Processes of Emergence of Systems and Systemic Properties: Towards a General Theory of Emergence, Proceedings of the International Conference, Castel Ivano,Italy, 18-20 Ocotber 2007 9812793461

Citation preview

Processes of Emergence of Systems and Systemic Properties Towards a General Theory of Emergence

This page intentionally left blank

Processes of Emergence of Systems and Systemic Properties Towards a General Theory of Emergence Proceedings of the International Conference Castel Ivano, Italy

18 – 20 October 2007

editors

Gianfranco Minati Italian Systems Society, Italy

Mario Abram Italian Systems Society, Italy

Eliano Pessa University of Pavia, Italy

World Scientific NEW JERSEY

•

LONDON

•

SINGAPORE

•

BEIJING

•

SHANGHAI

•

HONG KONG

•

TA I P E I

•

CHENNAI

Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

PROCESSES OF EMERGENCE OF SYSTEMS AND SYSTEMIC PROPERTIES Towards a General Theory of Emergence Copyright © 2009 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN-13 978-981-279-346-1 ISBN-10 981-279-346-1

Printed in Singapore.

The proceedings of the fourth national conference of the Italian Systems Society (AIRS) are dedicated to the memory of Evelyne Andreewsky, passed away on December 2007. Several members of AIRS had the honour to be her colleague and friend.

Evelyne Andreewsky was born in Paris. She earned an engineering degree in Electronics from E.S.E., Paris, and a "Docteur ès Sciences" degree (PhD) in Computer Science (Neurolinguistic Modelling), from Pierre & Marie Curie University, Paris VI. She was Senior Researcher at the French National Research Institute I.N.S.E.R.M. She has switched from (a straight) Computer Scientist career (as research engineer, chief of information processing public labs, consultant for government's policies, UNESCO expert...) to (pure) Research, trying to develop new multidisciplinary systemic approaches to Cognition and Language (over 150 papers in scientific international journals, books, chapters of books + books editor, guest editor of journals).

v

vi

Dedication

She was founder and honorary president of the Systems Science European Union (UES). She was actively involved in the board of scientific societies, namely AFSCET (French Systems Science Society) and MCX (European Program for Modelling Complexity). She belonged to the editorial board of scientific journals, namely "Cybernetics and Human Knowing" and Res-Systemica. She has organized or co-organized a number of national and international congresses, symposia and summer schools. She has been elected (1999) to Honorary Fellowship of the World Organisation of General Systems and Cybernetics (WOSC), founded by Professor John Rose, and has been invited to give courses or lectures in various countries. We will never forget her and her dedication to systems science. Thank you Evelyne.

PREFACE

The title of this fourth national conference of the Italian Systems Society (AIRS), Processes of emergence of systems and systemic properties − Towards a general theory of emergence, has been proposed to emphasize the importance of processes of emergence within Systemics. The study of this topic has a longstanding tradition within AIRS. Namely this conference can be considered as a continuation of the previous 2002 conference, Emergence in Complex Cognitive, Social and Biological Systems, and 2004 conference, Systemics of Emergence: Research and Applications. In the preface of the 2004 conference the editors wrote: “Emergence is not intended as a process taking place in the domain of any discipline, but as ‘trans-disciplinary modeling’ meaningful for any discipline. We are now facing the process by which General System Theory is more and more becoming a Theory of Emergence, seeking suitable models and formalizations of its fundamental bases. Correspondingly, we need to envisage and prepare for the establishment of a Second Systemics − a Systemics of Emergence…”. We had intense discussions in the periodic meetings of AIRS, focused on the large, increasing amount of contributions available in the scientific literature about emergence. In this regard we remark that AIRS members were and actually are involved in research projects in several disciplinary fields, having the experience of applying the view of emergence outlined above, for instance, in Architecture, Artificial Intelligence, Biology, Cognitive Science, Computer Science, Economics, Education, Engineering, Medicine, Physics, Psychology, and Social Sciences. As a consequence of this intense activity we felt an increasing need to better specify the principles to be adopted when dealing with this evolving, interdisciplinary study of emergence. With this point of view in mind, which could be viewed as a generalization of other instances, historically at the basis of birth of different systems societies in the world (e.g., Cybernetics, General System Theory, Living Systems Theory, Systems Dynamics, Systems Engineering, Systems Theory, etc.), in October 2006 the Italian Systems Society approved a Manifesto, available at our web site www.AIRS.it . It relates to our vision of the current situation of the role of world-wide Systems Societies, as well as of problems and perspectives of Systemics. In the Manifesto we outlined some fundamental aspects of our identity, such as the necessary role of disciplinary knowledge for Systemics, as

vii

viii

Preface

well as of inter- and trans-disciplinary knowledge, the meaning of generalization, the need for rigorousness and the non-ideological valence of reductionism. We quote the concluding statements of the Manifesto: “The purpose of systems societies should be to identify and, where possible, produce contributions to Systemics taking place in disciplinary and multidisciplinary research, making them general and producing proposals for structuring and generalizing disciplinary results. Examples of theoretical aspects of such an effort are those related the establishment of a General Theory of Emergence, a Theory of Generalization, Logical Philosophical models related to Systemics and the issue of Variety in different disciplinary contexts.” The general theory of emergence we figure out is not a unique, caseindependent and scale-independent approach, having general disciplinary validity. Instead we have in mind different, dynamical and interacting levels of description within a constructivist view able to model processes of emergence, in order not to reduce all of them to a single description, but to introduce multimodeling and modeling hierarchy as a general approach to be used in principle. A related approach has been introduced in literature with the DYnamic uSAge of Models (DYSAM) and logical openness, i.e. meta-level modelling (models of models). We make reference to a constructivist science, as dealing with the constructive role of the observer in processes of emergence. The latter is related to his/her cognitive model allowing the recognition of acquired systemic properties, which occurs when the hierarchical processes generating these properties cannot be modeled by using traditional causal approaches. In other words, according to a constructivist view on one side the observer looks for what is conceivable by using the assumed cognitive model, and, on the other side, he/she can introduce methodologies allowing the possibility of producing incongruence, unexpected results and inconsistence. The latter process asks for a new cognitive model generating paradigm shifts and new theoretical approaches, such as in the case of abduction, as introduced by Peirce. All this is endowed with a deep, general, cultural meaning when the focus is on scientific aspects where it is possible to test, compare, validate and formulate new explicative theories. Moreover, we believe that the subject of emergence is a sort of accumulation point of increasing, mutually related conceptual links to disciplinary open questions, such as the ones mentioned in the topics of the conference.

Preface

ix

The study of processes of emergence implies the need to model and distinguish, in different disciplinary contexts, the establishment of structures, systems and systemic properties. Examples of processes of emergence of systems are given by the establishment of entities which constructivistically the observer detects to possess properties different from those of the component parts, such as in the case of collective behaviors giving rise to ferromagnetism, superconductivity and superfluidity and to social systems such as markets and industrial districts. It must be noted that in a constructivist view the whole is not constituted by parts, but rather the observer identifies parts by using a model in the attempt to explain the whole (observer and designer coincide only for artificial systems). A different partitioning corresponds to different, mutually equivalent or irreducible, models. Systems do not only possess properties, but are also able, in their turn, to make emergent new ones. Examples of emergence of systemic properties in systems (i.e., complex systems) are given by the establishment of properties such as cognitive abilities in natural and artificial systems, collective learning abilities in social systems such as flocks, swarms, markets, firms and functionalities in networks of computers (e.g., in Internet). Evolutionary processes establish properties in living systems. The models of these processes introduced so far are based on theories of phase transitions, of bifurcations, of dissipative structures, and of Multiple Systems (Collective Beings). On the one hand the ability to identify these processes allows effectiveness without confusing processes of a different nature but having in common the macroscopic and generic establishment of systems. This concerns a number of disciplinary contexts such as Physics, Cognitive Science, Biology, Artificial Intelligence, Economics. On the other hand the attempt to build a General Theory of Emergence corresponds to Von Bertalanffy’s project for a General System Theory. The conference will then focus upon these issues from theoretical, experimental, applicative, epistemological and philosophical points of view. We take this opportunity to mention an important, even if not explicit, outcome of the conference. The scientific committee and we, the editors, had the duty and benefit of this outcome and now we have the pleasure of sharing it with the readers. As it is well known, the scientific and cultural level of scientific journals and edited books is assumed to be assured by a good refereeing by the editorial board and the scientific committee. The task is supposed to be quite “easy”

x

Preface

when dealing with topics having general acceptance in academic and research contexts, robust methodologies, and consolidated literature. Consistency is assumed to be assured, in short, by the complete state of the art, and consequently grounded on the application of well-described approaches, consistent reasoning, supporting examples, validation procedures, so as to get coherent conclusions. Traditionally, the systemic community (the one we criticize in the Manifesto) has always tolerated low ‘grades’ in those areas as balanced by the need to break disciplinary well-defined barriers and approaches and encourage focus on new aspects not regulated by classic rules of acceptance. The purpose was to don’t take the risk of suffocating ideas able to generate interesting cultural processes despite their imprecise formulation, even presenting an interesting inconsistence. This was the age when to be inter- and transdisciplinary was a challenge (actually, it is still so in several universities). As emphasized in our Manifesto, disciplinary scientific research had the need to become more and more interdisciplinary, independently from roles, efforts and recommendations of system societies. The challenge for the systemic movement is, in our view, to convert this need into a theoretical result stemming from a General Theory of Emergence intended as a Theory of Change. The challenge is not only at theoretical level, but also at educational level (e.g., in which university department make such a research?). At the same time we have today available an enormous amount of knowledge and we have to face the temptation to explain-all-with-previousknowledge (like in Science). In this context we may lack approaches suitable for recognize and establish new paradigms, inhomogeneous in principle with the old ones. At the same time we lack ways to assure quality levels (e.g. “What if Simplicio had had computers available?”). One consequence of the unavailability of a General Theory of Emergence as a Theory of Change is the unavailability of a robust methodology for evaluating contributions having this mission. The attempt to evaluate each contribution as a disciplinary contribution may imply the lack of appreciation for innovative, inter- and trans-disciplinary systemic meaning. The problem relates to the production of scientific knowledge and educational systems having to deal with an enormous amount of available knowledge by using often old approaches, methodologies and technologies. How to recognize that a wrong, intelligent idea may be more important than a right, not-so-intelligent idea expected to be homologated because of its homogeneity with the established knowledge?

