depicted by Anderson (1976), who introduced the famous notion of declarative and procedural forms of knowledge. I shall emphasize why these two forms of knowledge integrate both stability and change. Second, I shall discuss these notions at the organizational level for showing how organizational is different from individual knowledge. I shall embrace this debate in an evolutionary perspective for discussing change at the organizational and institutional level.

21.2 INDIVIDUAL FORMS OF KNOWLEDGE 21.2.1

Anderson’s Distinction between Two Forms of Knowledge

Many authors regard the distinction between procedural knowledge and declarative knowledge as fundamental to understanding the development of routines. For example M. Cohen (1991), based on Anderson’s works, distinguishes these two forms of knowledge. Taking into account the studies on artificial intelligence (Winograd, 1975) and a criticism of behaviourist analyses – which focused on the external manifestation of behaviour and not its content – Anderson observed the course of mental processes in order to understand their driving forces. Based on Simon’s distinction between short- and long-term memory, he introduced a key notion – that of declarative memory, which plays a key role between ‘working memory’ and ‘production memory’. Anderson developed the idea that one form of memory collected facts and that another stored and recorded them (Anderson, 1976). This notion was refined and clarified in 1983, a time during which he distanced himself from the Simonian approach in order to concentrate on cognitive mechanisms per se. Later on, Anderson’s research used the criticisms of artificial intelligence to go beyond the metaphor of the mind as a simple computer and to explore new

490

Handbook of knowledge and economics

directions. Indeed, starting from Ryle’s distinction between ‘know that’ and ‘know how’, Winograd (1975) underlines why the expert system can encode only two representations of knowledge: the knowledge encoded independently from the programme and the knowledge mobilized when the programme is being used. In this context, declarative knowledge may exist independently of its use, whereas procedural knowledge translates into in a specific behaviour. This makes it possible to operationally separate the knowledge that is easily accessible, communicable and usable because it is independent and varied from the knowledge that is used to solve a problem. Declarative memory concerns more specifically the recollection of facts, events and propositions (Anderson, 1983; Cohen, 1991). It is not linked to a specific use and can be used for several purposes. In particular, it can be ‘re-organized’ in order to find the solution to a problem. It therefore mobilizes facts and technical or scientific principles that are different from know-how. Procedural memory, on the other hand, concerns know-how, how things are done, the knowledge that is put to use. Part of this knowledge can be expressed through routines and rests on ‘patterned sequences of learned behavior involving multiple actors’ (Cohen and Bacdayan, 1994, p. 557). Anderson focuses more particularly on the creation of procedural knowledge. Indeed, he believes that declarative knowledge is converted into procedural knowledge thanks to the processes through which declarative knowledge is interpreted and selected. Interpretation leaves a trace in the working memory and repetition converts this declarative knowledge into procedural knowledge thanks to a compilation mechanism. The creation of some ‘production rules’ and their successful repetition increases their efficiency and, thus, their probability of being selected again. Anderson explains that ‘knowledge automatization’ occurs once knowledge has been used frequently over time for a specific purpose, consequently increasing performance and speed. This observation was introduced for explaining several clinical studies that have shown that the two forms of memory may be dissociated. A patient suffering from amnesia, for example, might no longer memorize declarative knowledge but his memorization of procedural knowledge is not affected. The case of a patient who did not remember the daily visits of his doctor, but could still play chess with great precision (Cohen, 1991), is but one of many examples that have made it possible to distinguish several types of memorization in individuals (Cohen, 1984; Cohen and Eichenbaum, 1993). These two forms of memory do not have implications at the level of individual memorization alone; they also generate specific representations of knowledge, creating different forms of organizational memory.

Evolution of individual and organizational knowledge 491 Furthermore, they can promote the creation of cognitive automatisms that play a role in the individual and collective routinization process. This is why Anderson’s distinction is based on two questions that have been fundamental to the study of routines: 1. 2.

Are there automatic procedures that make it possible to convert declarative knowledge into procedural knowledge? Are there cognitive automatisms that can govern the human spirit?

These questions are examined in the following subsections. 21.2.2

Procedural Form of Knowledge and Potential Cognitive Automatism

The question of cognitive automatisms is not new and has been the object of much research. Bergson proposed, long ago, a philosophical answer to this question by emphasizing that the will was inalienable and that consciousness could not be reduced to mere repetition.1 Recent studies that have also shown that cognitive automatisms evolve in parallel with consciousness and deliberation processes deserve to be examined more closely because they are central to the questions of change in memorization forms. 21.2.2.1 The classical debate on cognitive automatisms The question of cognitive automatisms was first addressed from the perspective of individuals’ attention and their limited capacities. R.M. Shiffrin, with Atkinson, started his work on this in 1968, focusing more particularly on memory control processes (Atkinson and Shiffrin, 1968). A few years later, he and Schneider started their research on cognitive automatisms (Shiffrin and Schneider, 1977), distinguishing two types of information processing. The controlled process is performed more slowly because it is maintained in the working memory, which requires conscious effort and sustained attention (Camus, 1988, p. 64). The automatic process, on the contrary, does not require attention in order to be performed because it is anchored in long-term memory. Furthermore, as Camus underlines, the automatic process can be mobilized consciously (ibid., p. 66). Shiffrin and Schneider’s research has influenced the research protocol in cognitive science by showing how the visual automatism is different from motor-sensory automatism. In the context of motor-skill development, automatism is comparable to a flexible and parametrizable pattern rather than to a rigid process (Schmidt, 1975, p. 83). In keeping with this line of research on visual capacities and their automatic encoding, Kahneman distinguishes several levels of automatism: a highly

492

Handbook of knowledge and economics

automatic type of information processing that does not require any particular attention; a partly automatic process which attention can influence, and finally information processing that is occasionally automatic and requires attention (Kahneman and Charzick, 1983; Kahneman and Treisman, 1984; Lazaric, 2011). This distinction has led cognitive science to put in perspective these forms of automatism: There is widespread consensus around the notion that, notwithstanding exceptions, a behavior understood as an observable response to a given situation, cannot be considered as totally automatic: Only some components of the processing underlying this behavior can be considered totally automatic. (Perruchet, 1998, p. 9)

These studies concur with and complement the work of Anderson by putting in perspective the automatic process implemented by individuals. In the so-called proceduralization phase, knowledge is directly incorporated into procedures for the execution of skills, which makes it possible to mobilize working memory less, but can also lead to errors or misses if the compilation phase is too short. In other words, the transition from declarative knowledge to procedural knowledge remains a delicate operation because the automatic process can lock some know-how into tight procedures. Human judgement is thus necessary to update these procedures. 21.2.2.2 Bargh’s contribution John Bargh (1997) progressively integrates the principles of motivations such as they are described in the ‘self determination theory’.2 Bargh’s theory could be considered in direct continuity to Anderson’s works because he focuses on the procedural dimension of cognitive mechanisms. He also observes to what extent the emotional, cognitive and motivational conditions that characterize an environment can serve as the basis for a preconscious psychological state that can generate an automatic response – automatic in that it escapes the individual’s awareness and direct consciousness. This hypothesis is summarized in Figure 21.1. The underlying idea – which Bargh borrowed from Whitehead (1911) and Shiffrin and Dumais (1981) – is that the routinization of certain procedures helps an individual focus his/her attention on essential, new and creative tasks (Lazaric, 2011). What is new compared to the traditional theory on cognitive automatisms is the manner in which Bargh analyses motivation. Indeed, nothing happens by accident. First of all, before walking may become an automatic process, we have learnt how to walk; and second of all we intend to walk (Bargh, 1997, p. 28). Bargh even talks of an ‘auto-motive model’ to explain to what extent mental representations are essential to the development of cognitive mechanisms.

Evolution of individual and organizational knowledge 493 Evaluative system

Environmental features

Motivational system

Behaviour

Perceptual system

Source: Bargh (1997, p. 26).

Figure 21.1

Parallel forms of preconscious analysis

The motivation/free will hypothesis enables Bargh to free himself from the computer metaphor in which cognition is reduced to serial processing. Mental processes evolve in parallel with various perceptual and motivational mechanisms. As highlighted by the Simonian approach (Simon, 1967), cognition is not the only driving force behind this dynamic, nor is it the only element of the decisional process. The interactions of cognition and motivation are therefore essential and must be taken into account. Consciousness is essential in that it initiates the process of skill acquisition with possible tensions during this learning stage: But even in the case of these automatic motivations, it is possible for a person to become aware of his or her actions and, as in the case of bad habits, attempt to change those behavior patterns. This question of how automatic and conscious motivations interact when in conflict is one of practical as well theoretical importance, and we are now investigating parameters of this interaction. (Bargh, 1997, p. 52)

21.2.2.3 Consciousness and free will in memorization processes The question of consciousness in mental processes has always been a thorny one. Currently research tries to show why the co-evolution of the potential automaticity and of free will is far from being a myth (Bargh and Chartrand, 1999) for understanding the shift from a so-called ‘more automatic’ process to a so-called ‘more deliberate’ and conscious phase.3 Although the debate over the role of consciousness in memorization processes is not new, it was for a long time perceived as a philosophical

494

Handbook of knowledge and economics

issue rather than as a real question about a psychological process that deserved to be studied. Bergson (1991) had already answered this question in a literary manner by highlighting why repetition is not the contrary of free will. Indeed, in order to memorize something, one must make sense of it, otherwise the body does not follow the process. Furthermore, Bergson showed that although our memory is tuned to our body, ‘what is automatic in the evocation of remembrances’ has been greatly exaggerated (ibid., p. 117). And in this automatic memorization tendency there are ‘inner movements of repeating and recognizing that are like a prelude to voluntary attention. They mark the limit between the voluntary and the automatic’ (ibid., p. 145, emphasis added). What Bergson highlighted, at the dawn of the twentieth century, without being able to demonstrate it scientifically (i.e. experimentally with subjects in the process of learning) was that consciousness, freedom and free will were all involved in the same memorization process. Contemporary research has confirmed and refined this hypothesis, but it took decades to do so because ‘Free will has always been something of a problem for behavioral scientists’ (Carver, 1997, p. 98). And because the studies in biology and psychology were all based on the behaviourist paradigm, it took a long time to dismantle it and construct new hypotheses. Recent studies converge on the fact that the consciousness versus automaticity opposition is a dichotomy that is no longer valid because it has now become clear that consciousness accompanies, rather than replaces, the processes of automatization (Baumeister and Sommer, 1997; Tzelgov, 1997; Bargh and Chartrand, 1999; Carver, 1997; Gardner and Cacioppo, 1997). Psychologists agree that both processes evolve together, which leaves unanswered the question of what context and environment cause this process to stop. Furthermore, acknowledging the role of consciousness and free will in memorization processes implies recognizing that chance and the environment have a limited role. In terms of memorization, this boils down to no longer focusing all attention on the mechanisms of procedural knowledge learning, and to acknowledging the fact that declarative knowledge is essential. In other words, the transition from representation to action is a mechanism that needs to be explained if we are to understand how our procedural knowledge evolves and why there is a gap between what an individual thinks he/she does and what he/she actually does (Tzelgov, 1997). In short, declarative and procedural knowledge appear to be indissociable, and understanding the origin of their co-evolution is critical for perceiving the link between individual and organizational routines. However, if ‘human behavior requires a fluid interaction between controlled and automatic processes and between cognitive and affective systems . . . since we see only the top

Evolution of individual and organizational knowledge 495 of the automatic iceberg, we naturally tend to exaggerate the importance of control’ (Camerer, et al., 2005, p. 11). Finally, the question of consciousness and of the modification of our forms of memorization must also be considered in relation to changes. Individuals, as well as organizations, must learn to manage these changes, and to channel the emotions generated by changes in the collective representations. Indeed, ‘the sum total of emotionally intelligent individuals might produce an emotionally handicapped organization’ (Huy, 1999, p. 325). In short, one may note that the debate on routines and automatisms has always had a more or less positive connotation because, even though contemporary research has intentionally distanced itself from the literal meaning of the term ‘routine’, a good number of researchers seeking to observe organizational routines have always associated this research with processes that actually present little interest: ‘It suggests the routine, mindless operation of thousands of people turned into mechanized pieces’ (Cohen, 1997, p. 129). I shall discuss later why this debate on consciousness (mindful versus mindless) has been at the heart of the questions of organizational attention and routines; I shall now discuss why the link between cognition and motivation needs careful examination. 21.2.3