Preface

xi

Is the system community, in force of its historical attention and mission related to inter- and trans-disciplinarity, able to face this challenge in general, i.e. propose innovative approaches and methodologies able to guarantee, test and validate inter- and trans-disciplinary consistency and robustness? We will try to contribute, on the basis of our experience and research activity, to the introduction of proposals and methodologies. The Italian Systems Society is trying to play a significant role in this process. The conference was articulated in different sessions able to capture both the theoretical aspects of emergence as introduced above and the applicative ones: 1. Emergence in Architecture. 2. Processes of emergence in Economics and Management. 3. Emergence. 4. Emergence in social systems. 5. Emergence in Artificial Intelligence. 6. Emergence in Medicine. 7. Models and systems. 8. Theoretical problems of Systemics. 9. Cognitive Science. We conclude by emphasizing that we are aware of how much the scientific community focuses on the available knowledge, as a very comprehensible attitude. By the way, we also have the dream of inter-related forms of knowledge, one represented and modelled into the other, in which meanings have simultaneous multi-significance contributing to generate hierarchies allowing to deal with the meaning of human existence. With this dream in mind we use the bricks of science to contribute to make emergent a new multidimensional knowledge. Gianfranco Minati AIRS president Eliano Pessa Co-Editor Mario Abram Co-Editor

This page intentionally left blank

PROGRAM COMMITTEE

G. Minati (chairman)

Italian Systems Society

E. Pessa (co-chairman)

University of Pavia

L. Biggiero

LUISS University, Rome

G. Bruno

University of Rome “La Sapienza”

V. Coda

“Bocconi” University, Milan

S. Della Torre

Polytechnic University of Milan

V. Di Battista

Polytechnic University of Milan

S. Di Gregorio

University of Calabria

I. Licata

Institute for Basic Research, Florida, USA

M.P. Penna

University of Cagliari

R. Serra

University of Modena and Reggio Emilia

G. Tascini

University of Ancona

G. Vitiello

University of Salerno

This page intentionally left blank

CONTRIBUTING AUTHORS

Abram M.R. Alberti M. Allievi P. Arecchi F.T. Argentero P. Arlati E. Avolio M.V. Battistelli A. Bednar P.M. Bich L. Biggiero L. Bonfiglio N. Bouchard V. Bruno G. Buttiglieri F. Canziani A. Carletti T. Cirina L. Colacci A. Collen A. D’Ambrosio D. Damiani C. David S. Del Giudice E. Dell’Olivo B.

Della Torre S. Di Battista V. Di Caprio U. Di Gregorio S. Ferretti M.S. Filisetti A. Giallocosta G. Giunti M. Graudenzi A. Gregory R.L. Guberman S. Ingrami P. Lella L. Licata I. Lupiano V. Magliocca L.A. Marconi P.L. Massa Finoli G. Minati G. Mocci S. Montesanto A. Mura M. Odoardi C. Paoli F. Penna M.P.

Percivalle S. Pessa E. Picci P. Pietrocini E. Pinna B. Poli I. Puliti P. Ramazzotti P. Ricciuti A. Rocchi C. Rollo D. Rongo R. Sechi C. Serra R. Setti I. Sevi E. Sforna M. Spataro W. Stara V. Tascini G. Terenzi G. Trotta A. Villani M. Vitiello G.

This page intentionally left blank

CONTENTS

Dedication

v

Preface

vii

Program Committee

xiii

Contributing Authors

xv

Contents

xvii

Opening Lecture Coherence, Complexity and Creativity Fortunato Tito Arecchi

3

Emergence in Architecture Environment and Architecture – A Paradigm Shift Valerio Di Battista

37

Emergence of Architectural Phenomena in the Human Habitation of Space Arne Collen

51

Questions of Method on Interoperability in Architecture Ezio Arlati, Giorgio Giallocosta

67

Comprehensive Plans for a Culture-Driven Local Development: Emergence as a Tool for Understanding Social Impacts of Projects on Built Cultural Heritage Stefano Della Torre, Andrea Canziani Systemic and Architecture: Current Theoretical Issues Giorgio Giallocosta

xvii

79 91

xviii

Contents

Processes of Emergence in Economics and Management Modeling the 360° Innovating Firm as a Multiple System or Collective Being Véronique Bouchard The COD Model: Simulating Workgroup Performance Lucio Biggiero, Enrico Sevi Importance of the Infradisciplinary Areas in the Systemic Approach Towards New Company Organisational Models: the Building Industry Giorgio Giallocosta

103 113

135

Systemic Openness of the Economy and Normative Analysis Paolo Ramazzotti

149

Motivational Antecedents of Individual Innovation Patrizia Picci, Adalgisa Battistelli

163

An E-Usability View of the Web: A Systemic Method for User Interfaces Vera Stara, Maria Pietronilla Penna, Guido Tascini

181

Emergence Evolutionary Computation and Emergent Modeling of Natural Phenomena R. Rongo, W. Spataro, D. D’Ambrosio, M.V. Avolio, V. Lupiano, S. Di Gregorio

195

A New Model for the Organizational Knowledge Life Cycle Luigi Lella, Ignazio Licata

215

On Generalization: Constructing a General Concept from a Single Example Shelia Guberman

229

General Theory of Emergence Beyond Systemic Generalization Gianfranco Minati

241

Uncertainty, Coherence, Emergence Giordano Bruno

257

Emergence and Gravitational Conjectures Paolo Allievi, Alberto Trotta

265

Contents

xix

Emergence in Social Systems Inducing Systems Thinking in Consumer Societies Gianfranco Minati, Larry A. Magliocca

283

Contextual Analysis. A Multiperspective Inquiry into Emergence of Complex Socio-Cultural Systems Peter M. Bednar

299

Job Satisfaction and Organizational Commitment: Affective Commitment Predictors in a Group of Professionals Maria Santa Ferretti

313

Organizational Climate Assessment: A Systemic Perspective Piergiorgio Argentero, Ilaria Setti Environment and Urban Tourism: An Emergent System in Rhetorical Place Identity Definitions Marina Mura

331

347

Emergence in Artificial Intelligence Different Approaches to Semantics in Knowledge Representation S. David, A. Montesanto, C. Rocchi Bidimensional Turing Machines as Galilean Models of Human Computation Marco Giunti

365

383

A Neural Model of Face Recognition: A Comprehensive Approach Vera Stara, Anna Montesanto, Paolo Puliti, Guido Tascini, Cristina Sechi

407

Anticipatory Cognitive Systems: A Theoretical Model Graziano Terenzi

425

Decision Making Models within Incomplete Information Games Natale Bonfiglio, Simone Percivalle, Eliano Pessa

441

Emergence in Medicine Burnout and Job Engagement in Emergency and Intensive Care Nurses Piergiorgio Argentero, Bianca Dell’olivo

455

xx

Contents

The “Implicit” Ethics of a Systemic Approach to the Medical Praxis Alberto Ricciuti Post Traumatic Stress Disorder in Emergency Workers: Risk Factors and Treatment Piergiorgio Argentero, Bianca Dell’Olivo, Ilaria Setti State Variability and Psychopathological Attractors. The Behavioural Complexity as Discriminating Factor between the Pathology and Normality Profiles Pier Luigi Marconi

473

487

503

Models and Systems Decomposition of Systems and Complexity Mario R. Abram How many Stars are there in Heaven ? The results of a study of Universe in the light of Stability Theory Umberto Di Caprio

533

545

Description of a Complex System through Recursive Functions Guido Massa Finoli

561

Issues on Critical Infrastructures Mario R. Abram, Marino Sforna

571

Theoretical Problems of Systemics Downward Causation and Relatedness in Emergent Systems: Epistemological Remarks Leonardo Bich

591

Towards a General Theory of Change Eliano Pessa

603

Acquired Emergent Properties Gianfranco Minati

625

The Growth of Populations of Protocells Roberto Serra, Timoteo Carletti, Irene Poli, Alessandro Filisetti

641

Investigating Cell Criticality R. Serra, M. Villani, C. Damiani, A. Graudenzi, P. Ingrami, A. Colacci

649

Contents

xxi

Relativistic Stability. Part 1 - Relation Between Special Relativity and Stability Theory in the Two-Body Problem Umberto Di Caprio

659

Relativistic Stability. Part 2 - A Study of Black-Holes and of the Schwarzschild Radius Umberto Di Caprio

673

The Formation of Coherent Domains in the Process of Symmetry Breaking Phase Transitions Emilio Del Giudice, Giuseppe Vitiello

685

Cognitive Science Organizations as Cognitive Systems. Is Knowledge an Emergent Property of Information Networks? Lucio Biggiero

697

Communication, Silence and Miscommunication Maria Pietronilla Penna, Sandro Mocci, Cristina Sechi

713

Music: Creativity and Structure Transitions Emanuela Pietrocini

723

The Emergence of Figural Effects in the Watercolor Illusion Baingio Pinna, Maria Pietronilla Penna

745

Continuities and Discontinuities in Motion Perception Baingio Pinna, Richard L. Gregory

765

Mother and Infant Talk about Mental States: Systemic Emergence of Psychological Lexicon and Theory of Mind Understanding D. Rollo, F. Buttiglieri

777

Conflict in Relationships and Perceived Support in Innovative Work Behavior Adalgisa Battistelli, Patrizia Picci, Carlo Odoardi

787

Role Variables vs. Contextual Variables in the Theory of Didactic Systems Monica Alberti, Lucia Cirina, Francesco Paoli

803

This page intentionally left blank

OPENING LECTURE

This page intentionally left blank

COHERENCE, COMPLEXITY AND CREATIVITY

FORTUNATO TITO ARECCHI Università di Firenze and INOA, Largo E. Fermi, 6 - 50125 Firenze, Italy E-mail: [email protected]

We review the ideas and experiments that established the onset of laser coherence beyond a suitable threshold. That threshold is the first of a chain of bifurcations in a non linear dynamics, leading eventually to deterministic chaos in lasers. In particular, the so called HC behavior has striking analogies with the electrical activity of neurons. Based on these considerations, we develop a dynamical model of neuron synchronization leading to coherent global perceptions. Synchronization implies a transitory control of neuron chaos. Depending on the time duration of this control, a cognitive agent has different amounts of awareness. Combining this with a stream of external inputs, one can point at an optimal use of internal resources, that is called cognitive creativity. While coherence is associated with long range correlations, complexity arises whenever an array of coupled dynamical systems displays multiple paths of coherence. What is the relation among the three concepts in the title? While coherence is associated with long range correlations, complexity arises whenever an array of coupled dynamical systems displays multiple paths of coherence. Creativity corresponds to a free selection of a coherence path within a complex nest. As sketched above, it seems dynamically related to chaos control. Keywords: heteroclinic chaos, homoclinic chaos, quantum uncertainty, feature binding, conscious perception.

1. Introduction - Summary of the presentation Up to 1960 in order to have a coherent source of light it was necessary to filter out a noisy regular lamp. Instead, the laser realizes the dream of shining a vacuum state of the electromagnetic field with a classical antenna, thus inducing a coherent state, which is a translated version of the vacuum state, with a minimum quantum uncertainty. As a fact, the laser reaches its coherent state through a threshold transition, starting from a regular incoherent source. Accurate photon statistics measurements proved the coherence quality of the laser as well the threshold transition phenomena, both in stationary and transient situations. The threshold is the first of a chain of dynamical bifurcations; in the 1980’s the successive bifurcations leading to deterministic chaos were explored. Furthermore, the coexistence of many laser modes in a cavity with high Fresnel

3

4

F.T. Arecchi

number gives rise to a complex situation, where the modes behave in a nested way, due to their mutual couplings, displaying a pattern of giant intensity peaks whose statistics is by no means Gaussian, as in speckles. Among the chaotic scenarios, the so called HC (Heteroclinic chaos), consisting of trains of equal spikes with erratic inter-spike separation, was explored in CO2 and in diode lasers with feedback. It looks as the best implementation of a time code. Indeed, networks of coupled HC systems may reach a state of collective synchronization lasting for a finite time, in presence of a suitable external input. This opens powerful analogies with the feature binding phenomenon characterizing neuron organization in a perceptual task. The dynamics of a single neuron is suitably modeled by a HC laser; thence, the collective dynamics of a network of coupled neurons can be realized in terms of arrays of coupled HC lasers [5,23]. Thus, synchronization of an array of coupled chaotic lasers is a promising tool for a physics of cognition. Exploration of a complex situation would require a very large amount of time, in order to classify all possible coherences, i.e. long range correlations. In cognitive tasks facing a complex scenario, our strategy consists in converging to a decision within a finite short time. Any conscious perception (we define conscious as that eliciting a decision) requires 200 ms, whereas the loss of information in the chaotic spike train of a single neuron takes a few ms. The interaction of a bottom-up signal (external stimulus) with a top-down modification of the control parameters (induced by the semantic memory) leads to a collective synchronization lasting 200 ms: this is the indicator of a conscious perception. The operation is a control of chaos, and it has an optimality; if it lasts less than 200 ms, no decisions emerge, if it lasts much longer, there is no room for sequential cognitive tasks. We call creativity this optimal control of neuronal chaos. It amounts to select one among a large number of possible coherences all present in a complex situation. The selected coherence is the meaning of the object under study. 2. Coherence 2.1. Classical notion of coherence Before the laser, in order to have a coherent source of light it was necessary to filter out a noisy regular lamp. Fig. 1 illustrates the classical notion of coherence, with reference to the Young interferometer. A light source with aperture ∆ x illuminates a screen with two holes A and B (that we can move to positions A’ and B’). We take the source as made of the superposition of

5

Coherence, Complexity and Creativity

source

detector

Figure 1. Young interferometer: a light source of aperture x illuminates a screen with two holes in it. Under suitable conditions, the phase interference between the fields leaking trough the two holes gives rise to interference fringes, as the point like detector is moved on a plane transverse to the propagation direction.

independent plane waves, without mutual phase relations. The single plane wave is called mode, since it is a solution of the wave equation within the cavity containing the source. Each mode, passing through ∆ x , is diffraction broadened into a cone of aperture θ = λ ∆ x . At left of the screen, the light from A and B is collected on a detector, whose electrical current is proportional to the impinging light power, that is, to square modulus of the field. The field is the sum of the two fields E1 and E2 from the two holes. The modulus must be averaged over the observation time, usually much longer than the optical period; we call |E1 + E2|2 this average. The result is the sum of the two separate intensities I1 = |E1|2 and I2 = |E2|2 , plus the cross phased terms E1∗ E2 + E2∗ E1 . These last ones increase or reduce I1 + I 2 , depending on the relative phase, hence interference fringes are observed as we move the detector on a plane transverse to the light propagation, thus changing the path lengths of the two fields. Fringe production implies that the phase difference be maintained during the time of the average … , this occurs only if the two fields leaking through the two holes belong to the same mode, that is, if observation angle, given by the distance AB divided by the separation r between screen and detector, is smaller than the diffraction angle θ = λ ∆ x . If instead it is larger, as it occurs e.g. when holes are in positions A′ , B′ , then the detector receives contributions from distinct modes, whose phases fluctuate over a time much shorter than the averaging time. Hence, the phased

6

F.T. Arecchi

terms are washed out and no fringes appear. We call coherence area that area SAB on the screen which contains pairs of points A, B such that the collection angle be at most equal to the diffraction angle. SAB subtends a solid angle given by

S AB =

λ2 ⋅ r 2 (∆ x) 2

(1)

The averaged product of two fields in positions 1 = A and 2 = B is called first order correlation function and denoted as

G (1) (1,2) = E1∗ E2

(2)

In particular for 1 = 2, G(1)(1,1) = E1*E1 is the local intensity at point 1. Points 1 and 2 correspond to holes A and B of the Young interferometer; their separation is space-like if the detector is at approximately the same distance from the holes. Of course, fringes imply path differences comparable with the wavelength, but anyway much shorter than the coherence time

τ = 1 ∆ω

(3)

of a narrowband (quasi-monochromatic) light source¸ indeed if the line breadth is much smaller than the optical frequency, ∆ω > T . In the case of the Michelson interferometer, 1 and 2 are the two mirror positions, which are separated time-like. Fringe disappearance in this case means that the time separation between the two mirrors has become larger than the coherence time. 2.2. Quantum notion of coherence The laser realizes the dream of shining the current of a classical antenna into the vacuum state of an electromagnetic field mode, thus inducing a coherent state as a translated version of the vacuum state, with a minimum quantum uncertainty (Fig. 2). We know from Maxwell equations that the a single field mode obeys a harmonic oscillator (HO) dynamics. The quantum HO has discrete energy states equally separated by ω starting from a ground (or vacuum) state with energy ω 2 . Each energy state is denoted by the number (0,1,2, … ,n, …) of energy quanta ω above the ground state. In a coordinate q representation, any state with a fixed n is delocalised, that is, its wavefunction is spread inside the region

Coherence, Complexity and Creativity

7

Figure 2. Quantum harmonic oscillator in energy-coordinate diagram. Discrete levels correspond to photon number states. A coherent state is a translated version of the ground state; its photon number is not sharply defined but is spread with a Poisson distribution.

confined by the parabolic potential (see e.g. the dashed wave for n = 5). Calling p = mv the HO impulse, the n state has an uncertainty in the joint coordinateimpulse measurement increasing as

∆ q ∆ p = (n + 1 2) .

(4)

The vacuum state, with n = 0, has the minimum uncertainty

∆q ∆p = 1 2 .

(4’)

If now we consider a version of the vacuum state translated by (where is proportional to q), this is a quantum state still with minimum uncertainty, but with an average photon number equal to the square modulus |α2| (in the example reported in the figure we chose |α2| = 5). It is called coherent state. It oscillates at the optical frequency in the q interval allowed for by the confining potential. It maintains the instant localization, at variance with a number state. The coherent state pays this coordinate localization by a Poisson spread of the photon number around its average |α2|. The quantum field vacuum state shifted by a classical current had been introduced in 1938 by Bloch and Nordsieck; in 1963 R. Glauber showed that these states have maximal coherence, and that a laser emits such a type of light, since the collective light emission in the laser process can be assimilated to the radiation of a classical current.

8

F.T. Arecchi

While the fringe production is just a test of the modal composition of the light field, the Hanbury Brown and Twiss interferometer (HBT) implies the statistical spread of the field amplitude. HBT was introduced in 1956 as a tool for stellar observation (Fig. 3) in place of Michelson (M) stellar interferometer. M is based on summing on a detector the fields from two distant mirrors, in order to resolve the angular breadth of a star (that is, its diameter, or the different directions of two stars in binary components). The more distant are the mirrors the higher the resolution. However the light beams deflected by the two mirrors undergo strong dephasing in the horizontal propagation and this destroys the fringes. In HBT, the two mirrors are replaced by two detectors, whose output currents feed a correlator; now the horizontal path is within a cable, hence not affected by further dephasing. The working principle is intensity correlation (rather than field), which for a Gaussian statistics (as expected from thermal sources as the stars) yields the product of the two intensities plus the square modulus of the field correlation as provided by a standard interferometer, that is, 2

G ( 2) (1,2) = E1 | E2 |2 = I1I 2 + | G (1) |2

(5)

Instead, Glauber had proved that for a coherent state, all the higher order correlation functions factor as products of the lowest one, that is,

G ( n ) (1,2,

, n) = G (1) (1) G (1) (2)

G (1) (n)

(6)

in particular, for n = 2 , we have

G ( 2) (1,2) = G (1) (1) G (1) (2) .

(6’)

G (1) is just the intensity; thus a coherent state should yield only the first term of HBT, without the correlation between the two distant detectors. In 1966 the comparison between a laser and a Gaussian field was measured time-wise rather than space-wise as shown in Fig. 4. The laser light displays no HBT, as one expects from a coherent state. The extra term in the Gaussian case doubles the zero time value. As we increase the time separation between the two “instantaneous” intensity measurements (by instantaneous we mean that the integration time is much shorter than the characteristic coherence time of the Gaussian fields), the extra HBT terms decays and eventually disappears. We have scaled the time axis so that HBT for different coherence times coincide.

Coherence, Complexity and Creativity

Optical field

Photon detector

Electric signal

Electronic correlator

9

Figure 3. Left: the Michelson stellar interferometer M; it consists of two mirrors which collect different angular views of a stellar object and reflect the light to a single photon detector through long horizontal paths (10 to 100 meters) where the light phase is affected by the ground variations of the refractive index (wavy trajectories). Right: the Hanbury-Brown and Twiss (HBT) interferometer; mirrors are replaced by detectors and the current signal travel in cables toward an electronic correlator, which performs the product of the two instant field intensities E1* E1, E2* E2 and averages it over a long time [32].

Coherence times are assigned through the velocities. of a random scatterer, as explained in the next sub-section. Notice that Fig. 4 reports coherence times of the order of 1 ms. In the electronic HBT correlator, this means storing two short time intensity measurements (each lasting for example 50 ns) and then comparing them electronically. If we tried to measure such a coherence time by a Michelson interferometer, we would need a mirror separation of the order of 300 km! 2.3. Photon statistics (PS) As a fact, the laser reaches its coherent state through a threshold transition, starting from a regular incoherent source. Accurate photon statistics measurements proved the coherence quality of the laser as well the threshold transition phenomena, both in stationary and transient situations. We have seen in Fig. 2 that a coherent state yields a Poisson spread in the photon number, that is, a photon number statistics as

p ( n) =

n n − e n!

n

(7)

10

F.T. Arecchi

Figure 4. Laboratory measurement of HBT for Gaussian light sources with different coherence times; for each case, the first order correlations between the signals sampled at different times decay with the respective coherence time, and asymptotically only the product of the average intensities (scaled to 1) remains. The laser light displays no HBT, as one expects from a coherent state. [14]

where n = |α2| is the average photon number. This provides high order moments, whereas HBT for equal space-time positions 1 = 2 would yield just the first and the second moment. Thus PS is statistically far more accurate, however it is confined to within a coherence area and a coherence time. If now we couple the coherent state to an environment, we have a spread of coherent states given by a spread P(α). The corresponding PS is a weighted sum of Poisson distributions with different average values n = |α2| . In Fig. 5 we report the statistical distributions of photocounts versus the count number. If the detector has high efficiency, they well approximate the photon statistics of the observed light source. A few words on how to build a Gaussian light source. A natural way would be to take a black-body source, since at thermal equilibrium P(α) is Gaussian. However its average photon number would be given by Planck’s formula as

n =

1 exp(− ω / kT ) − 1

.