Some Recent Works in Cognitive Science

Squire (2004) offers a new perspective on the cognitive science debate in which he pays homage to the pioneers of this field, on the one hand, and shows the physiological area of emotions in the brain, on the other. Several authors have tried to highlight the importance of representations in cognition and emotions. I shall briefly discuss here Damasio’s (1994) contribution in this regard. 21.2.3.1 Squire’s extension and refinement of Bergson’s theory According to Squire (2004), two pioneers have had great influence in the field of cognitive science – William James (1890) with his work, Principles of Psychology, and Henri Bergson, who was a French philosopher (1859–1941).4 Bergson’s works were pioneering in their conceptualization of memorization processes (Squire, 2004). Indeed, for this French philosopher there are two forms of memorization: memorization in the form of representations; and memorization in action. Bergson distinguishes a memory that stores the facts of our daily life, such as images, from a memory that materializes into motor mechanisms through the recollection of stored facts. Bergson takes the example of the lesson learnt by heart, which, through

496

Handbook of knowledge and economics

repetition, results in a form of automatism. He likens this form of memory to habit-memory, opposing it to image-memory, which does not rely on pure repetition and maintains an important degree of imagination: ‘To call up the past in the form of an image, we must be able to withdraw ourselves from the action of the moment, we must have the power to value the useless, we must have the will to dream’ (Bergson, 1991, p. 94). According to Bergson, there is permanent tension between the memory turned towards the present action and the memory that tends to distance itself from action. Hence the selection between motor habits and their representations: ‘The nascent generality of the idea consists, then, in a certain activity of the mind, in a movement between action and representation’ (ibid., p. 324). Squire (2004) uses and extends Bergson’s distinction to the principle of declarative and non-declarative knowledge. Like James, he underlines that there are many forms of memorization. Some are more able to collect facts, whereas others are directly operational, while others still are related to more emotional forms of memorization located in other parts of the brain, particularly in the ‘amygdala’ area. Each specific form of memorization is associated with a physiological area. This does not constitute a break from Anderson’s work, but a diversification of the non-declarative form of memorization, which helps better understanding of the various learning processes where emotions and reflexes are included as specific types of functioning located in a particular area. According to Squire (2004, pp. 172–3), Declarative memory allows remembered material to be compared and contrasted. It supports the encoding of memories in terms of relationships among multiple items and events. The stored representations are flexible and can guide performance under a wide range of test conditions. Declarative memory is representational. It provides a way to model the external world and as a model of the world it is either true or false. It is dispositional and expressed through performance rather than recollection. Non-declarative forms of memory occur as modifications within specialized performance systems. The memories are revealed through reactivation of the systems within the learning originally occurred.

Squire’s approach might initially be disconcerting because it seems to point to the existence of a multitude of memorization processes. But in fact it is more a decomposition of the procedural memory into different facets and learning zones that it describes. Indeed, in this perspective, one could subdivide memory into a declarative form (where we build representations) and a non-declarative form (zone of reflex and automatisms). Emotions are part of this configuration and correspond to a specific area that stores long-term memories.

Evolution of individual and organizational knowledge 497 Squire’s theory has now been largely accepted within the cognitive science community (Eichenbaum, 1997; Eichenbaum and Cohen, 2001; Milner et al., 1999; Schacter et al., 2000). His argument is partly based on James’s pioneering works, which underlined the flexibility of habits and showed that the various forms of memorization are found in different parts of the brain and are interconnected (James, 1890 in Thompson, 1990). In this vein, Eichenbaum argues that declarative knowledge is essential here in that it makes it possible to avoid oversimplified behaviours characterized by some rigidity in a context of repetition. This knowledge is characterized by ‘representational flexibility’, enabling it to adapt to new situations, contrary to the non-declarative forms of memorization that are thought to be more rigid (Eichenbaum, 1997, p. 554). 21.2.3.2 Damasio’s theory of emotions The question here is therefore to determine, in cognitive and social sciences, the degree of plasticity of both forms of memory. Indeed, although non-declarative memory involves an emotional zone that has an impact on motor-sensory processes, it is highly probable that these forms of memorization are far from inert. This is precisely what Damasio suggests when he argues that emotions play an important role in our cognitive faculties (Damasio, 1994). Our memory does not merely consist of archiving or storing documents or images that enable us to observe the past. On the contrary, our memory rests on representations made of recollections that enable us to implement our potential representations and use them. This notion of images could also be found in the works of Boulding (1956), according to whom our knowledge base is built around our recollection of images that guide our future and present behaviour, a ‘vision of the world’, so to speak. Representations serve to build the future, to create new knowledge and to stabilize the knowledge we use daily. They are made not only of images but also of emotions. When an individual faces a difficult situation and needs to elaborate a strategy, the challenge is to be able to implement a solution that will not be the fruit of past learning only. The act of inventing then rests on analogies and action combined with intuition and reason. Damasio (1994) shows that if a truly new situation emerges, intense emotions play a role in the resolution of this problem. Emotions serve to project oneself into the future and to make decisions; they are not the enemy of the mind but accompany the latter; otherwise, there could be no true invention or creativity. Damasio’s work derives from the somatic-markers hypothesis, which, based on clinical observations, has shed light on how the brain, when faced with a new situation,

498

Handbook of knowledge and economics

activates part of its resources in different ways. Indeed, the attention required to carry out a new task or face a new problem, mobilizes a certain somatic state (ibid., p. 278). The concept of somatic markers illustrates that every time a cognitive effort is made, there is a localized activation of our neurons. This example tries to show why the creation of individual routines may be conceived beyond pure automaticity, and may imply intense attention and decision making. More generally, in the theory of somatic markers, emotions serve, according to Damasio, as a booster for continued working memory and attention. Many researchers have shown that the question of emotions and reason is linked to the question of their intersubjective dimension. Indeed, the social environment can facilitate the regulation of somatic states while correcting the cognitive automatisms that are produced by these states. Although the social dimension has not been explicitly discussed by Damasio (1994), it is currently arousing growing interest because it helps to better understand how the automatic and deliberate processes develop through a form of empathy for others (Rolls, 2000; Camerer et al., 2005). Emotions create ‘motivated cognition’ and have powerful effects on memory as well on the perception of risk (ibid., p. 29). Indeed, learning occurs through imitation and through the concomitant activation of the motor circuits, which makes it possible to learn through a mirror effect and during the acquisition of the procedural memory (Gallese and Goldman, 1998). Many pragmatist authors, such as William James or John Dewey, had already highlighted this reflexive dimension of learning (see in this respect, Thompson 2001 for a comparison of Damasio and Dewey). Dewey’s perspective, in particular, has been re-examined so as to take more thoroughly into account the role of social interactions in cognition (Adler and Obstfeld, 2007; Cohen, 2007). At this level of analysis, a critical issue is how a close examination of the individual dimension can enrich the debate on collective knowledge and its organizational and institutional roots. Indeed, rather than reducing organizational knowledge to the sum of its individual dimensions, one needs to examine the co-evolution of both dimensions. The following section will be devoted more extensively to this latter question.

21.3 COLLECTIVE KNOWLEDGE AND ITS ORGANIZATIONAL AND INSTITUTIONAL ROOTS For some authors, methodological individualism is a prerequisite for understanding the individual component of knowledge and its diverse

Evolution of individual and organizational knowledge 499 forms (Abel et al., 2007); for others this debate makes little sense because it minimizes the importance of the social dimension and denies the heritage of the American institutionalists (Hodgson, 2007). I shall try to clarify my viewpoint by showing why the transition from the individual to the organizational level can generate methodological difficulties, which are not, however, insurmountable. Furthermore, I shall emphasize that looking at the organizational dimension is necessary for understanding efforts made by individuals when faced with change. Nevertheless, focusing on this dimension should not prevent us from also examining the role of institution and its implication at individual and organizational level. I shall introduce this discussion with a discussion around the role and the limit of entrepreneur ability to introduce change in habits and routines and notably for making change in his/her declarative knowledge, which represent his/her traditional ‘business concept’. 21.3.1

Does what we Know about Individuals’ Memorization Processes Apply to Organizations?

When the analysis shifts from the individual to the organizational level, one faces two types of difficulties: the first is that of anthropomorphism, which, according to Walsh and Ungson (1991), is the attribution to organizations of human qualities or characteristics that prevents one from understanding the specificity of organizations. The second difficulty is that of methodological individualism, which reduces the debate to the microeconomic dimension (Hodgson, 2007). Let us examine these two dimensions. 21.3.1.1 The risk of anthropomorphism Researchers often borrow references from cognitive science to explain individual memorization, but this can make it difficult for them to shift their analysis to the organizational level. However, even if individuals do not memorize in the same way that organizations do, it is important to recognize this specific level of analysis in order to understand how it may impact and foster collective memorization. This debate has raised the methodological question of how analyses of individual knowledge could be transposed to the level of organizational knowledge. Several authors consider that organizations do not ‘remember’ in the literal sense of the term. Bartlett (1961), although he shared Bergson’s point of view concerning the re-composition within individual learning processes, had his doubts concerning organizational memory. He favoured the term ‘memory within the group’ over the term ‘memory of the group’ (Paoli and Prencipe, 2003). Thus it appears that the concept of ‘organizational

500

Handbook of knowledge and economics

memory’ should not be understood literally but at a metaphoric level (Divry and Lazaric, 1998). As summed up by Walsh and Ugson (1991), organizations do not have the same attributes as human beings, and taking the analogy between both processes of memorization too far would lead to a real risk of anthropomorphism. However, the daily interactions of individuals promote the development of common beliefs and of intersubjectively shared meanings (ibid., p. 60). The interpretation and selection of collective representations is the essential organizational mechanism that transcends the individual level. It is also the process through which an organization can preserve the knowledge of past events even if some members, who played key roles, left. The social learning processes help to better understand how one can shift from the individual to the organizational level and why there is not only one organizational memory but processes of memorization that are more or less well coordinated. Indeed, organizational memory is not the sum of individual knowledge, but part of the latter, which will be used and activated within organizational routines. Thus organizational memory could be defined as ‘collective beliefs, behavioral routines, or physical artifacts that vary in their content, level, dispersion and accessibility’ (Moorman and Miner, 1997, p. 93), which may have a significant impact on individual memory (see also HowardGrenville, 2005, p. 623 on this point). 21.3.1.2 From the individual to the organizational level I shall now examine the debate on methodological individualism. Although memorization is done by individuals, organizations are characterized by a collective form of memorization. An organization must act as a coherent entity and develop a common organizational culture to which individuals can refer. To avoid reducing the analysis to the micro level, it is interesting to examine the emergence of the mechanism of change brought about by the individuals, and that has an impact on the organization (‘upward causation’), and the changes within organizations that affect individuals (‘reconstitutive downward causation’; Hodgson, 2007, p. 108). Routines clearly lie between these two levels of analysis because they are enacted by individuals in a social context. Coordination serves to balance these tensions by providing the organization with a unity and a memory so that it can look beyond the individual procedural memories. Routines therefore rest on intra-organizational coordination and on the interactions of the various people with their environment (Vromen, 2006, p. 551). This double recursiveness in routines is not new. Its existence was highlighted by Giddens, among others – and more recently by Clark (1999) – but what is of interest here is that it reveals the possible free will of individuals in this

Evolution of individual and organizational knowledge 501 process, which leads us to a definition of routines that does not limit the latter to mere cognitive automatisms. Michael Cohen and Bacdayan (1994) are confronted with the same question in their observation of collective forms of memorization that go beyond individual forms of memorization: ‘the routine of a group can be viewed as the concatenation of such procedurally stored actions, each primed by and priming the actions of others’ (ibid., p. 557). This issue has been examined from a different perspective by Hutchins (1995), who shows how the social and technical environment conditions individual processes of memorization. Indeed, in his study of a marine crew, the social rules and the labor division are far from neutral and partly condition the ‘recurrent interaction patterns’. The technical environment plays a crucial role as it constitutes, for the crew members, a veritable source of external memories that reinforce and complement the individual processes of memorization. A double recursive loop that enables individuals to draw from the group’s memory (which is either in the crew or in the equipment) and enables the group to regenerate itself through the contributions of individual memorization is present. This leads Hutchins (1995) to conclude that there definitely is a type of memory distributed at the group level: Organized groups may have cognitive properties that differ from those of the individuals who constitute the group. These differences arise from both effects of interactions with technology and effects of a social distribution of cognitive labor. The system formed by the navigation team can be thought of as a computational machine in which social organization is computational architecture. The members of the team are able to compensate for local breakdowns by going beyond the normative procedures to make sure that representational states propagate when and where they should. (ibid., p. 228)

Furthermore, in this coordination of tasks, most members of the crew conduct their cognitive activity in an active manner, that is, they go beyond what is formally prescribed so as to be able to anticipate possible breakdowns or future problems (ibid., p. 201). The question here is that of the part played by the individual in the development and control of these cognitive processes. How do variations in procedural and declarative memory occur and under what conditions? 21.3.1.3

Organizational memory and striking the right balance between declarative and procedural memories The literature on organizational memory provides an explanation of why organizations’ memory is not the sum of individual knowledge, but part of the latter, which will be used and activated within organizational routines.

502

Handbook of knowledge and economics

Organizational memory is thus ‘collective beliefs, behavioural routines, or physical artefacts that vary in their content, level, dispersion and accessibility’ (Moorman and Miner, 1997, pp. 93–4). On the basis of this definition, several authors have attempted to determine the impact of the content of organizational memory. Indeed, the centralized or decentralized nature of an organization, as well as the level of memorization (intense or low) has an impact on the firm’s creativity – for example on its economic performance and creativity during the launch of new products. Thus Moorman and Miner (1997) underline the difficulty of determining, in practice, the right dosage of procedural memory and of declarative memory. Too much procedural memory within organizations can hinder creativity because firms find it difficult to absorb new knowledge. This balance is very sensitive to changes in the environment. If organizational memory is dispersed but contains a minimal level of procedural knowledge, firms can then innovate and create new products even in a turbulent environment. However, the dispersion of knowledge can have a negative impact on competitiveness in the case of a turbulent environment, and a positive impact on creativity and performance in the context of a relatively stable environment. The conclusions of the study show that a sufficient degree of procedural knowledge enables firms to absorb new knowledge, but that the nature of this knowledge (centralized or scattered) has a structuring impact in the ability to assimilate (Moorman and Miner, 1997). A complementary study was conducted in some firms of the foodprocessing industry; it shows that a high degree of procedural knowledge has a positive impact on efficiency but a negative impact on creativity, and that it is necessary for both forms of memory to evolve together in order to stabilize the organization and promote its creativity. Indeed, foodprocessing firms tend to extensively codify their practices, which creates important amounts of procedural knowledge but reduces their innovation capacity. Companies have been formalizing next product development activities, following ISO certification programmes, project management models .  .  . promulgated by the industry. In pursuit of this effort they try to codify successful own or others’ practices and experience into a set of rules or recipes . . . to speed up activities and build up skills. One implication of our findings for product development practitioners is that companies need to exercise moderation in these practices. (Kyriapoulos and Ko de Ruter, 2004, p. 1491)

Indeed, the combination of large amounts of procedural knowledge with internal or external information flows does not allow for creativity; this is due to the fact that the high degree of existing procedural knowledge

Evolution of individual and organizational knowledge 503 hinders the absorption and assimilation of new information (Kyriapoulos and Ruter, 2004). This result is in keeping with other studies concerning these standards that underline why the implementation of these standards has a positive impact on attention and detail at the expense of innovation capacities. 21.3.2

Broadening the Debate with Evolutionary Insights: Routines as the Knowledge Activated at the Organizational Level

The evolutionary approach inspired by Veblen (1898, 1914), Nelson and Winter (1982) and Murmann (2003) implies a dynamic of evolution of routines and institutions (Hodgson, 2004; Becker et al., 2005). Furthermore, the analysis of routines cannot focus exclusively on the organizational or macroeconomic level (with a holistic framework of analysis), but must envisage routines as the result of the interaction of the individual and collective dimensions as explained above. An analysis at organizational level therefore helps explain this apparent dichotomy. Thus: Organizational routines are a unit of analysis that allows capturing a level of granularity significant for organizational change. (An analysis that remains too much on a macro-level will be systematically incapable of capturing many interactions and their effects on actors and the environment.) Considering routines enables the researcher to ‘zoom in’ on micro-level dynamics and identify driving forces of change on that level. (Becker et al., 2005, p. 776)

The hypothesis is that change in organizational practices follows the Veblenian logic of ‘cumulative causality’. In this respect, the evolution of practices is cumulative and incremental. This does not exclude the existence of what Rosenberg called threshold effects, which occur after a certain amount of incremental changes have accumulated over time. In other words, ‘cumulative causality’ helps to account for the emergence of new intentional and unintentional behaviours as well as their coevolution with certain aspects of their environment (Hodgson, 2001). Indeed, the question is not how a set of behaviours or actions becomes stable and balanced over time, but how it evolves (Veblen, 1919, p. 8). In this context, individuals have certain habits and behaviours that are conditioned by experience (ibid., p. 79), which is why the cumulative and self-reinforcing process of a set of routines and habits on which the economic order rests has to be depicted. These habits and propensities, embedded in social structures, tend to reproduce themselves; hence their potential inertia: At the same time men’s present habits of thought which tend to persist indefinitely, except as circumstances enforce a change. These institutions which have

504

Handbook of knowledge and economics

so been handed down, these habits of thought, point of view, mental attitudes and aptitudes, or what not, are therefore themselves a conservative factor. This is the factor of social inertia, psychological inertia, conservatism. (Veblen, 1898, pp. 190–91)

Veblen (1914) highlighted the changes in routine behaviours and the profound evolutions in the capitalist system. Although routines and ‘habits’ are disrupted by and depend upon artisanal knowledge, they are not exclusively determined by it (Veblen, 1914, p. 236). This framework, which helps to take into account the embeddeness of individual habits and organizational routines but also the emergence of routinization processes within a population of firms, is in keeping with the recent approach to organizational routines (Lazaric and Denis, 2005; Becker and Lazaric, 2003; Nelson, 1995) and the co-evolution theory. Indeed, several studies have shown the importance of stability and change in organizational routines (Feldman, 2000). Routines are repertoires of knowledge partly activated by the members of an organization (Lazaric, 2000, p. 164); consequently they are generative, dynamic systems, not static objects (Pentland and Feldman, 2005). Finally, the notion of co-evolution captures and sums up this debate because it does not consider the causality of two variables separately, but rather reveals their mutual contribution to the evolution process. Co-evolution .  .  . means not the restricted sense that two things are jointly evolving together but in the broader sense that multiple things are jointly evolving . . . Two evolving populations coevolve if and only if they both have a significant causal impact on each other’s ability to persist. (Murmann, 2003, pp. 21–2)

Co-evolution leads us to broaden the organizational perspective to the institutional level without restricting or subordinating one peculiar dimension to another (Hodgson, 2007). The entrepreneur’s ability to introduce change in his habit and his routines will provide an illustration of potentiality and limits inside the process of knowledge regeneration. 21.3.3

The Institutional Level: Images, Regularities and the Potential Institutionalized Mind

21.3.3.1 Images as regularities and frames for entrepreneurs For Boulding (1956), images play this role of intermediation between the perception of raw data and the internal value system. Each human action is induced by man’s own images, but these images can be altered or revised by the actions themselves.

Evolution of individual and organizational knowledge 505 Every man has a self-image, which includes a picture of his location in space, his acknowledgement of being part of a time flow, the perception of the universe around him as a world of regularities and the sensation of being part of a human relational network. He then has an image to evaluate reality (‘value image’) which intervenes in his relationship with the external environment, embedding information with meanings, an ‘affectional or emotional image’ which provides him with feelings, attitudes and motivations, and a ‘public image’ which helps to compare his personal views with those shared collectively. (Boulding, 1956, p. 14)

In this perspective, images are a way to interpret information and to make sense of the environment. Images create temporarily stable cognitive frameworks bearing individual and collective regularities: ‘In the meanwhile, individual imagery has a relevant social function because it enables collective sharing of values and meanings. From this point of view the image has cohesive power which may exert a strategic function in organisational contexts and in cooperative interaction’ (Patalano, Chapter 6, this volume). But these images are not inert, and they evolve thanks to the experiences that are likely to come into contradiction with the existing images: ‘As each event occurs, however, it alters my knowledge structure or my image . . . Every time a message reaches him, his image is likely to be changed in some degree by it, and as his image is changed, his behaviour pattern will be changed likewise’ (Boulding, 1956, p. 5). This insight highlights that action might also modify cognition either deliberately or simply by accident (Weick, 1990; Greve, 1998). In other words, these constitutive regularities are entangled and the interpretative frameworks not only guide performance but are shaped by this latter, as ‘ostensive routines’ are guided by ‘performative routines’ or might be constrained or triggered by them (Feldman, 2000; Pentland and Feldman, 2005). Innovation may, in this way, be created by an innovator who may modify the current thinking on the economic activity thanks to the emergence of less stereotyped images in some specific context (Patalano, Chapter 6, this volume). The fear of ‘newness’ that may appear, at first sight, as rather counter-intuitive, takes root in the vision of the active entrepreneurs who are surviving in a competitive environment. But those entrepreneurs – as human beings – have their own frameworks, their own declarative knowledge and their intrinsic motivation, leading them either to resist such organizational change or to be at the source of such a change. The cognitive frameworks result both from these internal processes and from the local and cultural environment (Bandura, 1986; Witt, 2000). Cognitive frameworks emerge from the co-constitution of action and perception, as defended by the constructivist approach (see notably Weick, 1990 on this dimension).