(8)

For visible light ω = 2 eV and current blackbody temperatures (remember that 104 K ≈ 1 eV ) we would have n > c / 2L, simultaneously above threshold. In such a case the nonlinear mode-mode coupling, due to the medium interaction, gives an overall high dimensional dynamical system which may undergo chaos. This explains the random spiking behavior of long lasers. The regular spiking in time associated with mode

Coherence, Complexity and Creativity

21

Figure 17. Photorefractive oscillator, with the photorefractive effect enhanced by a LCLV (Liquid Crystal Light Valve). Experimental setup; A is an aperture fixing the Fresnel number of the cavity, z =0 corresponds to the plane of the LCLV; z1, z2, z3 are the three different observation planes. Below: 2-dimensional complex field, with lines of zero real part (solid) and lines of zero imaginary part (dashed). At the intersection points the field amplitude is zero and its phase not defined, so that the circulation of the phase gradient around these points is non-zero (either ±2 ) yielding phase singularities. [15,4,10]

locking is an example of mutual phase synchronization, akin to the regular spiking reported in Fig. 16. 3.2.2. b) transverse case Inserting a photorefractive crystal in a cavity, the crystal is provided high optical gain by a pump laser beam. As the gain overcomes the cavity losses, we have a coherent light oscillator. Due to the narrow linewidth of the crystal, a single longitudinal mode is excited; however, by an optical adjustment we can have large Fresnel numbers, and hence many transverse modes. We carried a research line starting from 1990 [15,16, for a review see 4]. Recently we returned to this oscillator but with a giant photorefractive effect provided by the coupling of a photorefractive slice to a liquid crystal [10,6,12] (Fig. 17). The inset in this figure shows how phase singularities appear in a 2D wave field. A phase gradient circulation ±2π is called a topological charge of ±1 respectively. A photodetector responds to the modulus square of the field amplitude. To have a phase information, we superpose a plane wave light to the 2D pattern, obtaining results illustrated in Fig. 18. For a high Fresnel number we have a number of

22

F.T. Arecchi

Figure 18. Left: a phase singularity is visualized by superposing an auxiliary coaxial plane wave to the optical pattern of the photorefractive oscillator; reconstruction of the instantaneous phase surface: perspective and equi-phase plots. Right: if the auxiliary beam is tilted, we obtain interference fringes, interrupted at each phase singularity (± correspond to ±2 circulation, respectively). The digitized fringe plots correspond to: upper plot (Fresnel number about 3): 6 defects of equal topological charge against 1 of opposite charge; lower plot (Fresnel number close to 10): almost 100 singularities with balanced opposite charges, besides a small residual unbalance [16].

singularities scaling as the square of the Fresnel number [9]. Referring to the inset of Fig. 17, when both intersections of the two zero lines are within the boundary, we expect a balance of opposite topological charges. However, for small Fresnel numbers, it is likely that only one intersection is confined within the boundary; this corresponds to an unbalance, as shown in Fig. 18, upper right. The scaling with the Fresnel number is purely geometric and does not imply dynamics. The statistics of zero-field occurrences can be predicted on purely geometric considerations, as done for random speckles. If instead we look at the high intensity peak in between the zeros, the high fields in a nonlinear medium give a strong mode-mode coupling which goes beyond speckles. This should result from the statistical occurrence of very large peaks. In order to do that, we collect space-time frames as shown in Fig. 19, with the help of the CCD +grabber set up shown in Fig. 17. We don’t have yet a definite 2D comparison with speckles. However, a 1D experiment in an optical fiber has

Coherence, Complexity and Creativity

23

Figure 19. Photorefractive oscillator: Spatiotemporal profile extracted from the z2 movie. [10]

produced giant optical spikes with non-Gaussian statistics [43]. The author draw an analogy with the so called “rogue” wave in the ocean which represent a frequent problem to boats, since satellite inspection has shown that they are more frequent than expected on a purely linear basis. We consider the anomalous statistics of giant spikes as a case of complexity, because the mutual coupling in a nonlinear medium makes the number of possible configurations increasing exponentially with the Fresnel number, rather than polynomially. The rest of the paper explores this question: how it occurs that a cognitive agent in a complex situation decides for a specific case, before having scanned all possible cases, that is, how we “cheat” complexity.

4. Physics of cognition – Creativity 4.1. Perception and control of chaos Synchronization of a chain of chaotic lasers provides a promising model for a physics of cognition. Exploration of a complex situation would require a very large amount of time. In cognitive tasks facing a complex scenario, our strategy consists in converging to a decision within a finite short time. Various experiments [36,38] prove that a decision is taken after 200 ms of exposure to a sensory stimulus. Thus, any conscious perception (we define conscious as that

24

F.T. Arecchi

Figure 20. Feature binding: the lady and the cat are respectively represented by the mosaic of empty and filled circles, each one representing the receptive field of a neuron group in the visual cortex. Within each circle the processing refers to a specific detail (e.g. contour orientation). The relations between details are coded by the temporal correlation among neurons, as shown by the same sequences of electrical pulses for two filled circles or two empty circles. Neurons referring to the same individual (e.g. the cat) have synchronous discharges, whereas their spikes are uncorrelated with those referring to another individual (the lady) [42].

eliciting a decision) requires about 200 ms, whereas the loss of information in a chaotic train of neural spikes takes a few msec. Let us consider the visual system; the role of elementary feature detectors has been extensively studied [34]. By now we know that some neurons are specialized in detecting exclusively vertical or horizontal bars, or a specific luminance contrast, etc. However the problem arises: how elementary detectors contribute to a holistic (Gestalt) perception? A hint is provided by [42]. Suppose we are exposed to a visual field containing two separate objects. Both objects are made of the same visual elements, horizontal and vertical contour bars, different degrees of luminance, etc. What are then the neural correlates of the identification of the two objects? We have one million fibers connecting the retina to the visual cortex. Each fiber results from the merging of approximately 100 retinal detectors (rods and cones) and as a result it has its own receptive field. Each receptive field isolates a specific detail of an object (e.g. a vertical bar). We thus split an image into a mosaic of adjacent receptive fields. Now the “feature binding” hypothesis consists of assuming that all the cortical neurons whose receptive fields are pointing to a specific object synchronize the corresponding spikes, and as a consequence the visual cortex

Coherence, Complexity and Creativity

25

Figure 21. ART = Adaptive Resonance Theory. Role of bottom-up stimuli from the early visual stages an top-down signals due to expectations formulated by the semantic memory. The focal attention assures the matching (resonance) between the two streams [27].

organizes into separate neuron groups oscillating on two distinct spike trains for the two objects. Direct experimental evidence of this synchronization is obtained by insertion of microelectrodes in the cortical tissue of animals just sensing the single neuron (Fig. 20) [42]. An array of weakly coupled HC systems represents the simplest model for a physical realization of feature binding. The array can achieve a collective synchronized state lasting for a finite time (corresponding to the physiological 200 ms!) if there is a sparse (non global) coupling, if the input (bottom-up) is applied to just a few neurons and if the inter-neuron coupling is suitably adjusted (top-down control of chaos) [5,23]. Fig. 21 shows the scheme of ART [27]. The interaction of a bottom-up signal (external stimulus) with a top-down change of the control parameters (induced by the semantic memory) leads to a collective synchronization lasting 200 ms: this is the indicator of a conscious perception. The operation is a control of chaos, and it has an optimality; if it lasts less than 200 ms, no decisions emerge; on the contrary, if it lasts much longer, there is no room for sequential cognitive tasks (Fig. 22). The addition of extra degrees of freedom implies a change of code, thus it can be seen as a new level of description of the same physical system.

26

F.T. Arecchi

Figure 22. Chaos is controlled by adding extra-dynamic variables, which change the transverse instability without affecting the longitudinal trajectory. In the perceptual case, the most suitable topdown signals are those which provide a synchronized neuron array with an information lifetime sufficient to activate successive decisional areas (e.g. 200 ms), whereas the single HC neuron has a chaotic lifetime of 2 ms. If our attentional-emotional system is excessively cautious, it provides a top-down correction which may stabilize the transverse instability for ever, but then the perceptual area is blocked to further perceptions.

4.2. From perception to cognition - Creativity We distinguish two types of cognitive task. In type I, we work within a prefixed framework and readjust the hypotheses at each new cognitive session, by a Bayes strategy. Bayes theorem [21] consists of the relation:

P(h | data) = P (data | h)

P ( h) P(data)

(9)

That is: the probability P(h | data ) of an hypothesis h, conditioned by the observed data (this is the meaning of the bar | ) and called a-posteriori probability of h, is the product of the probability P(data | h) that data are generated by an hypothesis h, times the a-priori probability P (h) of that hypothesis (we assume to have a package of convenient hypotheses with different probabilities) and divided the probability P(data) of the effectively occurred data. As shown in Fig. 23, starting from an initial observation and formulating a large number of different hypotheses, the one supported by the experiment suggests the most appropriate dynamical explanation. Going a step forward and repeating the Bayes procedure amounts to climbing a probability mountain along a steepest gradient line.

27

Coherence, Complexity and Creativity final condition INFORMATION Fitness = Probability mountains

a-posteriori probability a-priori probability

Darwin = Bayesian strategy

initial condition

Figure 23. Successive applications of the Bayes theorem to the experiments. The procedure is an ascent of the Probability Mountain through a steepest gradient line. Each point of the line carries an information related to the local probability by Shannon formula. Notice that Darwinian evolution by mutation and successive selection of the best fit mutant is a sequential implementation of Bayes theorem. [19,18]

On the other hand, a complex problem is characterized by a probability landscape with many peaks (Fig. 24). Jumping from a probability hill to another is not Bayesian; I call it type II cognition. A deterministic computer can not do it. In human cognition, Type II is driven by hints suggested by the context (semiosis) yet not included in the model. Type II task is a creativity act because it goes beyond it implies a change of code, at variance with Type I, which operates within a fixed code. The ascent to a single peak can be automatized in a steepest gradient program; once the peak has been reached, the program stops, any further step would be a downfall. A non-deterministic computer can not perform the jumps of Type II, since it intrinsically lacks semiotic abilities. In order to do that, the computer must be assisted by a human operator. We call “meaning” the multi-peak landscape and “semantic complexity” the number of peaks. However, this is a fuzzy concept, which varies as our comprehension evolves (Fig. 25). Let us discuss in detail the difference between type I cognitive task, which implies changing hypothesis h within a model, that is, climbing a single mountain, and Type II cognitive task which implies changing model, that is, jumping over to another mountain.

28

F.T. Arecchi complexity MEANING

INFORMATION

STOP!!! Bayes without semiosis

complication

Figure 24. Semantic complexity - A complex system is one with a many-peak probability landscape. The ascent to a single peak can be automatized by a steepest gradient program. On the contrary, to record the other peaks, and thus continue the Bayes strategy elsewhere, is a creativity act, implying a holistic comprehension of the surrounding world (semiosis). We call “meaning” the multi-peak landscape and “semantic complexity” the number of peaks. It has been guessed that semiosis is the property that discriminates living beings from Turing machines [39]; here we show that a nonalgorithmic procedure, that is, a non-Bayesian jump from one model to another is what we have called creativity. Is semiosis equivalent to creativity? [19,18].

We formalize a model as a set of dynamical variables xi (i = 1,2, , N ) , N being the number of degrees of freedom, with the equations of motion

xi = Fi ( x1 ,

, x N ; µ1 ,

, µM )

(10)

where Fi are the force laws and the M numbers µ represent the control parameters. The set {F, x, µ} is the model. Changing hypotheses within a model means varying the control parameters, as we do when exploring the transition from regular to chaotic motion in some model dynamics. Instead, changing code, or model, means selecting different sets y ,ν , G of degrees of freedom, control parameters and equations of motion as follows:

yi = Gi ( y1 ,

, y R ;ν 1 ,

,ν L )

(11)

where R and L are different respectively from N and M. The set {G , y, ν } is the new model. While changing hypotheses within a model is an a-semiotic procedure that can be automatized in an computerized expert system, changing model implies catching the meaning of the observed world, and this requires what has been

29

Coherence, Complexity and Creativity Re-coding = creativity C computation Semiosis

Newton

Scientific Theory 0

K

Figure 25. C-K diagram (C = computational complexity; K = Information loss rate in chaotic motion): Comparison between the procedure of a computer and a semiotic cognitive agent (say: a scientist). The computer operates within a single code and C increases with K. A scientist explores how adding different degrees of freedom one can reduce the high K of the single-code description. This is equivalent to the control operation of Fig. 22; it corresponds to a new model with reduced C and K. An example is offered by the transition from a molecular dynamics to a thermodynamic description of a many body system. Other examples are listed in Table 1. The BACON program [41] could retrieve automatically Kepler’s laws from astronomical data just because the solar system approximated by Newton two-body interactions is chaos-free.

called embodied cognition [46]. Embodied cognition has been developed over thousands of generations of evolutionary adaptation, and we are unable so far to formalize it as an algorithm. This no-go statement seems to be violated by a class of complex systems, which has been dealt with successfully by recursive algorithms. Let us consider a space lattice of spins, with couplings that can be ferro or anti-ferromagnetic in a disordered, but frozen way (spin glass at zero temperature, with quenched disorder). It will be impossible to find a unique ground state. For instance having three spins A, B, and C in a triangular lattice, if all have ferromagnetic interaction, then the ground state will consist of parallel spins, but if instead one (and only one) of the mutual coupling is anti-ferromagnetic, then there will be no satisfactory spin orientation compatible with the coupling (try with: A-up, Bup, C-up; it does not work; then try to reverse a single spin, but it does not work either). This model has a cognitive flavor, since a brain region can be modeled as a lattice of coupled neurons with coupling either excitatory or inhibitory, thus resembling a spin glass, [33,1,45]. We have a large number of possible ground

30

F.T. Arecchi Table 1. From complication to complexity: four cases of creativity. 1 - electricity - magnetism - optics

Electromagnetic equations (Maxwell)

2 - Mendeleev table

Quantum atom (Bohr, Pauli)

3 - zoo of 200 elementary particles

Quarks (M. Gell Mann)

4 - scaling laws in phase transitions

Renormalization group (K. Wilson)

states, all including some frustration. Trying to classify all possible configurations is a task whose computational difficulty (either, program length or execution time) diverges exponentially with the size of the system. Sequentially related changes of code have been successfully introduced to arrive at finite-time solutions. [37,44]. Can we say that the mentioned solutions realize the reductionistic dream of finding a suitable computer program that not only climbs the single probability peak, but also is able to choose the highest peak? If so, the optimization problem would correspond to understanding the meaning of the object under scrutiny. We should realize however that spin glasses are frozen objects, given once for ever. A clever search of symmetries has produced a spin glass theory [37] that, like the Renormalization Group (RG) for critical phenomena [47] discovers a recursive procedure for changing codes in an optimized way. Even though the problem has a large number of potential minima, and hence of probability peaks, a suitable insight in the topology of the abstract space embedding the dynamical system has led to an optimized trajectory across the peaks. In other words, the correlated clusters can be ordered in a hierarchical way and a formalism analogous to RG applied. It must be stressed that this has been possible because the system under scrutiny has a structure assigned once for ever. In everyday tasks, we face a system embedded in an environment, which induces a-priori unpredictable changes in course of time. This rules out the nice symmetries of hierarchical approaches, and rather requires an adaptive approach. Furthermore, a real life context sensitive system has to be understood within a reasonably short time, in order to take vital decisions about it.

Coherence, Complexity and Creativity

31

We find again a role of control of chaos in cognitive strategies, whenever we go beyond the limit of a Bayes strategy. We call creativity this optimal control of neuronal chaos. Four cases of creative science are listed in Table 1. Furthermore, Fig. 24 sketches the reduction of complexity and chaos which results from a creative scientific step. Appendix. Haken theory of laser threshold [28,29,30,34] We summarize in Table 2 the Langevin equation for a field E, ruled by a dynamics f (E ) corresponding to the atomic polarization and perturbed by a noise. The noise has zero average and a delta-like correlation function with amplitude D given by the spontaneous emission of the N2 atoms in the upper state. The time dependent probability P( E , t ) for E obeys a Fokker-Planck equation. In the stationary limit of zero time derivative, the Fokker-Planck equation is easily solved and gives a negative exponential on V (E ) which is the potential of the force f (E ) . Below laser threshold, f (E ) is linear, V quadratic and P(E ) Gaussian. Above threshold, f has a cubic correction, V is quartic and P(E ) displays two peaks at the minima of the quartic potential. Table 2. Langevin equation

E = f (E) + ξ

ξ 0 ξ1 = 2 Dδ (t ) D = γ spont N 2 ∂2P ∂ ∂P f (E) + D 2 =− ∂E ∂t ∂E P( E ) ≈ e −V ( E )

Fokker-Planck equation Stationary solution

D

V ( E ) = − f ( E )dE

f ( E ) = −αE 2

f ( E ) = +αE − β E E

Force laws, over/under threshold

32

F.T. Arecchi

References Papers of which I am author or co-author can be found in my home page: www.inoa.it/home/arecchi , List of publications - Research papers in Physics

1. D.J. Amit, H. Gutfreund, H. Sompolinski, Phys. Rev A 32, 1007 (1985). 2. F.T. Arecchi, Phys. Rev. Lett. 15, 912 (1965). 3. F.T. Arecchi, Proc. E. Fermi School 1967 in Quantum Optics, Ed. R.J. Glauber (Academic Press, New York, 1969), pp. 57-110.

4. F.T. Arecchi, in Nonlinear dynamics and spatial complexity in optical systems (Institute of Physics Publishing, Bristol, 1993), pp. 65-113.

5. F.T. Arecchi, Physica A 338, 218-237 (2004). 6. F.T. Arecchi, in La Fisica nella Scuola, Quaderno 18, (Epistemologia e Didattica della Fisica) Bollettino della Assoc. Insegn. Fisica 40(1), 22-50 (2007).

7. F.T. Arecchi, E. Allaria, A. Di Garbo, R. Meucci, Phys. Rev. Lett 86, 791 (2001). 8. F.T. Arecchi, A. Berné, P. Burlamacchi, Phys. Rev. Lett. 16, 32 (1966). 9. F.T. Arecchi, S. Boccaletti, P.L. Ramazza, S. Residori, Phys. Rev. Lett. 70, 2277, (1993).

10. F.T. Arecchi, U. Bortolozzo, A. Montina, J.P. Huignard, S. Residori, Phys. Rev. Lett. 99, 023901 (2007).

11. F.T. Arecchi, V. Degiorgio, B. Querzola, Phys. Rev. Lett. 19, 1168 (1967). 12. F.T. Arecchi, V. Fano, in Hermeneutica 2007, Annuario di filosofia e teologia, (Morcelliana, Brescia, 2007), pp. 151-174.

13. F.T. Arecchi, W. Gadomski, R. Meucci, Phys. Rev. A 34, 1617 (1986). 14. F.T. Arecchi, E. Gatti, A. Sona, Phys. Lett. 20, 27 (1966). 15. F.T. Arecchi, G. Giacomelli, P.L. Ramazza, S. Residori, Phys. Rev. Lett. 65, 25312534 (1990).

16. F.T. Arecchi, G. Giacomelli, P.L. Ramazza, S. Residori, Phys. Rev. Lett. 67, 3749 (1991).

17. F.T. Arecchi, M. Giglio, A. Sona, Phys. Lett. 25A, 341 (1967). 18. F.T. Arecchi, R. Meucci, F. Salvadori, K. Al Naimee, S. Brugioni, B.K. Goswami, S. Boccaletti, Phil. Trans. R. Soc. A, doi:10.198/rsta, 2104 (2007).

19. F.T. Arecchi, A. Montina, U. Bortolozzo, S. Residori, J.P. Huignard, Phys. Rev. A 76, 033826 (2007). F.T. Arecchi, G.P. Rodari, A. Sona, Phys. Lett. 25A, 59 (1967). T. Bayes, Phil. Trans. Royal Soc. 53, 370-418 (1763). G.J. Chaitin, Algorithmic information theory, (Cambridge University Press, 1987). M. Ciszak, A. Montina, F.T. Arecchi, arXiv, nlin.CD:0709.1108v1 (2007). R.J. Glauber, Phys. Rev. 130, 2529 (1963). R.J. Glauber, Phys. Rev. 131, 2766 (1963). R.J. Glauber, in Quantum Optics and Electronics, Ed. C. DeWitt et al., (Gordon and Breach, New York, 1965). 27. S. Grossberg, The American Scientist 83, 439 (1995). 28. H. Haken, Zeits. Phys. 181, 96-124 (1964), 182; 346-359 (1964). 29. H. Haken, Phys. Rev. Lett. 13, 329 (1964).

20. 21. 22. 23. 24. 25. 26.

Coherence, Complexity and Creativity

30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48.

33

H. Haken, Laser Theory, (Springer, Berlin, 1984). H. Haken, H. Risken, W. Weidlich, Zeits. Phys. 204, 223 (1967); 206, 355 (1967). R. Hanbury Brown, R.Q. Twiss, Nature, 4497, 27 (1956). J.J. Hopfield, Proc. Nat. Aca. Sci., USA 79, 2554 (1982). D.H. Hubel, Eye, Brain and Vision, Scientific American Library, No. 22, (W.H. Freeman, New York, 1995). M. Lax, Phys. Rev. 145, 110-129 (1966). B. Libet, E.W. Wright, B. Feinstein, D.K. Pearl, Brain 102, 193 (1979). M. Mezard, G. Parisi, M.A. Virasoro, Spin glass theory and beyond (World Scientific, Singapore, 1987). E. Rodriguez, N. George, J.P. Lachaux, J. Martinerie, B. Renault, F. Varela, Nature 397, 340-343 (1999). T.A. Sebeok, Semiotica 1341(4), 61-78 (2001). L.P. Shilnikov, Dokl. Akad. Nauk SSSR, 160, 558 (1965). A. Shilnikov, L. Shilnikov, D. Turaev, Int. J. Bif. And Chaos 14, 2143 (2004). H.A. Simon, Cognitive Science 4, 33-46 (1980). W. Singer, E.C.M. Gray, Annu. Rev. Neurosci. 18, 555 (1995). D.R. Solli, C. Ropers, P. Koonath, B Jalali, Nature 450, 1054 (2007). S. Solomon, in Ann. Rev. of Comp. Physics II, (World Scientific,1995), pp. 243-294. G. Toulouse, S. Dehaene, J.P. Changeux, Proc. Nat. Aca. Sci. USA 83, 1695 (1986). F. Varela, E. Thompson, E. Rosch, The Embodied Mind (MIT Press, Cambridge, MA, 1991). K.G. Wilson, Rev. Mod. Phys. 47, 773 (1975).

This page intentionally left blank

EMERGENCE IN ARCHITECTURE

This page intentionally left blank

ENVIRONMENT AND ARCHITECTURE – A PARADIGM SHIFT

VALERIO DI BATTISTA Politecnico di Milano Dipartimento Building Environment Science and Technology – BEST

The interaction of human cultures and the built environment allows a wide range of interpretations and has been studied inside the domain of many disciplines. This paper discusses three interpretations descending from a systemic approach to the question: - architecture as an “emergence” of the settlement system; - place (and space) as an “accumulator” of time and a “flux” of systems; - landscape as one representation/description of the human settlement. Architecture emerges as a new physical conformation or layout, or as a change in a specific site, arising from actions and representations of political, religious, economical or social powers, being shaped at all times by the material culture belonging to a specific time and place in the course of human evolution. Any inhabited space becomes over time a place as well as a landscape, i.e. a representation of the settlement and a relationship between setting and people. Therefore, any place owns a landscape which, in turn, is a system of physical systems; it could be defined as a system of sites that builds up its own structure stemming from the orographical features and the geometry of land surfaces that set out the basic characters of its space. Keywords: Architectural Design, Architecture, Built Environment, Landscape.