506

Handbook of knowledge and economics

Regularities also take their origin in ‘cognitive automatisms’, which are generated by a stabilization inside the ‘procedural knowledge’ that allows memorizing more quickly when circumstances appear to be similar (Bargh, 1997; Cohen and Bacdayan, 1994). These potential automatisms, also taking their origin in the ‘declarative knowledge’ – that is, the representational level – help human beings to find a predictable behaviour in a dynamic environment and to integrate some plasticity in the solving of new problems that the mind has not yet memorized (see Lazaric et al., 2008, for a longer discussion on this point). Images are part of this system as they produce regularities inside the procedural knowledge as well as new insights in the declarative knowledge not always put in practice, that is, transformed by the mind into a purposeful cognitive act. 21.3.3.2 Potential automaticity and the ‘institutionalized mind’ The interplay of the individual and the collective levels of action is far from being neutral for our purpose (Dopfer, 2007). Indeed, entrepreneurs shape their judgement, beliefs and acts not only by themselves but also in interactions with others. And the nature of these micro-interactions can produce ‘recurrent interacting patterns’ that need to be carefully observed (Cohen et al., 1996). In the old evolutionary debate, Commons (1934, 1950) has proposed a taxonomy of interactions linked to the type of knowledge involved (Dutraive, Chapter 5, this volume).‘Routine transactions’ are those related to habitual activities involving stabilized knowledge (embodied in rules) and ‘strategic transactions’ are those related to situations of novelty implying new practices and new opportunities for which there is no stabilized knowledge and rules of thumb. In other words, ‘routine transactions’ are stabilized procedures deeply entrenched in the procedural memory of the entrepreneur, whereas ‘strategic transactions’ concern new ways of doing things not yet classified by the human mind. Indeed, for Commons, deliberation and calculative processes are not always mobilized and can consciously trigger past habits when they are appropriate. However, in some circumstances the mind may reveal ‘a creative agency looking towards the future and manipulating the external world and other people in view of expected consequences’ (Commons, 1934, p. 7; Hodgson, 1988). In short, an institution has to be understood as the working rules of collective action that may restrain the individual deliberation and play a cognitive role by creating ‘institutionalized minds’ and ‘institutionalized personalities’ (Commons, 1934, p. 874). This entanglement between individual and collective actions is very clear: When a new worker goes into a factory or on a farm, or when a beginner starts in a profession or a business, everything may be novel and unexpected because

Evolution of individual and organizational knowledge 507 not previously met in his experience. Gradually he learns the ways of doing things that are expected from him. They become familiar. He forgets that they were novel when he began. He is unable even to explain them to outsiders. They have become a routine, taken for granted. His mind is no longer called upon to think about them. We speak of such minds as institutionalized. But all minds are institutionalized by whatever habitual assumptions they have acquired and take for granted, so that they pay no attention to them except when some limiting factor emerges and goes contrary to what they are habitually expecting. (ibid., pp. 697–8)

Commons – but also Veblen and the main authors belonging to the ‘old’ evolutionary thought – invites us to scrutinize both the mechanisms of change brought about by the individuals and the changes within the organization (Hodgson, 1988). This interplay of the individual and collective dimensions has been very well illustrated in the literature, where entrepreneurs are not always capable of taking the so-called ‘best’ decision because of a vast amount of unreliable information. For this reason, they may have relevant heuristics describing similar contexts, which they use in order to analyse the competitive structure of its environment (Porac and Thomas, 1990). Their individual images are also framed by collective actions of the local environment. Thus entrepreneurs may act in a tied manner, not because they have their own cognitive limits but because the vast quantity of information they might collect is not always suitable for taking the right decision. This may push them to adopt mimetic behaviours dealing with the uncertainty they may have in the creation of their judgment (Greve, 1998). Adopting mimetic behaviour avoids in some circumstances scrutinizing all the possible actions they could trigger (Kahneman, 2003). This myopia is explained by the voluntary ignorance of facts and data but also by the willingness to reduce learning and costs of search of information (Kirzner, 1979). This localized learning – induced by various vicarious learning processes – is not only present at the industrial level but also on the local level (Maskell and Malmberg, 2007). This can produce a deliberate willingness not to absorb new knowledge in order to avoid redefining the deeply entrenched procedural knowledge that matches the current vision. This willingness to stay on ‘routine transactions’ and to keep away from the creation of new ‘strategic transactions’ is illustrated by the famous exploration/exploitation dilemma (Levinthal and March, 1993; Greve, 2007). This compromise shows that exploitation not only increases the probability of performing organizational routines again but simultaneously avoids exploration by reducing the resources available for research. This issue is crucial for entrepreneurs who could act in a local environment with reduced resources, which would prevent them from fully investing in the renewal of practices.

508

Handbook of knowledge and economics

For this reason, the fear of ‘newness’ is not always a pure cognitive process but also a dynamic in which resources are linked to the size of the firm, leading it to overestimate the cost of organizational change induced by innovation. For this purpose, organizational inertia should not be seen as a pure defensive attitude but also as a way to maintain endogenous changes in the micro-interaction helping to contain them within some predictable limits. 21.3.4

Motivation is the Prerequisite for Knowledge Absorption and Knowledge Change

How knowledge is created and diffused at a local level is a crucial issue for academics and practitioners. This puzzle has a long history in the debate about the geography of innovation. Audretsch and Feldman (1996), in a seminal work on the investment of R&D, underline the propensity for industrial activity to cluster spatially in order to benefit from knowledge externalities. However, the real question is not the clustering effect, which is very ancient, but the disparity in localization and the various abilities for (or not) capturing these externalities. These differences could be explained by the ‘knowledge filter’ or the ability to transform opportunities present in the territory in effective innovation and products (Acs et al., 2003). Indeed, knowledge spillovers have to be explained beyond the traditional explanation that knowledge – defined as codified R&D – could be automatically transformed into industrial products. New knowledge could be exploited by agents, but these opportunities, which have to be discovered, are not present spontaneously. On the contrary, before being an opportunity, knowledge has to be identified clearly. An important gap between the stock of knowledge and its relevance exists that could be defined by the ‘knowledge filter’, that is, a certain ‘absorptive capacity’ between firms for knowledge to be transmitted easily (Acs et al., 2003). The basic attribute of knowledge concerns its various externalities and its potential openness. Knowledge is distributed among various decentralized units and needs to be shared and absorbed in a local context. As Schumpeter long ago claimed, knowledge has to be combined to produce innovation. This kind of alchemy is far from automatic because some opportunity must pre-exist in order to enable viable interactions and should be worthwhile for valuable benefits overtaking the uncertainty associated with this process (Nahapiet and Ghoshal, 1998). This highlights the importance of motivation for benefiting from such exchange. Indeed, without engagement of firms and actors the internal stickiness of knowledge will be always present, creating the ‘knowledge filter’ discussed above (Szulanski et al., 2004).

Evolution of individual and organizational knowledge 509 Mindful reflexivity (Langer and Moldoveanu, 2000) and motivation about organizational change are thus necessary to overcome these obstacles, but not always sufficient (Howard-Grenville, 2005). This implies that motivational factors within current practices should be in accordance with the change introduced at a cognitive level. The perception and image of change are crucial here and concern both the declarative and procedural forms of knowledge, that is, the representation of change and its effective implementation. In this perspective, the change in routines should not be seen as a fateful coincidence related to external and disruptive factors, but as a crucial ingredient in the revitalization of individuals and organizations. This leads us to reconsider the very meaning of the term ‘routine’ and to focus on individual and collective memorization processes (Weick and Sutcliffe, 2006, p. 522). In order to create reflexivity about knowledge, mindfulness matters and is materialized by an intention and a capacity to absorb change at both the motivational and cognitive levels (Huet and Lazaric, 2008; Lazaric et al., 2008). A mindful attitude may be defined, for our specific field, as a capacity to go beyond ‘routine transactions’ in order to change the procedural knowledge embedded in entrepreneurs’ minds and entrepreneurs’ ways of doing things. A mindful attitude is an explorative behaviour that has to be adopted to generate ‘strategic transactions’ – that is, transactions that are not always known in advance and that may trigger unpredictable change within organizations. Motivation is a significant trigger for introducing change inside current procedural knowledge and for avoiding the entrenchment of cognitive automatism inside memory. As Locke and Latham (2004) remind us, there has been a long tradition that tries to distinguish knowledge from motivation as two separate streams of research: ‘In most studies of motivation researchers attempt to hold cognition (knowledge) constant as not to confound their separate effects on performance. But in reality, they always go together. Thus we need to learn about how each affects the other (Locke and Latham, 2004, p. 398). Through their concept of ‘volition’ Bono and Locke (2000) try to reconcile these dimensions by showing how free will may be reintroduced without being contradictory to the recommendations of psychological science. Their theory, similar to Bandura’s insights and with the mindful attitude described above, is not in contradiction with old and new evolutionary insights. For example, in this perspective, Witt explains the social diffusion of the ‘business conception’ and shows how cognitive frames originally created by an entrepreneur may be socially shared by employees. The expansion and the adoption of these cognitive frames explicate the difficult growth of the firm. Original business frames may be put into question by employees as soon as the firm grows. The

510

Handbook of knowledge and economics

growth of the firm will create new challenges for diffusion of the cognitive frames and may imply some opportunities for creating new frames inside or outside the firm by diversification or spin-off. The cognitive leadership will be challenged to adapt them to a changing environment, leading to the recreation of new declarative knowledge and a new vision for the ‘business conception’ (Buenstorf and Witt, 2006). Spin-offs are a good example of such entrepreneurship because they represent employees who may not espouse any more current cognitive frames created by the initial entrepreneur and who feel more motivated to create a new firm in order to create their own business conception, their own vision and their own declarative knowledge. These spin-offs are in fact not an exception and show clearly the limits of the cognitive leadership for employees in the adoption of a cognitive frame in which they are the only actor. Motivation may explain a part of change inside declarative and procedural forms of knowledge, but also their co-evolution and their transformation over time. By reintroducing the motivational dimension, we may have a better explanation of the evolution of individual, organizational and institutional knowledge. However, feedbacks between these various levels need to be explained more fully at the empirical level and more attention should be paid to this co-evolution to yield a better knowledge of diverse triggers of change inside individual and organizational knowledge and better ways of observing and eventually measuring them (for a discussion see especially Huet and Lazaric, 2008).

21.4 CONCLUSION In order to understand what drives change in knowledge, I have conducted a close examination of the heart of cognitive mechanisms so as to observe the development and evolution of declarative and procedural knowledge. New dimensions must be taken into account in order to understand the creation of procedural knowledge: individuals’ emotions, free will and motivation. From this point of view, Bargh’s insight, by highlighting the importance of motivation, is symptomatic of the recent reconciliation between the forms of memorization considered ‘automatic’ and the more controlled or deliberate processes. The debate in cognitive science concurs with the debate on organizations. Individual and organizational forms of memorization are distinct but they face the same difficulties: how can representations be made to change, how can a repertoire of knowledge used daily be changed and improved, and how can new knowledge be created? More generally, the cognitive science debate is a detour that has enabled us to get to the heart of processes of individual memorization. These dimensions must be considered when observing individuals’ acquisition

Evolution of individual and organizational knowledge 511 of new procedural knowledge (their resistance or their possible acceptance) and when trying to determine how organizations interfere with this process by ‘smoothing’ this learning process. A careful re-examination of the concepts of habits and routines shows the similarity of the notions, in terms of their properties and of their ability to integrate changes. In this respect, a re-examination of institutions’ role would enable us to better identify and understand the forces behind these changes, which are not exclusively related to cognitive contingencies (Nelson, 1994; Nelson and Sampat, 2001). The persistence of old habits and the difficulty of creating novel routines provide us with the opportunity to examine the organizational and institutional dynamics from an evolutionary perspective. Entrepreneurial frames and exisiting cognitive regularities show why motivation is necessary to renew existing transactions and consequently organizational routines in order to implement a real endogeneous change and to go beyond individual and collective procedural knowledge. Routines may be path dependent if local entrepreneurs are not able to go beyond their prior cognitive regularities in creating new transactions that are not reduced to the product of the past with some ramdom variations but integrate some degree of newness in their mind and in their organization.

NOTES 1. Henri Bergson wrote the first edition of Matter and Memory in 1896 and Creative Evolution in 1907. He was highly influenced by Spencer’s work. Throughout his works, he strongly argued that intuition was deeper than pure reason and was also influenced by Spencer’s evolutionist ideas. However, in Creative Evolution, Bergson argues that creative urge, not the Darwinian concept of natural selection, is at the heart of selection. It is also said that Bergson (who married Marcel Proust’s cousin in 1881) gave Marcel Proust the idea for his great novel Remembrance of Things Past (1913–27). 2. For a discussion on the self-determination theory, see Deci and Ryan (1985). 3. This synthesis is recent; indeed for many years psychology attempted to separate these processes instead of viewing them as working together, and empirical studies focused on one of the phases without managing to coordinate all the interacting cognitive processes. 4. There are many similarities between Bergson’s work and his book Matter and Memory and William James’s approach to habits. Bergson grants great importance to the notion of perception initially developed by William James. Just as James did, he thought that perception and emotion play a key role in our cognitive faculties. Bergson was awarded the Nobel Prize for Literature in 1927.