1. Introduction A number of studies, both international (Morin, 1977 [19]; Diamond, 1997 [6]), and national (Bocchi and Ceruti, 2004 [1]; La Cecla, 1988, 1993 [14,15]) have recently highlighted a new and wider understanding of human cultures and their interaction with their settlements and the built environment. A part of the Milanese School of Architecture has been interested in these questions for a long time: I would like to recall, among the others, Guido Nardi’s work on dwelling (Nardi, 1986 [21]) and some of our own considerations about the settlement system and the “continuous project” and its double nature – both intentional and unintentional (Di Battista, 1988, 2006 [7,9]). This framework allows a range of interpretations: • architecture as an “emergence” of the settlement system; • place (and space) as an “accumulator” of time and a “flux” of systems; • landscape as one representation/description of the human settlement.

37

38

2.

V. Di Battista

Architecture (be it “high” or not) as an “emergence” of the settlement system

If we define architecture as “the set of human artefacts and signs that establish and denote mankind’s settlement system” (Di Battista, 2006 [10]), we agree that architecture always represents the settlement that generates it, under all circumstances and regardless of any artistic intention. Architecture emerges as a new physical conformation or layout, or as a change in a specific site, arising from actions and representations of political, religious, economical or social powers, being shaped at all times by the material culture belonging to a specific time and place in the course of human evolution. As these actions constantly signal our way of “belonging to a place”, they consequently promote cultures of site and dwelling that denote each dimension of the settlements: from the large scale of the landscape and the city to the small scale of homes and workplaces. These cultures of different settlements involve both human history and everyday life. The “settlement culture” (that is, the culture belonging to a specific settlement) reveals itself by means of its own techniques and artefacts – terracings, buildings, service networks, canals… – and their peculiar features, related to religion, rites, symbols and style. Artefacts and techniques derive from a social and economic environment and highlight psychological and cultural peculiarities of the settled population. Therefore, our artefacts shape and mark places for a long time; moreover, they come from the past continuously reflecting changes occurring in the settlement and in the built environment. All this means that architecture often outlives its generating system, becoming a heritage to the following ones, thus acting as memory – an identity condition linking people and places to their past systems. This peculiarity, signalling both continuity and inertia of the built environment, derives from the many factors that shape the relation between people and places over time. 3. The variable of time and the built environment Whenever we observe a system of objects, the landscape we are facing, this represents both what has been conserved and what has been transformed; it displays geometric shapes, dimension, materials, colors in their relationships and presents a great variety of information about the conditions and means by which every item has been produced and used, in any time. Every description always takes note only of the state of what has been conserved, because the information

Environment and Architecture – A Paradigm Shift

39

about what has been transformed has been irretrievably lost. But even what we perceive as “conservation”, is actually the result of transformation; only a very keen anamnesis and a historical and documental reconstruction can recognise the size and distance in time of past transformation; every backward enquiry puts what we observe to our scientific and cultural models and, paradoxically, the more recent and keener, the more questionable it is. Moreover, no “case history” will ever be able to describe each and every interaction between the built environment we observe today and the settlement system it once belonged to. Every possible assumption about past events is always an interpretation biased by today cultural models and their leading values. This means that memory acquires and processes materials in order to describe a past that always – in different ways – gets to us through our current reality; it, unavoidably, produces a project – be it intentional or unintentional – that regards future. 4. The bonds and values of time Our built environment is the solid outcome of the different lifetimes of all the various things that today represent the system. They represent the “state of reality”, but also refer to the settlements that produced and selected them in time. Therefore, the built environment is the resultant of the many settlements that came one after the other in the same place, the resultant of the un-realized imagination and the enduring conditions; and it is the summation of all the actions – conservation and transformation, addition and subtraction – that have been performed over time in the same place we now observe. It means that today a place is the way it is (be it anyhow) just and because in it a number of things happened, built up and dissolved in a number of times. Every place is the resultant of a huge quantity of things and times: N things N lives N events N times = place N This mound where things and human lives heap together, this summation of times, of every thing and of every human being that ever was in this place, this is what we can read today in our landscapes. This huge amount of past lives we perceive, even if confusedly, may be the reason why we are so spellbound by historical and archaeological finds. Maybe we perceive more keenly our own brief existence, the continuous change of our landscapes, when we face those places where the past and the mound of time become more evident. Actually, today every open space is the background of an ever-changing setting of movable things; this transient scene repeats itself with equivalent components, depriving the background of any meaning. This may be the reason

40

V. Di Battista

why, in our culture, some monuments and place retain acknowledged values and sometimes they become “sacred” in a strange way, being “consumed” by tourism in a sort of due ritual. The hugeness of past, that belongs to every place, cannot be perceived anywhere and anytime; it can be lost when there are not – or there are no more – traces; in these cases, the links between a place and its times wear out in the speed of actions that happen without leaving any mark. 5. Architecture and society No memory can be recalled when every trace of time past has disappeared, but no trace can reach across time if nobody calls it back by inquiry. What is the filter that defines the time of things? No project, no purpose of duration, no painstaking production can guarantee permanence. Only the strict bond between observed system and observing system gives body and meaning to the time of things in a given place. Our built environments, our settlements, are the references – which are in turn observing and observed – of the meanings connecting places and time. Therefore space receives time, it has received it in the past, is sees it flow in the present, it longs for it and it fears it in the future. In the present, the different speeds of change in settlements (for instance, economic values change much faster than social ones) meet the existence cycles of existent buildings; this raises two major issues: • the difference of speed of different settlements in different places of the world; • the virtual availability of all places and landscapes of the earth. This relativization seems to lessen values; indeed, it might offer a new meaning both to “different” conditions and to the material constitution and duration of the places where we live, even the more ordinary ones. In this new relationship with “virtuality” we always find a condition of “dwelling” always claiming a perceptible, physical relationship between us and the space – very variable in character and dimension – that receives our existence, our time, our observations, our decisions, our actions. How do the various existences of places and their specific things meet the occurrences of the human beings that produce, use, conserve, transform or destroy those places?

Environment and Architecture – A Paradigm Shift

41

To understand what happens in our built environments and dwelling places, we could imagine what happens in some of our familiar micro-landscape, such as our bedroom and the things it contains. We could consider the reasons – more or less profound – that organize its space, its fittings, its use, the way we enter it, its outlook and so on. We could also consider the meaning of the different things that characterize that place where we live day after day, discovering and giving way to emotions and rationality, needs and moods, function and symbols: all of these things being more or less inextricable. Now, let’s try and move these reasons and actions and emotions to the wider landscape of social places. Let’s consider the number of subjects acting, of things present in our settlement; let’s multiply the spurs and the hindrances for every subject and every thing. Finally, let’s imagine how many actions (in conservation, transformation, change of use etc.) could affect every single point and every link in the system. If we imagine all this, we will realize that the configuration and the global working of the system is casual, but nevertheless the organization of that space, the meanings of that place – of that built environment – are real. They can exist in reality only as an emergence (a good or bad one, it does not matter) of the settlement system that inhabits that same place. 6. Built environment and landscape Any inhabited space becomes over time a place (a touchstone both for dwelling and identity) as well as a landscape, i.e. a representation of the settlement ad a relationship between setting and people. Therefore, any place owns a landscape which, in turn, is a system of physical systems; it could be defined as a system of sites that builds up its own structure stemming from the orographical features and the geometry of land surfaces that set out the basic characters of its space. It is a multiple space that links every place to all its neighbours and it is characterized by human signs: the agricultural use of land, the regulation of land and water, all the artefact and devices produced by the settled population over time. Thus every place builds up its own landscape, describing its own track record by means of a complex system of diverse signs and meanings. Every landscape displays a dwelling; it changes its dimensions (is can widen up to a whole region, or shrink to a single room) according to the people it hosts and their needs (identity, symbol, intentions of change, use, image…) and their idea of dwelling. This landscape is made of signs that remind of past decisions, projects, actions; it gets its meaning, as a description of material culture, from everything

42

V. Di Battista

that has been done and conceived in it up to our age. And as soon as this space becomes a settlement – and therefore it is observed, described, acted in – it becomes not only a big “accumulator” of permanencies and past energies, but also a big “condenser” of all relations that happen and develop in that place and nowhere else. This local peculiarity of relations depends in part upon geography and climate, in part upon the biological environment (plants, animals), in part upon the characters of local human settlements. In the time t0 of the observation, the place displays the whole range of its current interactions, and that is its identity. Landscapes narrates this identity, that is the whole system of those interactions occurring in the given time: between forms of energy, matter, people, information, behaviors, in that place. Every inhabited place represents, in the time t0 of the observation, the emergence of its settlement system; therefore, as it allows for an infinite number of descriptions, both diachronic and synchronic, it also happens to be – all the time – the “describer” of our settlement system. Every local emergence of a human settlement represents (regarding the time of observation) at the same time both the condition of state t0 of the whole system, and the becoming (in the interval t 0 + t1 * t n ) of its conditions, as the systems within and without it change continuously. Therefore, a place is the dynamic emergence of an open system, the more complex and variable as the more interactive with other systems (social, economic, political…) it is. Observing a place during a (variable) length of time allows us to describe not only its permanence and change – with entities appearing and disappearing – but also its existence flow. This idea – the existence flow of a place, or of a settlement – gives a new meaning to the architectural project in the built environment. 7. The existence flow of a settlement system Every system of relations between human beings and their settlement shapes and gives meaning to its built environment in specific and different ways, according to the different geographic conditions and cultures. We agree that the built environment represents the balance, gained over time, between those environmental conditions and the cultural models belonging to that specific civilization. Some recent studies in environmental anthropology have investigated the feedback from the built environment to the social behavior, and it would be useful to consider the cognitive functions that the “built environment”, in a broader sense, could represent.

Environment and Architecture – A Paradigm Shift

43

Anyway this balance (environment – culture), within the same space, displays a variation in conditions that can be considered as a true flow of elements (and existence) inside the place itself. Resorting to the coincidence “inhabited place/settlement system”, we can describe the space of a place (location and basic physical conditions) as a permanence feature, the unchanging touchstone of all the succeeding systems and their existence cycles. Therefore, could we actually investigate one dynamic settlement system, proceeding in the same place along a continuous existence flow, from its remote foundation to our present time? Or should we analyze by discontinuous methods this same flow as it changes over time and articulates in space? It depends upon the meaning and purpose of our description level. Every place is an evidence of the whole mankind’s history; our history, in turn, changes according to places. The whole flows of events deposits artefacts and signs in places: some of them remain for a long time, some get transformed, some disappear. Generally speaking, natural features such as mountains, hills, plains, rivers, change very slowly, while their anthropic environment (signs, meanings, resources) changes quickly. The duration of artefacts depends upon changes in values (use values, financial values, cultural values etc.), and many settlements may follow one another in the same place over time. It is the presence and the variation of values belonging to the artefacts that establishes their duration over time. On the other side, a built environment crumble to ruins when people flee from it: in this case, it still retains materials and information slowly decaying. Radical changes in the built environment, otherwise, happen when changes in the settlement system establish new needs and requirements. As a settlement changes its structures (social, economic, cultural ones) by imponderable times and causes, so does the built environment – but in a much slower way and it could probably be investigated as an existence flow. In this flow relevant factors can be observed. Permanent and changing elements rely upon different resources and energies, and refer to different social values. Usually, the “useful” elements are conserved; when such elements embody other meanings (such as religious, symbolic, political ones) that are recognized and shared by a large part of the population, their value increases. Sometimes, elements remain because they become irrelevant or their disposal or replacement is too expensive. Duration, therefore, depends upon the complex weaving over time of the needs and the values belonging to a specific settlement

44

V. Di Battista

system, which uses this woven fabric to filter the features (firmitas, utilitas, venustas…) of every artifact and establish its destiny. Landscapes, as system of signs with different lifespan, have two conditions: On one side, in the flowing of events, symbolic values (both religious and civil ones) have a stronger lasting power than use and economic ones, which are more responsive to the changes in the system. On the other side, what we call “a monument” – that is, an important episode in architecture – is the result (often an emergence) of a specific convergence of willpower, resources, many concurrent operators and other favorable conditions. This convergence only enables the construction of great buildings; only commitment and sharing allows an artefact to survive and last over time. It is the same today: only if a system has a strong potential it will be able to achieve and realize major works, only shared values will guarantee long duration of artefacts. 8. Characters of the urban micro-scale The multiple scale of settlement system allows for many different description levels. There are “personal” spaces, belonging to each individual, and the systems where they relate to one another; these spaces include dwellings and interchange places, thus defining an “urban micro-scale” that can be easily investigated. Such a scale displays itself as a compound sign, a self-representation of the current cultural model which is peculiar to every settlement system; it has different specific features (geographical, historical etc.) characteristic of its meanings, roles, identities inside the wider settlement system. The structure of images and patterns (Lynch, 1960, 1981 [16,17]; Hillier and Hanson, 1984 [12]), the urban texture and geometric features, the characters of materials – such as texture, grain and color – their looking fresh or ancient, indicate such things as cultural models, the care devoted to public space by the population and their public image and self-representation. Urban micro-space always represents an open system, a plastic configuration of meanings, where different flows of values, energy, information, resources, needs and performances disclose themselves as variations in the relationship between longlasting and short-lived symbols and signs, which together constitute the landscapes where we all live. Such an open system allows for different levels of description, it requires a recognition, an interpretation of its changes and some partial adjustment and tuning.

Environment and Architecture – A Paradigm Shift

45

9. Inhabited places: use, sign, meanings It would be necessary to investigate the complex interactions that link some features of the cultural model of a population in the place of its settlement (history and material culture, current uses and customs), the way inhabited places are used, the configuration of the ensuing signs (Muratori, 1959 [20]; Caniggia, 1981 [2]), and the general meaning of the place itself. The ways a space is used, the conditions and needs of this use, generate a flow of actions; hence derives to the system a casual configuration which is continuously – though unnoticeably – changing. This continuous change produces the emergence of configurations, systems of signs, which possess a strong internal coherence. Just think of the characteristics of architecture in great cities, corresponding to the periods of great wealth in their history. This emergence of things and built signs, and the mutual relations among one another and their geographic environment is peculiar to all places, but it appeals in different ways to our emotions and actions. The appeal of a place depends upon the different mix of values, linked either to use, information, aesthetics, society etc., that the observer attaches to the place itself; this mix depends upon the observer’s own cultural and cognitive model, as well as his/her needs and conditions (Maturana and Varela, 1980 [18]). In this conceptual framework, every built environment brings forward a set of values which are shared in different ways by the many observing systems inside and outside it. In their turn, such values interfere with the self-regulation of the different flows that run through the environment: flows of activities, personal relationships, interpretations, emotions, personal feelings that continuously interact between humans and things. This generates a circular flow between actions and values, where the agreement connecting different parties is more or less strong and wide and in variable ways regulates and affects flows of meaning as well as of activity and values. 10. Project and design In the open system of the built environment and in the continuous flow of human settlements that inhabit places, there are many reasons, emotions, needs, all of which are constantly operating everywhere in order to transform, preserve, infill, promote or remove things. These intentional actions, every day, change and/or confirm the different levels of our landscape and built environment. This flow records the continuous variation of the complex connexions between people and places. This flow represents and produces the implicit project that all

46

V. Di Battista

built environments carry out to update uses, values, conditions and meaning of their places. This project is implicit because it is self-generated by the random summation of many different and distinct needs and intentions, continuously carried out by undefined and changing subjects. It gets carried through in a totally unpredictable way – as it comes to goals, time, conditions and outcomes. It is this project anyway, by chaotic summations which are nevertheless continuous over time, that transforms and/or preserves all built environments. No single project, either modern or contemporary, has ever been and will ever be so powerful as to direct the physical effects and the meanings brought about by the implicit project. Nevertheless, this very awareness might rouse a new project culture, a more satisfactory public policy, a better ethic in social and economic behaviour. A multiplicity of factors affects this project resulting, in turn, in positive and negative outcomes (Manhattan or the Brazilian favelas?). Which can be the role of intentional projects – and design – in the context of the implicit project, so little known and manageable as it is? How could we connect the relations between time and place deriving from our own interpretation of human settlements, to an implicit project that does not seem to even notice them? Moreover, how could today practice of architectural design – as we know it – cope with such complex and elusive interrelations, in the various scales of space and time? What signs, what meanings do we need today to build more consistent and shared relationships in our built environments? Is it possible to envisage some objective and some design method to improve the values of the built environment? How can we select, in an accurate way, what we should conserve, transform, renew, dispose of? How can we improve something that we so little know of? We need a project of complex processes to organize knowledge and decisions, to find effective answers to many questions, to carry out positive interactions to the flow of actions running through our settlements. 11. Self-regulation, consent, project The issue of consent and shared agreement about the shape and meaning of the human settlement is quite complex and subtle: it deals with power and control, with democracy, with participation. Decisions, choice and agreement cherish each other and correspond to the cultural and consume models of the population. Consent, through the mediation of local sets of rules, turns into customs and icons of the local practices in building and rehabilitation activities. This usually

Environment and Architecture – A Paradigm Shift

47

degenerates into the squalor of suburban housing; but it also makes clear that every human place bears the mark of its own architecture, through a sort of homeostatic self-regulation (Ciribini, 1984, 1992 [3,4]). Such self-regulation modifies meanings by mean of little signs, while upgrading signs by adding new meanings. The continuous variation of these two factors in our environment is the result of a systemic interaction of a collective kind: it could not be otherwise. Will it be possible to improve our capacity in describing such a continuous, minute action we all exert upon every dimension of the built environment? Will it be possible to use this description to modify and consciously steer the flow of actions toward a different behavior? Which role can the single intention/action play and, in particular, what could be the role of the project, as a mock description of alternative intentional actions within the collective unintentional action? How does the cultural mediation operate, between project and commonplace culture? How can it be transferred into the collective culture – that is, into culture’s average, historically shared condition? 12. Research A small settlement can represent, better than a urban portion, a very good place to investigate such complex relations; the good place to understand the role, conditions, chances and limits of the process of architecture – making (Di Battista, 2004 [8]). In a small settlement, the flows and variations of needs, intentions and actions seem more clear; their project comes alive as a collective description that signifies and semantically modifies itself according to values and models which are mainly shared by the whole community. Here the implicit project (Di Battista, 1988 [7]; De Matteis, 1995 [5]), clearly displays itself as a summation of intentions and actions that acknowledges and signifies needs and signs settled over time; in doing this, it preserves some of these signs – giving them new meanings – while discarding others, in a continuous recording of variations that locally occur in the cultural, social, economic system. This continuous updating links the existent (the memory of time past) to the perspective of possible futures. Moreover, it also links the guess upon possible change in real life and in dreams (that is, in hope) with the unavoidable entropy of the physical system. In this sense, the collective project that knows-acts-signifies the complex environment represent its negentropy (Ciribini, 1984, 1992 [3,4]). It would be advisable that such a project could acquire a better ethical consciousness. Thus, inside the collective project, the individual project would became the local

48

V. Di Battista

action confirming and feeding into the whole; or else, it could aim to be the local emergence, finally taking the lead of the collective model. The signs drawn from this continuous, circular osmosis (of models, intentions, predictions, actions, signs and meanings), over and over reorganize, by local an global actions, the existing frame of the built environment. This osmosis never abruptly upsets the prevalence of present signs and meaning: it makes very slow changes, which can be fully perceived over a time span longer than a human lifetime. This self-regulation allows the physical system to survive and change, slowly adjusting it to the continuous change of the cultural and dwelling models; it preserves the places’ identity while updating their meanings. When the implicit project lumps diverging intentions and actions together, the whole meaning becomes confused and hostile. Today, many economical and social relationships tend to “de-spatialize” themselves; many organizations and structures displace and spread their activities, individuals and groups tend to take up unstable relations and may also inhabit multiple contexts. Architecture seems to have met a critical point in shattering one of its main reasons: the capability to represent the relationship between the local physical system and the self-acknowledgement of the social system settled in it. Actually, this de-spatialization is one of the possible identities that individuals, groups and social/economical systems are adopting, and this could be the reason why many places are becoming more and more individualized/socialized. Therefore, the problems of globalization, of social and identity multiplicity, cause such an uncertain and fragmentary forecast that it urges the need and the quest for places that can balance such upsetting; that is why places with memory and identity are so strongly sought for. “Landscape” can be one of the significant centers for this re-balancing action. Landscape is perhaps the most powerful collective and individual representation of the many models we use to describe ourselves – the philosophical and religious as well as the consumerist and productive or the ethical and symbolic ones. It is also the more direct portrayal of many of our desires and fears, both material and ideal. Landscape and architecture are not mere pictures, nor embody only aesthetical and construction capabilities; they are meaningful representations of the time and space emergence of the continuous flow of actions; they self-represent the settlement system in space (Norberg-Schulz, 1963, 1979 [22,23]) and time, and the deepest existential and symbolic relationships of mankind (Heidegger, 1951 [11]; Jung, 1967 [13]):

Environment and Architecture – A Paradigm Shift

49

they are so rich and complex that we still find very hard to describe and even imagine them. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

G. Bocchi and M. Ceruti, Educazione e globalizzazione (Cortina, Milano, 2004). G. Caniggia, Strutture dello spazio antropico (Alinea, Firenze, 1981). G. Ciribini, Tecnologia e progetto (Celid, Torino, 1984). G. Ciribini, Ed., Tecnologie della costruzione (NIS, Roma, 1992). G. De Matteis, Progetto implicito (Franco Angeli, Milano, 1995). J. Diamond, Guns, Germs and Steel. The Fates of Human Societies (1997); (it. ed.: Armi, acciaio e malattie (Einaudi, Torino, 1998)). V. Di Battista, Recuperare, 36, (Peg, Milano, 1988). V. Di Battista, in Teoria Generale dei Sistemi, Sistemica, Emergenza: un’introduzione, G. Minati, (Polimetrica, Monza, 2004), (Prefazione). V. Di Battista, Ambiente Costruito (Alinea, Firenze, 2006) V. Di Battista, in Architettura e approccio sistemico, V. Di Battista, G. Giallocosta, G. Minati, (Polimetrica, Monza, (2006), (Introduzione). M. Heiddeger, Costruire, abitare, pensare (1951), in Saggi e discorsi, Ed. G. Vattimo, (Mursia, Milano, 1976). B. Hillier, J. Hanson, The social logic of space (Cambridge University Press, 1984) C.G. Jung. Man and His Symbols (1967), (it. ed.: L’uomo e i suoi simboli, (Longenesi, Milano, 1980)). F. La Cecla, Perdersi. L’uomo senza ambiente (Laterza, Bari-Roma, 1988). F. La Cecla, Mente locale. Per un’antropologia dell’abitare (Elèuthera, Milano, 1993). K. Lynch, The image of the City, (it. ed.: L’immagine della città (Marsilio, Padova, 1960)). K. Lynch, Good City Form (1981), (it. ed.: Progettare la città: la qualità della forma urbana (Etaslibri, Milano, 1990)). HR. Maturana, F.J. Varala, Autopoiesis and Cognition (1980), (it. ed.: Autopoiesi e cognizione. La realizzazione del vivente (Marsilio, Padova, 1985)). E. Morin, La Methode (1977), (it. ed.: Il metodo (Raffaello Cortina, Milano, 2001)). S. Muratori, Studi per una operante storia urbana di Venezia (Istituto Poligrafico dello Stato, Roma, 1959) G. Nardi, Le nuove radici antiche (Franco Angeli, Milano, 1986). C. Norberg-Schulz, Intentions in Architecture, (1963), (it. ed.: Intenzioni in architettura (Lerici, Milano). C. Norberg-Schulz, Genius Loci (Electa, Milano, 1979).