REFERENCES Abel, P., Felin, T. and Foss, N. (2007), ‘Building micro-foundations for the routine, capabilities, and performance links’, DRUID working paper. Acs, Z., Audretsch, D., Braunerhjelm, P. and Carlsson, B. (2003), ‘The missing link:

512

Handbook of knowledge and economics

the knowledge filter and endogenous growth’, Working Paper presented at DRUID Conference on Creating, Sharing and Transferring Knowledge: The Role of Geography, Institutions and Organization. Adler, P.S. and Obstfeld, D. (2007), ‘The role of affect in creative projects and exploratory search’, Industrial and Corporate Change, 16 (1), 19–50. Anderson, J.R. (1976), Language, Memory and Thought, Hillsdale, NJ: Erlbaum. Anderson, J.R. (1983), The Architecture of Cognition, Cambridge, MA: Harvard University Press. Atkinson, R.C. and Shiffrin, R.M. (1968), ‘Human memory: a proposed system and its control process’, in K.W. Spence and J.T. Spence (eds), The Psychology of Learning and Motivation, Vol. 2, New York: Academic Press, pp. 89–195. Audretsch, D.B. and Feldman, M.P. (1996), ‘R&D spillovers and the geography of innovation and production’, American Economic Review, 86 (3), 630–40. Bandura, A. (1986), Social Foundations of Thought and Action: A Social Cognitive Theory, Englewood Cliffs, NJ: Prentice-Hall. Bargh, J. (1997), ‘The automaticity of everyday life’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 1–61. Bargh, J.A. and Chartrand, T.L. (1999), ‘The unbearable automaticity of being’, American Psychologist, 54, 462–79. Bartlett, J. (1961), Remembering, Cambridge: Cambridge University Press. Baumeister R.E. and Sommer Kristin, L. (1997), ‘Consciousness, free choice and automaticity’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 75–81. Becker, M. and Lazaric, N. (2003), ‘The influence of knowledge in the replication of routines’, Economie Appliquée, 16 (3), 65–95. Becker, M., Lazaric, N., Nelson, R.R. and Winter, S.G. (2005), ‘Toward an operationalization of the routines concept’, Industrial and Corporate Change, 14 (5), 775–91. Bergson, H. (1991) [1896], Matter and Memory, trans. N.M. Paul and W.S. Palmer, New York: Zone Books. Bono, J.E. and Locke, E.A. (2000), ‘Personality and job satisfaction: the mediating role of job characteristics’, Journal of Applied Psychology, 85, 751–65. Boulding, K. (1956), The Image, Ann Arbor, MI: University of Michigan Press. Buenstorf, G. and Witt, U. (2006), ‘How problems of organizational growth in firms affect industry entry and exit’, Revue de l’Observatoire Français des Conjonctures Economiques, June, 47–62. Camerer, C., Loewenstein G. and Prelec, D. (2005), ‘Neuroeconomics: how neuroscience can inform economics’, Journal of Economic Literature, XLIII, 9–64. Camus, J.F. (1988), ‘La distinction entre les processus contrôlés et les processus automatiques chez Schneider et Shiffrin’ (Distinction between controlled and automatic processes in Schneider and Shiffrin), in P. Perruchet (ed.), Les automatismes cognitifs (Cognitive automatisms), Liege and Brussels: Pierre Mardaga, pp. 55–80. Carver, C.S. (1997), ‘Associations to automaticity’, in John A. Bargh and Robert S. Wyer Jr  (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 95–103. Clark, P. (1999), Organisations in Action: Competition between Contexts, London: Routledge. Cohen, M.D. (1984), ‘Preserved learning capacity in amnesia: evidence for multiple memory systems’, in L. Squire and N. Butters (eds), Neuropsychology of Memory, New York: Guilford Press, pp. 88–103. Cohen, D. (1997), ‘Ifs and thens in cultural psychology’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 121–31. Cohen, M.D. (1991), ‘Individual learning and organisational routine: emerging connections’, Organization Science, 2 (1), 135–9. Cohen, M.D. (2007), ‘Reading Dewey: reflections on the study of routine’, Organization Studies, 28 (5), 773–86.

Evolution of individual and organizational knowledge 513 Cohen, M.D. and Bacdayan, P. (1994), ‘Organisational routines are stored as procedural memory: evidence from a laboratory study’, Organization Science, 5 (4), 554–68. Cohen, M.D. and Eichenbaum, H. (1993), Memory, Amnesia and the Hippocampal System, Cambridge, MA: MIT Press. Cohen, M.D., Burkhart, R., Dosi, G., Egidi, M., Marengo, L., Warglien, M. and Winter, S. (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5 (3), 653–97. Commons, J.R. (1934), Institutional Economics: Its Place in the Political Economy, New Brunswick, NJ: Transaction Publishers. Commons, J.R. (1950), The Economics of Collective Action, New York: Macmillan. Damasio, A. (1994), Descartes’ Error: Emotion, Reason and the Human Brain, New York: Putnam’s. Deci, E.L. and Ryan, R.M. (1985), Intrinsic Motivation and Self-Determination in Human Behavior, New York: Plenum. Divry, C. and Lazaric, N. (1998), ‘Organisational memory’ (original French title ‘Mémoire organisationnelle’), Revue Internationale de Systémique, 12 (1), 3–13. Dopfer, K. (2007), ‘The evolutionary signature of Leibenstein’s work. Operant and generic layers of behavioural economics’, in R. Frantz (ed.), Renaissance in Behavioural Economics, London: Routledge, pp. 123–57. Eichenbaum, H. (1997), ‘Declarative memory: insights from cognitive neurobiology’, Annual Review of Psychology, 48, 547–72. Eichenbaum, H. and Cohen, N.J. (2001), From Conditioning to Conscious Recollection: Memory Systems of the Brain, New York: Oxford University Press. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11 (6), 611–29. Gallese, V. and Goldman, A. (1998), ‘Mirror neurons and the simulation theory of mindreading’, Trends in Cognitive Sciences, 2, 493–501. Gardner, W.L. and Cacioppo, J.T. (1997), ‘Automaticity and social behavior: a model, a marriage, and a merger’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 133–41. Greve, H.R. (1998), ‘Managerial cognition and the mimetic adoption of market positions: what you see is what you do’, Strategic Management Journal, 19, 967–88. Greve, H.R. (2007), ‘Exploration and exploitation in product innovation’, Industrial and Corporate Change, 16 (5), 945–75. Hodgson, G. (1988), Economics and Institutions: A Manifesto for a Modern Institutional Economics, Cambridge: Polity Press. Hodgson, G. (1993), Economics and Evolution: Bringing Life Back into Economics, Cambridge: Polity Press. Hodgson, G. (2001), ‘Darwin, Veblen and the problem of causality in economics’, History and Philosophy of the Life Sciences, 23 (3/4), 383–422. Hodgson, G. (2004), The Evolution of Instutitional Economics: Agency, Structure and Darwinism in American Institutionalism, London: Routledge. Hodgson, G.M. (2007), ‘Institutions and individuals: interaction and evolution’, Organization Studies, 28 (1), 95–116. Howard-Grenville, J.A. (2005), ‘The persistence of flexible organizational routines: the role of agency and organizational context’, Organization Science, 6, 618–36. Huet, F. and Lazaric, N. (2008), ‘Capacité d’absorption et d’interaction: une étude de la coopération dans les PME française’, Revue ďéconomie Industrielle, no. 121, premier trimestre, 65–84. Hutchins, E. (1995), Cognition in the Wild, Cambridge, MA: MIT Press. Huy, Q.N. (1999), ‘Emotional capability, emotional intelligence, and radical change’, Academy of Management Review, 24 (2), 325–46. Kahneman, D. (2003), ‘Maps of bounded rationality: a perspective on intuitive judgment and choice’, in T. Frangsmyr (ed.), Les Prix Nobel: The Nobel Prizes 2002, Stockholm, Sweden: Almquist & Wiksell International, pp. 449–89.

514

Handbook of knowledge and economics

Kahneman, D. and Charzyk, D. (1983), ‘Test of the automaticity of reading: dilution of the stroop effects by color irrelevant stimuli’, Journal of Experimental Psychology: Human Perception and Performance, 9, 497–509. Kahneman, D.A. and Treisman, A. (1984), ‘Changing views of attention and automaticity’, in R. Parasuram (ed.), Varieties of Attention, London: Academic Press, pp. 29–61. Kirzner, I.M. (1979), Perception, Opportunity and Profit, Chicago, IL: University of Chicago Press. Kyriapoulos, K. and Ruter, Ko de (2004), ‘Knowledge stocks and information flows in new product development’, Journal of Management Studies, 41 (8), 1469–98. Langer, E.J. and Moldoveanu, M. (2000), ‘The construct of mindfulness’, Journal of Social Issues, 56 (1), 1–9. Lazaric, N. (2000), ‘The role of routines, rules and habits in collective learning: some epistemological and ontological considerations’, European Journal of Economic and Social Systems, 14 (2), 157–72. Lazaric, N. (2011), ‘Routinized learning and cognitive automatisms’, in Norbert Steel (ed.),  Encyclopedia of Sciences of Learning, Springer Sciences Media, in press: DOI 1007/978. Lazaric, N. and Denis, B. (2005), ‘Routinisation and memorisation of tasks inside a workshop: an illustration through a case study’, Industrial and Corporate Change, 14 (5), 873–96. Lazaric, N., Longhi, C. and Thomas, C. (2008), ‘Gatekeepers of knowledge versus platforms of knowledge: from potential to realized absorptive capacity’, Regional Studies, 42 (6), 837–52. Levinthal, D.A. and March, J.G. (1993), ‘The myopia of learning’, Strategic Management Journal, 14, 95–112. Locke, E.A. and Latham, G. (2004), ‘What should we do about motivation theory? Six recommendations for the twenty-first century’, Academy of Management Review, 29, 388–403. Lorenz, E. (2000), ‘Organisational routines in the light of the old evolutionary economics: bringing back into the study of organisational learning’, European Journal of Economic and Social Systems, 14 (2), 191–207. Maskell, P. and Malmberg, A. (2007), ‘Myopia, knowledge development and cluster evolution’, Journal of Economic Geography, 7 (5), 603–18. Milner, B., Squire, L.R. and Kandel, E.R. (1998), ‘Cognitive neuroscience and the study of memory’, Neuron, 20, 445–68. Moorman, C. and Miner, A.S. (1997), ‘The impact of organizational memory on new product performance and creativity’, Journal of Marketing Research, 3 (1), 91–106. Murmann, J.P. (2003), Knowledge and Competitive Advantage: The Coevolution of Firms, Technology, and National Institutions, Cambridge: Cambridge University Press. Nahapiet, J. and Ghoshal, S. (1998), ‘Social capital, intellectual capital, and the organizational advantage’, The Academy of Management Review, 23, 242–66. Nelson, R.R. (1994), ‘Routines’, in G. Hodgson et al. (eds), The Elgar Companion to Institutional and Evolutionary Economics, vol. 2, Aldershot, UK and Brookfield, US: Edward Elgar, pp. 249–53. Nelson, R.R. (1995), ‘Recent evolutionary theorizing about economic change’, Journal of Evolutionary Economics, 33 (1), 48–90. Nelson, R. and Sampat, B.N. (2001), ‘Making sense of institutions as a factor shaping economic performance’, Journal of Economic Behavior & Organization, 44, 31–54. Nelson, R. and Winter, S. (1982), An Evolutionary Theory of Economic Change, Boston, MA: Harvard University Press. Paoli, M. and Prencipe, A. (2003), ‘Memory of the organisation and memories within the organisation’, Journal of Management and Governance, 7, 145–62. Pentland, B.T. and Feldman, M.S. (2005), ‘Organizational routines as a unit of analysis’, Industrial and Corporate Change, 14 (5), 793–815. Perruchet, P. (ed.) (1998), Cognitive Automatisms (original French title Les automatismes cognitifs), Liege–Bruxelles: Pierre Mardaga.

Evolution of individual and organizational knowledge 515 Porac, J. and Thomas, H. (1990), ‘Taxonomic mental models in competitor definition’, Academy of Management Review, 15, 224–40. Rolls, E.T. (2000), ‘Précis of The Brain and Emotion’, Behavioral and Brain Sciences, 23, 177–234. Schacter, D.L., Wagner, A.D. and Buckner, R.L. (2000), ‘Memory systems of 1999’, in E. Tulving and F.I. Fraik (eds), Oxford Handbook of Memory, New York: Oxford University Press, pp. 627–43. Schmidt, R.A. (1975), ‘A schema theory of discrete motor skill learning’, Psychological Review, 82, 225–60. Shiffrin, R.M. and Dumais, S.T. (1981), ‘The development of automatism, in J.R. Anderson (ed.), Cognitive Skills and their Acquisition, Hillsdale, NJ: Lawrence Erlbaum, pp. 111–40. Shiffrin, R.M. and Schneider, W. (1977), ‘Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory’, Psychological Review, 84, 127–90. Simon, H. (1967), ‘Motivational and emotional controls of cognition’, Psychological Review, 74 (1), 29–39. Squire, L.R. (2004), ‘Memory of the brain: a brief history and current perspective’, Neurobiology of Learning and Memory, 82, 171–7. Szulanski, G., Cappetta, R. and Jensen, J.R. (2004), ‘When and how trustworthiness matters: knowledge transfer and the moderating effect of causal ambiguity’, Organization Science, 15, 600–613. Thompson, E. (2001), ‘Empathy and consciousness’, Journal of Consciousness Studies, 8 (5–7), 1–32. Thompson, R.F. (1990), ‘The neurobiology of learning and memory: William James in retrospect’, Psychological Science, 1 (3), 172–3. Tzelgov, J. (1997), ‘Automatic but conscious: that is how we act most of the time’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ: Lawrence Erlbaum, pp. 217–29. Veblen, T. (1898), Why is Economics not an Evolutionary Science?, New York: Viking Press. Veblen, T. (1914), The Instinct of Workmanship and the State of the Industrial Arts, reprint with a new introduction by Murray G. Murphey, New Brunswick, NJ and London: Transaction Books, 1990. Veblen, T. (1919), The Place of Science in Modern Civilization and Other Essays, New York: Huebsch. Vromen, J.J. (2006), ‘Routines, genes and program-based behavior’, Journal of Evolutionary Economics, 16 (5), 543–60. Walsh, J.P. and Ungson, G.R. (1991), ‘Organizational memory’, Academy of Management Review, 16 (1), 57–91. Weick, K. (1990), ‘The vulnerable system: an analysis of the Tenerife air disaster’, Journal of Management, 16 (3), September, 571–93. Weick, K.E. and Sutcliffe, K.M. (2006), ‘Mindfulness and the quality of organizational attention’, Organization Science, 17 (4), 514–24. Whitehead, A.N. (1911), An Introduction to Mathematics, London: Williams and Northgate. Winograd, T. (1975), ‘Frame representation and the declarative–procedural controversy’, in D.G. Bobrow and A.M. Collins (eds), Representations and Understanding, New York: Academic Press, pp. 185–210. Witt, U. (2000), ‘Changing cognitive frames – changing organizational forms: an entrepreneurial theory of organizational development’, Industrial and Corporate Change, 9 (4), 733–55.

Index

aprioristic approach 54, 74, 102, 103, 328–9 abduction 103 absorptive capacity 217–18, 221–2, 349, 378, 502, 508 accessibility 185, 203–4 action behaviourism 131, 343, 347–40 beliefs and 101, 107 Boulding, see Boulding’s mental image cognition as basis and result of, see embodied cognition collective action 114–15, 506–7 logical and non-logical 26–33, 41 Marshallian organization 61–2, 65 Paretian, see Paretian theory situated action 343–4, 350 skilful action 386–7 transaction 102–3, 114–15 adaptation 354–5, 449–53 adaptive learning, see game theory Allais’s paradox 150, 188–90 allocation 274–6, 275–6, 277 altruism 156 artificial intelligence 194, 489–90 aspiration 157, 437–9 association 111 attention 491–2 Austrian tradition 12, 73–4 authority 102, 466–7, 472–3 automatic processes 10, 185, 203–7, 378, 389, 491–2, 493, 494 axiomatic approach 52–3, 68, 187 Allais paradox 150, 188–90 framing effects and 195–7 Bargh’s theory 492–3 Bayesian theory 320–21 behaviourism 131, 343, 347–40 belief 54, 67, 84 common knowledge and 419–20, 422, 424

demonstrable truth 52–3 as habitual disposition to act 101–2, 107 as preconception 107–8 as predictive 247–9, 258 Wieser, see Wieser: the role of beliefs Bergson’s theory 494, 495–7 bias 29–30, 199–202 Boulding’s mental image 121–6 change, adjustment/resistance to 123–5, 135–6, 505 differentiation 127 Hayekian theory and 132–6 image knowledge and perception 122–6, 133 imaginative capacity 126–7 information flow 123, 497, 504–5 overview 121 public image 125 regularities 123 scientific critique of 127–8 subjective nature of knowledge and 132–4 transcription/symbols 125 value image 122 see also framing effects boundaries of the firm 356–8, 428–9, 435 bounded learning 134–6 bounded memory 260–61 bounded rationality 145, 155–8, 161, 246, 438 capabilities 154, 391–2 capital 49, 292–3, 395 Carnegie-Mellon School 151–2 causality 106, 109–10, 113 change in knowledge 488–511 automaticity and the ‘institutionalized mind’ 506–8 automatism 491–2, 493, 494 Bargh’s focus on the procedural 492–3

517

518

Index

Boulding’s theory 123–4, 504–5 co-evolution of declarative and procedural memory 494–5, 504 cognition and motivation in memorization 495–7 cognitive automatisms 491–2 consciousness and freewill in memorization 493–5 consecutive change process 106–7 declarative and procedural memory 10, 205, 489–91, 501–3 emotions 497–8 evolutionary insights 494–5, 503–4 general equilibrium and 50 images as regularities and frames for entrepreneurs 504–6 individual forms of knowledge 489–98 individual to organizational 500–501 institutional level 504–8 mimetic/myopic behaviour 9–10, 507 mindful reflexivity 10–11, 498, 509 motivation 492–3 motivation, knowledge absorption and change 508–10 organizational and declarative and procedural memory 501–3 organizational level 499–504 organizational memory as anthropomorphic 499–500 overview 488–9, 510–11 plasticity of declarative and procedural memory 497–8 routines as doubly recursive 500–501 see also evolutionary perspectives Chicago School 154, 206 co-evolution 173, 412 co-relational structure of knowledge 212–17, 219, 222–3 coalitions 413–14 codification in academia 297–8 as transforming and refining knowledge 303 in business sector 298 DUI mode 306–7 experiential knowledge and 269–72, 278 expert systems 298–300 in relation to innovation 304–6

STI mode 304–6, 307 tacit knowledge and 229–31, 372–3, 377–8, 385 transcription/symbols 125 the work process 300–303 cognitive approach 183–207 ‘as if’ hypothesis 184, 190 axiomatic approach 187, 188–90, 195–7 bias: search processes and computational complexity 199–202 bounded rationality 184–5, 193–5 constructivist view 205–6 evolutionary aspects, introduction of 183–4 evolutionary justification of rationality 190–93 framing effects 196–7 Friedman’s theory 184, 187 Harrod’s evolutionary theory 183–4, 190–91 Hayek, see Hayekian theory industries’ marginalist approach 191 Kahneman and Tverksy’s approach 195–7 Kahneman’s accessibility 185, 203–5 Kauffman’s ‘traits’ (organisms) 191–2 overview 13–14, 151–2 Pareto’s action theory and 33–6 profit maximization debate 183 psychology and economics, separation of 186–7 rational choice and psychology of choice 185–6 sub-optimal solutions and strategies 198–201 tacit versus explicit knowledge 203–6 see also economic theories; memory/ memorization; psychologybased theories cognitive distance 340, 349, 350, 352, 356–7, 361–3 cognitive economics 128, 136, 151, 184–5 cognitive platforms 419–22 cognitive science 340, 491–2, 495–7, 510 coherence 140

Index collective action 114–15, 506–7 collective imagery 125 collective knowledge 498–510 collective opinions 315 common knowledge 419–20, 422, 424, 459 common values/trust 408, 411–13, 416 Common’s theory 110–15 collective action 114–16, 506–7 ‘futurity’ 112, 113–14 going concerns 115 influences on 111–12 interaction/transaction 12, 99, 114–15 rationality: habit and rules 115 typology of ideas 112 communication 125, 229–30, 371 communities in the firm, types 358–60, 414–15 communities of practice 358, 409–10 communities, scientific 327–8 community approach to theory of the firm 403–29 alignment of interests (hierarchical) 418–19 cliques in network theories and 414 coalitions and 413–14 cognitive platforms (common knowledge) 419–22 communities in the firm, types 358–60, 414–15 communities of practice 409–10 coordination advantages 415–17 coordination disadvantages 417–19 epistemic communities 408–9, 410 evolutionary economics approach 404–6, 429 firm as a platform for communities 419–26 firm as processor of knowledge 403 functional coordination groups and 413 hierarchical structures and 422–6 interaction between communities 420–21 knowing communities 407–14 knowing communities, properties of 410–14 knowing communities, types 408 project teams and 413

519

social anthropology of learning 404, 405–6 strategic management approach and 404, 405–6, 427–8 sunk costs 415–16 trust/common values 411–13, 416 see also knowledge generation and utilization; organizations: management of knowledge competition 31–2, 170, 231–2, 233–5 computational complexity 199–202 connectionism (neural nets) 346–8 consciousness 13–14 and mechanism 54–7 as a priori 54 automatism and 10, 185, 203–7, 491–2, 493, 494 free will and 56, 491, 493–5 pragmatist view 104 somatic markers 390–91, 497–8 subsidiary and focal awareness 387, 389 constructivist perspectives 205–6 consumerism, see demand analysis consumption, see demand analysis (consumption) contracts 469 convention, financial 325–9 coordination 74 advantages 415–17 by authority 466–7 balance of knowledge held and 463 costs 464–5 disadvantages 417–19 under a general equilibrium view 176–7 under a knowledge-based view 177 knowledge generation and utilization 226–9 knowledge integration and 470–72 knowledge integration, artefacts 477 knowledge integration, individual capacity 477 knowledge integration, processes 478–9 correlation and co-relation 215 creativity 62, 126–7 cultural determinants 5, 354

520

Index

Darwinian ideas 53–4, 66, 106, 347 decision making biases 199–202 codification in the business sector 298 conjectural division of (sub-problems) 7, 200 expected utility theory, see expected utility theory framing effects (choice problem) 7–8, 9–10 game theory, see game theory normative approach 186–7 organizational forms 441–3 rational choice theory 144–6, 149–50, 154 rules and aspirations, adaptive concept 438–9 Sah and Stiglitz’s theory 442–3 Simonian theory, see Simonian rationality somatic markers 390–91, 497–8 sub-optimal solutions 198–9 uncertainty, see uncertainty see also problem solving demand analysis (consumption) 58–9, 66–7, 76–7, 80–81 democracy 104 determinism 56, 263 deterministic and probabilistic choice 249–50, 251 Dewey’s social theories 102–3, 104, 498 distributed knowledge 458–81 authority (hierarchy) 466–7, 472–3 coordination and balance of knowledge 463 coordination costs 464–5 definitions 4–6, 458–61 diseconomies limiting 463 division of labour and 226–9, 464–5, 467–8, 479 economies of scale and 461 economies of specialization 462–3, 464 economies of speed 463 effects of, on organizations 469–70, 479–80, 502 exploration and exploitation balance 467–8 intermediate situation 461

knowledge integration and 470–72 knowledge integration, artefacts 477 knowledge integration, individual capacity 477 knowledge integration, processes 478–9 knowledge transfer 456–6 modularization 460, 463, 480 structure as coordination mechanism 473–7 study implications 479–81 uncertainty, pervasive 468–9 division of knowledge 79, 305 division of labour 51, 56–7, 224–6, 460 distributed knowledge coordination and 226–9, 464–5, 467–8, 479 doubt, see uncertainty Durkheimian political economics 313–14 e-commerce 326–7 economic theories absolute rationality 145 associationist psychology 108 Austrian tradition 73 economies of scale 461 economies of specialization 462–3, 464 economies of speed 463 efficiency theory 317–22 expected utility theory 149–50 futurity/security of expectations 113–14 general economic equilibrium theory 1–2 and knowledge 167–80 Marshall, see Marshallian theory neoclassical approach 154–5, 159–60, 286–7 public good 288–9 rational choice 287–8, 315 utility 186–7 see also cognitive approach economic theory and knowledge-based economy 167–80 collective dynamics, examples 177–8 decentralization and interdependence 167–8 general equilibrium model and 167–8, 169, 174–5, 176

Index information 168–71 information and knowledge, distinctions 171–2 locality/path dependent models 169–70, 391–2 macroeconomic perspectives 174–8 microeconomic perspectives 173–4 questions raised by 178–80 social distribution of information 169 see also knowledge pools economics Boulding’s ‘resistance to change’ 124 competition 234–5 coordination mechanisms 466–7 increasing return 63 knowledge, its economic characteristics 369–79 market failure 274–7 Marshall’s theories, see Marshallian theory money 58–9, 80–83, 85, 91–2 Pareto and, see Paretian theory and psychology, separation from 58, 186–7 tacit knowledge and 390–92 uncertainty, see uncertainty see also finance, objective value versus convention economics of knowledge 286–96 effectiveness 154–5 efficiency theory 317–22 efficient market 319 egotism 156 electrical circuits analogy 440 elites 36 cliques 414 protectionism 36–43 specialization 78–9, 82, 462–3 Wieser’s analysis of power 87–8 embodied cognition 339–64 boundaries of the firm and 356–8 cognitive distance 340, 349, 350, 352, 356–7, 361–3 cognitive identity 355 communities and firms 358–60 connectionism and neural Darwinism 346–8 embodied cognition perspective 348–51

521

levels and variety (interactionism) 351–3 linguistic semantics 341–2, 350–51 overview 339–40, 363–4 as beyond rational calculation 350 representational–computational view and 340–43 situated action 343–4 theories of organization (firm theory) 353–5 see also knowledge, experiential; knowledge, tacit and personal emotions 33, 34–5, 497–8 empiricism 51–3, 100, 111–12 entrepreneurship 9–10, 504–8 equilibration 65–7 equilibrium 50, 65–7, 68–70, 246 ésprit de géométrie 6–7 evolutionary economics 404–6, 429 evolutionary perspectives ‘bottom-up’ 178 continuum of transactions 104 game theory, see game theory individual and organizational knowledge, see change in knowledge industries’ marginalist principles 191 institutions, see Wieser: the role of beliefs Kauffman’s ‘traits’ (organisms) 191–2 locality/path-dependent models 169 Marshallian 53–4, 59–61, 65–6 network representation of knowledge 218–19 neurology and 347, 388–90 observation and modification 213–14 pragmatist philosophy 102–4, 105–10 progress and equilibrium 65–6 rationality, theoretical justifications of 155, 183–4, 190–91, 190–93 routine and 191–2, 503–4 Smith on (2002 Nobel Lecture) 70–71 see also change in knowledge expected utility theory 149–50, 158, 160, 186–7 Allais’s paradox 150, 188–90

522

Index

experiential knowledge, see knowledge, experiential experimentation 62 exploration/exploitation dilemma 467–8, 507 external environment and mental representation 212–16, 219–20 externalities 274–7, 278–90 facsimile reproduction 272 finalist and mechanist philosophy 109–10 finance, objective value versus convention 313–35 concept of modelling 319–20 consensual opinion 332 Durkheimian political economics 313–14 non-stationary system (conventional predictions) 324–9 objectivity of the future (efficiency theory) 317–22, 329–30 overview 315–17, 333–5 self-referential process 331–2 subjective estimate idea 321–9 uncertainty 320–22 see also economics firm, theory of the, see community approach to theory of the firm firms, see organizations, management of knowledge framing effects 7–8, 196–7 free will and determinism 56, 491, 493–5 Friedman on rational behaviour 153–6 functional groups 413 future/futurity 112, 113–14, 317–22 irreducible subjectivity of 322–9 game theory 159, 246–65 adaptive learning 247–50 backward oriented criteria of performance 249, 250–58 Bayesian learning and 247 beliefs 247–9, 258 bounded rationality and 246 contagion 263–4 deterministic and probabilistic choice 249–50, 251 deterministic process 263

expected payoffs 248–9, 258–64 Fictitious Play processes 259–61 forward-oriented criteria of performance 248–9, 258–64 framework of study 246–7 local interaction 261–4 overview 14 performance evaluation procedures 248–9 probability distributions 248, 260 realized payoffs (reinforcement) 251–3 received payoffs (imitation) 253–8 stochastic process 264 strategic environment 247–8, 250 general equilibrium theory 1–2, 50, 167–8, 169, 174–5, 176–7, 232, 318 generation, see knowledge generation and utilization 211–43 genetics 129, 132–3, 191–2, 226 globalization 303, 394–5 group dynamics communities 413–14, 417 distributed memory 501 knowledge transfer processes 465 Marshallian theory 62–3 Paretian action theory 36–43 Simmel’s theory 351–2 growth 60, 62, 173, 286 habit 101–2, 107, 111, 115 Harrod’s evolutionary justification 183, 190–91 Hayekian theory 1 authority/limited span of control 466–7 information and knowledge differentials 130, 134, 147–8 knowledge and uncertainty 152–3 knowledge, nature of 5–6, 56, 74, 132–4 learning, role of 135 ‘map’ and ‘model’ 129 neuropsychology of perception (genetics) 129–30, 132–3 perception 121, 129–30 heuristics 157 hierarchical structures 422–6, 442–4, 466–7, 472–3 hybrid forms 444–5, 447–9

Index management 404, 424–5 power/authority 36, 87–93, 102 Hume’s radical empiricism 111 idealism 100 images, see Boulding’s mental image imagination 121, 126–7, 137 imitation 82, 88, 89, 253–8 increasing return 63 individual and organizational knowledge, see change in knowledge individual forms of knowledge 489–98 individualism 75, 287 inductive and deductive cognizance 24–6, 103 inflation 59 information and retrievable/ interpretive knowledge structure 217–18 information as objective/factual 130, 217 information processing 6–7, 464, 491 information technology, see codification information, knowledge and economic theory collective dynamics, examples 177–8 corporate knowledge and 172–4 distinctions between information and knowledge 171–2, 217 general equilibrium (GE) and 1–2, 168–71 knowledge of mind, body and soul 291 macro theory and 174–7 innovation 171–2, 212 absorptive capacity and 217–18, 221–2, 502, 508 codification 304–6 image and entrepreneurship 126, 505 by interaction 360–63 procedural knowledge and 502 as revitalization of experiential knowledge 280 transfer and localization 392–5 inquiry 103, 106 institutional theory 73–4, 99, 107–8, 112–17, 488, 504–8

523

Commons’s theory, see Common’s theory Menger’s theory, see Mengerian theory Wieser’s theory, see Wieser: the role of beliefs see also organizational theory intellectual property rights, see property rights intelligence generation 464 interaction 9 communication 125, 229–30, 371 embodied cognition 343–4, 351–3 individual decision making and 159, 167 innovation and 360–63, 506–8 local 261–4 repetitive and qualitative 420–21, 425–6 social character of 292–3 internalized action 344–6 internet search engines 294 Kahneman’s theory 159, 160–61, 195–7, 203–4 Keynesian theory 15, 173, 331–2 Knight: uncertainty and risk 147, 148 knowing communities 407–14 knowledge, collective 498–510 knowledge, distributed, see distributed knowledge and its coordination knowledge, economic characteristics 369–79 codification and 372–4, 377–8 public- and private-good 374–9 see also knowledge generation and utilization knowledge, experiential 267–83, 294 allocation 274–6, 277 analytical framework 274–7 codification as depersonalizing 271–2, 278 codified script (reproduction) 269–71 as disturbing and disruptive 273 disinvention and deactivation of 272–4 economic aspects 272 externalities 274–7, 278–90 industrial organization (maintenance) 279–80

524

Index

innovation (maintenance) 280 integration (maintenance) 281–2 knowledge reproduction and transmission 268–72 market failures 274–7 museum solution to maintenance 277–8 partial reproduction 271 private and collective 268 status and maintenance of 278–82 see also embodied cognition; knowledge pools knowledge, growth of 50, 57–61, 63, 76 knowledge, individual and collective 75, 83, 85–6, 89–90, 136 knowledge, subjective nature of 5–6, 25, 26–7, 130, 132–4 knowledge, tacit and explicit 203–6, 219–31, 291–3, 296–306, 369, 403–4 knowledge, tacit and personal 383–98 biology of 388–90 decision making 390–91 implications for economics 390–92 information and knowledge distinctions 171–2, 391 knowledge management and 384–5 memory/rules 389–90 overview 396–8 Polanyian overview 3–4, 372, 383–4 skilful action 386–7 subsidiary and focal awareness 387 technological learning and capability 391–2 transfer and localization 392–5, 466 transfer and reproduction 268–9 see also embodied cognition knowledge and information 1–2, 130, 134, 147–8, 171–2, 391 knowledge and rationality, see Simonian knowledge and rationality knowledge as intersubjective 104 knowledge-based society 211 knowledge-based theory of the firm, see community approach to theory of the firm knowledge-based economy, see economic theory and knowledgebased economy

knowledge bases of firms 235–41 knowledge generation and utilization 211–43 adaptive learning 247–50 applicability and focus of study 211–12 co-relational examples 215–16, 219 co-relational structure of knowledge 212–17 ‘cognitive’ and ‘technological’ processes 224–5 connectivity 228 coordination 226–9 disciplines, observable spaces 219–20, 225–6, 227–8, 233 division of labour 224–6 econometric equations 216 external environment 212–16, 219–20, 227 inseparability of processes 224–5 knowledge bases concept and 235–41 lexicographic analysis 238–40 local character of knowledge 219–22 measurement of properties 240 measurement: relationship of KB properties and performance 241 mental models 231–2 network (static and dynamic) 218–19, 221 objective knowledge 223 observables and variables 213–15, 222–3 overview 241–3 pattern recognition 232 production of knowledge (generalities) 223–35 retrieval and interpretative structure 217–18, 219, 222 specialization/disciplines 224–6 storage, expression and transfer 370–72 tacit and codified knowledge 219–31 theories of knowledge and 222–3 see also community approach to theory of the firm; knowledge, economic characteristics knowledge integration 470–72, 477–9 knowledge management 422–6, 435–54 knowledge organization, see organizational design

Index knowledge pools 285–309 codification, innovation and 304–7 codification in academia 297–8 codification in the business sector 298–300 codification of the work process 300–303 economics of knowledge 286–96 know-what/why/how/who 293–5 knowledge of mind, body and soul 291–2 knowledge transfer 289–90, 307 neoclassical economics and 286–7 overview 308–9 public good (property rights) 288–90, 292–3, 294–6 rational choice 287–8 social interaction 292–3 tacit/codified knowledge and 296–306 see also economic theory and knowledge-based economy; knowledge, experiential knowledge reproduction and transmission 268–72 knowledge transfer 289–90, 307, 371–2, 392–5, 456 learning adaptive, see game theory biology of tacit knowledge 388–90 bounded learning 134–6 discretionary 301–2 DUI mode 306–7 image and 123 imitation 82 implicit 466 knowledge of mind, body and soul 291–2 knowledge transmission 269 lean production learning 303 locality/path dependent models 169–70 Menger’s emergence of money and 80–83 neurology of 129 organizational learning 198–9 productivity gains through specialization 462–3 reflexive 498

525

technological and sectoral 172–3, 391–2 levels of needs 351–2 lexicographic analysis 238–40 local character of knowledge 63–4, 219–22, 272 locality/path dependent models 169–70 localization 395–6, 508 logical and non-logical action 26–33, 41 Lucasian theory 320, 321 macro/microeconomics 174–8 management 384–5, 404, 424–5 management information systems 298–9 management literature on knowledge distribution 471–2, 480 marginalism 183, 190–93 market failure 274–7 marketability 80–81 Markov chain 260–61 Marshallian scholarship 51–3 Marshallian theory 49–71 activities and wants (demand) 58–9 character 57–8 consciousness and mechanism 54–7 dialectical processes in 60–61 evolutionary approach 53–5, 59–61 growth and equilibrium 50, 65–7 growth of knowledge 50, 57–61, 63 imagination 126 the industrial district 63–4 model of the mind 59–60 organization 61–5 overview 11–12, 49–50, 67–70 principle of substitution 60 productive activity 60–61 psychological elements 52, 53–4, 57–8 true knowledge/empiricism 51–3 mechanism 54–7, 202 memory/memorization 279, 389 bounded memory 260–62 cognition and motivation in 495–7 consciousness and freewill in 493–5 declarative and procedural memory 205, 489–91, 501–3 declarative and procedural memory, co-evolution of 494–5, 504

526

Index

declarative and procedural memory, plasticity of 497–8 emotions 497–8 individual to organizational, risks 499–500 local character of knowledge 220–21 see also cognitive approach Mengerian theory 75–83 complementarity and substitutability 79–80 consumption and knowledge (means and ends) 76–8 division of knowledge and 79 knowledge, growth of 76 knowledge, individual and collective 75, 83 learning (emergence of money) 80–83, 85, 91–2 organic and pragmatic institutions 74, 86 overview 94 selection process and prior knowledge 82 specialization (fragmented population) 79 subjectivist concept of value 75–6 time constraints 78–9 value 75–6 mental models 231–3, 343–4 metaphor 341–2, 350–51 microeconomics 2, 173–4 middlemen 79 mimetic behaviour 9–10, 507 mindfulness 10–11, 509 Misesian theory 74 modular systems 460 money 58–9, 80–83, 85, 91–2 motivation 10, 351–2, 492–3, 508–10 neoclassical approach 154–5, 159–60, 183, 286–7 network externalities/effects 82 heterogeneity 361 knowing communities and cliques 414 network economy 339 representation of knowledge 218–19, 221, 232, 239, 471

static and dynamic 218 neural Darwinism 347 neurophysiology 133, 134, 383, 388–90 New Economy convention 325–9 non-logical and logical action 26–33, 41 normative approach 186 objective end and subjective purpose 27–9 objective knowledge 223 ‘objective spirit’ 89–90 objective values, see finance, objective value versus convention observation and modification 213–14 opinion 314–15, 316, 331–3 organic and pragmatic institutions 74, 86 organization design (knowledge organization) 435–54 behavioral theory and 437–40 cognitive ability and screening function 440–41, 445–7, 450 cost 447 decomposability/adaptability 449–53 goals 437–8 hierarchy and polyarchy structures 442–5, 447–9 hybrid forms 444–5, 447–9 market structure and 447, 448 modeling 445–9 organization of knowledge as structure 441–5 organizational forms, typology 445, 457 organizations as architectures 440–49 overview 435–6, 439 problemistic search 439 organizational learning 198–9, 300–301 organizational theory boundaries 356–8 collective knowledge 498–510 communities 358–60 competence 353 coordination 464 corporate knowledge 172–4 firm size 356, 359–60

Index inclusiveness and tightness (myopia risk) 357–8 interaction, distance and innovation 360–63 legal identity 355 Marshallian theory 61–5 organizational focus 354–6, 357–8 organizational learning 198–9 selection and adaptation 354–5 Smith’s theory 70 theory of the firm (embodied cognition) 340, 353–6 work process 300–303 see also institutional theory organizations, management of knowledge 435–54 codification of ‘know-why’ 298 design of, see organization design distributed knowledge, see distributed knowledge expert systems 298–300 industrial districts and innovation 64–5 knowledge bases of firms 235–41 localization 395–6 maintenance of experiential knowledge 279–80 mindfulness and 10–11, 509 overview 435–6 performance, see knowledge generation and utilization scientific communities 327–8 Simonian theory, see Simonian rationality tacit knowledge and technology transfer 392–5 tacit routine 393–4 work process 300–303 see also community approach to theory of the firm Oxford Research Group 183 parallel distributed processing 347 Paretian theory 23–44 action theory 24–6 chart of free trade/protectionism choice 39–40 ‘derivations’ 34–5, 41 ‘elites’ 36

527

group dynamics (free trade/ protectionism) 36–43 inductive and deductive cognizance 24–6 intentional nature 27 ‘interests’ 35–6 knowledge and logical and non-logical actions 26–33 knowledge building, categories 33–6 logico-experimental method (economics and sociology) 24–6 method of successive approximations 25 overview 11, 23 pursuit curve 30–32 ‘residues’ 33–4 typology of actions 27–30 Peircian theory 101–2, 103, 104 Penrosian theory 60, 62–3, 127, 353, 394 perception 121, 129–30 performance, see knowledge generation and utilization Piaget and Vygotsky 344–5 pleasure 55–6 Polanyian theory 3–4, 372, 383–4, 386–7, 390 political economics 313–14 power/authority 36, 87–93, 102, 466–7, 472–3 pragmatist theory of knowledge 100–105 abduction 103 beliefs 101–2 collective dimension 104–5 definition and variants 100 evolutionary, see Veblenian theory institutionalist conception, see Common’s theory knowledge and beliefs 100–105 main propositions 105 mental processes and beliefs 100–102 naturalization of mind/function of thought 101 reflexive learning 498 relation and inquiry (ontology and method) 102–3 scientific methods 102, 103 private property, see property rights

528

Index

probability 248, 249–50, 251, 260, 318, 320–21, 468–9 problem solving 6–7 bias and non-logical actions 29–30 Rubik’s Cube 199–200 ‘satisficing approach’ 157, 195, 197 sub-optimal solutions 198–9 see also decision making procedural rationality 7, 66, 145, 157–62, 205 product development 471, 502 productivity gains 462–3 profit-maximization 183 property rights 36–43, 288–9, 292–3, 308, 375–9 prospect theory (Kahneman and Tversky) 159, 160–61, 195–7 protectionism 36–43 psychology-based theories ésprit de géométrie 6–7 historical development of 130–31, 147 implicit knowledge 387 internalized action 344–6 knowledge and uncertainty 147–9, 152–3 Marshallian theory and 52, 53–4, 56–8 memory processes 494 perception 129–30 see also cognitive approach public good 275, 276–7, 280, 369, 374–9 property rights and 288–90, 292–3, 294–6 R&D, see innovation rational choice theory 144–6, 149–50, 154, 184 rationalism 100 rationality 77, 286 evolutionary justification of 190–93 logical and non-logical action 26–33 rational choice 287–8, 315 Simonian, see Simonian knowledge and rationality reality/truth 5–13, 102, 212–13 reflexivity 498, 509 regularities 123, 504–6 regulation approach 178

reinforcement learning 251–3 relation, pragmatist ontology 102–3 religous belief 53 repetition 494 representational–computational view 340–41, 342 retrieval and interpretative structure of knowledge 217–18, 222 Rhône-Poulenc 238–40 risk 145, 147, 320, 468–9 routine double recursiveness 500–501 evolution of knowledge and 488, 495 evolutionary economics 404–5 habit 101–2, 107, 111, 115 knowledge, learning and 8–11 mechanism and 55–6, 60–61 organizational change 62–3, 503–4 organizational knowledge 478 sub-optimal solutions and strategies 198–202 tacit versus implicit knowledge 203–6 within firms 393 rules 6–7, 115, 389–90, 438–9 Sah and Stiglitz’s organization 441–4 Schumpeterian theory 68, 328 science–technology–innovation sequence 304–6 scientific community organization 327–8 scientific knowledge 273–4, 276–7, 281, 384 scientific method 102, 103, 222–3, 226 securities market 317, 323–9 selection process, Menger’s 82 semantics, linguistic 341–2, 350–51 shared knowledge 458–9, 465 Simonian knowledge and rationality 121, 144–62 absolute rationality 145 analysis comparative to Friedman 154–5 auxiliary assumptions 155–6 bounded and procedural rationality relationship 145, 157–8, 161 bounded rationality 155–6, 158, 193–5, 438

Index divergence (economic and human) 145 empirical foundations and social interactions 159 limited cognitive resources 467 mathematical models 439 nature of human rationality 146 organizaiton knowledge and design 436, 439 organizational learning 198–9 procedural rationality 7, 66, 145, 157–62, 205 procedurality, comparative analysis 158–61 rational choice and psychologybased theories 147 ‘satisficing approach’ 157, 195 substantive rationality 26–7, 32 uncertainty and knowledge 152–3, 157–8 uncertainty and rational choice 149–52 utility function 155–6 situated action 343–4, 350 situated knowledge 4–5 small decisions 275–6 Smith, Vernon 70, 205–6 Smithian tradition 51–2, 167, 224, 305, 308, 462 division of labour, see division of labour social dimensions democracy 104 economics, uncertainty and specialization 78–9 image and 125, 127 interaction, see interaction Pareto’s free trade/protectionism choice 36–43 social character of knowledge 292–3 socially distributed knowledge 6, 176 somatic states 498 transactions, going concerns and institutions 104, 114–16 social value 275 social welfare 37 somatic markers 390, 497–8 specialization/disciplines 51, 56–7, 78–9, 82, 224–5 Spencerian theory 54, 57

529

Squire’s cognitive science 495–7 stochastic process 264 strategic management approach 404, 405–6, 427–8 alignment of interests 418–19 knowledge management 422–6 structuralist theories of science 222 sub-optimal solutions and strategies 198–202 subjective expected utility 158 subjectivism 5–6, 25, 26–7, 130, 132–4 subjectivity of fundamentalist estimates 322–9 substantive rationality 26–7, 32, 116 success 90–91 Suchman theory 4 tacit and explicit knowledge 203–6, 219–31, 291–3, 296–306, 369, 403–4 tacit and personal knowledge, see knowledge, tacit and personal Taylorist organization 300–302 teams 413, 477 technological knowledge 172, 391–2, 501 teleological behaviours and cumulative causation 108–10 tenacity 101–2 terminology of knowledge 290–91 theology 52–3 time constraints 78–9, 136 transaction 102–3, 114–15, 320 trust 411–13, 416 truth/reality 51–3, 102, 212–13 Tversky, decision theory 203 uncertainty 101, 103, 124, 146, 152–3, 320–22, 328 and rational choice (expected utility theory) 149–51 distributed knowledge and 468–70 knowledge and 147–9 security of expectation 113–14, 144 unemployment 67 unified general theory 229 utility 155–6, 186–7 value 75–6 Commons’ futurity 113–14

530

Index

dualist explanations (subjective/ objective) 85–6, 89–90 objective, see finance, objective value versus convention subjective 75–6 Veblenian theory evolutionary economic theory 108 human behaviour (cumulative causation) 108–10, 503–4 perceptions (changing reality) 105–7 preconceptions, emergence and evolution 107–8 Walrasian theory 2, 69 Weberian theory 112, 216

Wieser: the role of beliefs 83–94 law of upward mobility of classes 91 Menger’s emergence of money and 85, 91–2 methodological individualism and holism 85–6, 89–90 notion of institutions 86–7, 93, 94 ‘objective spirit’ 89–90 overview of theory 83–4, 94 power (internal and external) 88–9 power (masses and leaders) 87–8, 92 power (supra-social nature of) 92–3 social egotism 90–91 work process 300–303 zone of proximal development 345