This page intentionally left blank

EMERGENCE OF ARCHITECTURAL PHENOMENA IN THE HUMAN HABITATION OF SPACE ARNE COLLEN Saybrook Graduate School and Research Center 747 Front Street, San Francisco, CA 94111 USA E-Mail: [email protected]

Considering the impact on human beings and human activities of architectural decisions in the design of space for human habitation, this chapter discusses the increasingly evident and necessary confluence in contemporary times of many disciplines and humanoriented sciences, with architecture being the meeting ground to know emergent phenomena of human habitation. As both a general rubric and a specific phenomenon, architectural emergence is the chosen focus of discussion and other phenomena are related to it. Attention is given to the phenomena of architectural induction, emergence, and convergence as having strategic and explanatory value in understanding tensions between two competing mentalities, the global domineering nature-for-humans attitude, in opposition to the lesser practiced humans-for-nature attitude. Keywords: architecture, convergence, design, emergence, human sciences, induction, systemics, transdisciplinarity.

1. Introduction What brought me to the subject of this chapter is my long-time interest in the occupancy and psychology of space. My approach to the subject is transdisciplinary and systemic, in that I think in contemporary times, we have to converge many fields of study and understand their interrelations to know the subject. What I find particularly interesting and relevant are reciprocal influences between one dynamic body of disciplines associated with architecture, art, design, and engineering the construction of human dwellings on the one side, and another body of disciplines associated with psychological and philosophical thought, human creativity and productivity, and well-being on the other side. Decades of research interest have transpired regarding the reciprocal influences between the two bodies of disciplines, but many would argue that the apparent marriage of architecture and psychology (to take one illustrative connection), through such a lens as environmental psychology [1] applied to architectural designs since the middle of the twentieth century, may have ended 51

52

A. Collen

in divorce by appearances of our human settlements of the early twenty-first century. From my reading of designers, architects and engineers, whose jobs are to design and construct the spaces we inhabit, in recent decades the development of our cities and living spaces constituting them have become subject to the same homogenizing and globalizing forces shaping our consumer products and human services. But for a minority of exceptions, overwhelmingly, the design and construction of human habitats have accompanied the industrialization, the standardization of the processes and products of production, and the blatant exploitation and disregard of the natural order and fabric of the physical world. From our architectural decisions and following them, subsequent actions to organize and construct our living spaces, we have today the accumulation of the physical, psychological, and social effects of them. Our intentions to live, collaborate, and perform in all kinds of human organizations do matter. We are subject to and at the effects of the spaces we occupy. This chapter is to discuss the relevance of trans-disciplinary and systemic approaches that may inform and three architectural phenomena that accompany the dwellings we occupy. 2. Two Attitudes What we do to our surroundings and each other in the forms of architectural decisions have lasting effects. If we believe our surroundings are there only to serve us to fulfill our needs to live, communicate, work, and breed, we have what may be termed the nature-for-humans attitude. Following this mentality, we freely exploit and redesign the natural world to suit ourselves. This attitude is rampant and we see the results everywhere on the planet today. The opposite mentality is the minority view. Adopting this critical interpolation of consciousness, if we believe we are here to serve our surroundings in a sustainable fashion to fulfill our needs, we have the humans-for-nature attitude. It is a pragmatic attitude in which every action takes into conscious regard the consequences of the action on the environment. Unfortunately, only a small proportion of humankind appears to manifest this mentality at this time in human history. We may increasingly question the dominant attitude, such that we may justifiably ask: What are we doing in the design and construction of our habitats to evidence that the humans-for-nature attitude underlies all that we do? Architectural phenomena and decision-making are foci to explore tensions between the two attitudes.

Emergence of Architectural Phenomena in the Human Habitation of Space

53

3. Human Activity Systems and Organized Spaces I have been involved with systems research and sociocybernetics for three decades [2]. I have been particularly interested in what we may call human activity systems [3]. A group of persons forms this kind of system when we may emphasize as the most important defining quality of such a system to be the interactions among these persons. The interactions constitute the activity of the system. The system is not very visible much of the time, but only in our imagination. However, when the people meet in person, or communicate by means of technology for example, the system is activated, it comes alive. It is the communications among the persons that make the system visible. In sum, it is what we mean by a human activity system. It is common that we are members of many human activity systems simultaneously and during our lives. The structures and places associated with human activity systems bring the subject matter of architecture to my research interest, because architecture I believe has a tremendous omnipresent influence on human activity systems. Typically today, we are separated from the natural environments that were common for most of humanity several generations ago. Most of us live our lives in cities. We live and work in contained and well-defined spaces. Considering the longevity of human history, the change from agrarian and nomadic non-city ways of life to the industrialized, consumer-oriented and modernized enclosed spaces of contemporary life has come fast. But an alternative way to think about it is to reflect upon the question: In what ways is the architecture of the life of a human being different today than two hundred years ago? This question is important, in that the architectural decisions of the past, as manifested in the dwellings we inhabit today, I submit have a profound influence on living, thinking, producing, and self-fulfillment. The idea of organized spaces need not be confined to physical housing as we know them. Dwellings, such as schools, offices, and homes, and the physical meeting places within them, such as countertops, dining tables, and workstations, are but nodes of vast and complex networks of persons spanning the globe, made possible via our electronic media technology. Thus, we have various levels of complexity for human activity open to us to consider what organized spaces entail, namely both real and virtual spaces. In fact, such devices as the mobile phone have profoundly altered our idea of what an organized space is. The interface between real and virtual space means that wherever we are in the physical world, there is increasingly present the potentiality of an invasive influential addition (radios, intercoms, cell phones, television and computer screens). These virtual avenues complicate our understanding of our inhabitation

54

A. Collen

of that physical space, because activation of a medium can at any time compete as well as complement our activity in that place. Being paged or phoned may distract or facilitate respectively from current events. The interface has become extremely important to communication, so much so, virtual devices are aspects included in the architectural decisions to design and construct human habitats, for example, placements of recreation and media rooms, and electrical wiring. As a result, various technological media devices are evidence of extensions of human activity systems into virtual realms not immediately visible to us with the physical presence of a group of persons at the same time in the same physical location. 4. Architecture Designs and Organized Space One form of expression of the design of space is architecture. To make a decision that organizes space is an essential element that creates architecture. To impose architecture in space is to organize the space for human habitation. Various organizations of space constitute architectural designs. This activity of ordering space, whether by design of the architect or the inhabitant, can lead to a range of consequences on human activity, from extreme control by others on the one hand to personal expression, happiness, and ornate displays on the other hand [4,5]. Beyond the basics of the perceptual cognitive relations involved in constituting design, the art and innovation in architecture tend to embroider and enhance its minimalism. However, contemporary approaches tend to challenge this view as too limiting, as evidenced for example when inhabitants of modernist architecture remodel and innovate to make their dwellings their own. Such secondary effects illustrate that we cannot take sufficiently into account the emergent consequences of imposing a given architectural design on human beings. Defining architecture, from Vitruvius to present day, and keeping it relevant to human settlements are challenges informatively described in terms of urbanizing concentrations of humanity as complex systems [6]. Further, a provocative journey through the development of architecture revisions the aesthetic of architecture and the primacy of beauty in contemporary terms of the pursuit of happiness that we can experience and manifest in the design and inhabitation of constructed spaces [7]. 5. Architectural Induction and Experiencing Space It is a non-controversial fact, an existential given, that the space a living being inhabits has a profound influence on that living being. Where the biologist may

Emergence of Architectural Phenomena in the Human Habitation of Space

55

point to primary examples of this fact by means of the phototropic and hydrotropic propensities in life forms, the anthropologist may cite the prevalence and placement of certain raw materials, infusing the artifacts of festivals, ceremonies and other cultural events, that are distinguishing markers among peoples. Interacting with the constituent make up of a living being, the environment is a determinant reality of that being. Arranging homes about a meeting place, limiting the heights for every urban dwelling, defining room sizes and their configuration to constitute the set of spaces of a dwelling are examples of architectural decisions. Architecture shapes and organizes the environment for human beings; de facto, architecture is a key environmental force. As a human being, my principal point of reference for existence is my being. To survive, I think in this way and relate to all other persons, things, and places from my personal point of view, my vantage point. Thus, cognition, perception, psychology, and phenomenology are particularly relevant for me to explain, understand, create, design, construct, and change the spaces in which I live, work, and relate with other human beings. At every moment, induction has much to do with my experiencing of the space I inhabit. What sights, sounds, smells, touches and tastes make my space of a place? The objects I perceive and my cognizance of their configuration about me constitute my ongoing experience. My experience is amplified because of my movement through space, which also means through time. My interactions with the objects are specific relations and my space a general relation, all of which are inductions. But those aspects of my experiencing of the space that may be attributed to decisions determining the overall design and organization of the space may be termed architectural induction. By means of perception, cognition and action, we experience space in chiefly four ways: 1) in a fixed body position, we sense what is; 2) we senses what is, while the body is in motion; 3) we interact with persons and objects that are what is; and 4) we integrate senses and actions of what is from multiple separate body positions. This internal frame of experiencing is an artificial articulation of course, because we are doing all four simultaneously most of the time. What becomes experience of a given space is determined in part by the internal frame and in part by the architecture of the space we occupy. The architecture induces and the frame influences. From the resultant confluence, experience emerges.

56

A. Collen

Figure. 1. Framing reconnects separated spaces.

6. Framing and Architectural Phenomena Framing is a natural inherent perceptual-cognitive process of being human (Fig. 1). To line out an area of space is to frame. It is to make separations in the space, to break the space into parts. What is included and excluded in the frame is an act of profound importance having major consequences in regarding architectural induction and emergence. One excellent example of framing in architectural design is making the window. The window is an elementary frame, depicted as a square, rectangle, triangle, circle, oval, or other such intended opening in what is otherwise a pure division of space. Let us consider the square window. What does each square window of a building, seen from a given vantage point communicate? What is its inducement? When a square is made as a window, doorway, recess, or projection, what is induced? Consider some possible relations, not as facts, but only hypotheses: The open square is separation, openness, or possibility; the double square is solidity, stability, or strength; the black-and-white or colored square is separation; the square with crossbars is confinement, imprisonment, or

Emergence of Architectural Phenomena in the Human Habitation of Space

57

control; the square of squares is separateness, security, or safety; and the square in a circle is fluctuation, alternation, tension, or creativity. Consistent with a phenomenology of experiencing space, the examples above are to illustrate the relevance of the experience of the beholder and occupier of the space, regarding the induction of the frame, in this case the square (like the window frame) and the consequent emergent elements of experience.

7. Arena of Inquiry Influences Architecture Inquiry is often discussed in terms of paradigm. We may also realize it is another example of framing. Philosophically, an arena of inquiry (paradigm) comes with an epistemology (knowing), ontology (being), axiology (set of values), and methodology (means of conducting inquiry). We want to know the space. There is knowledge of the place. We can experience the space by being in it and that is not the same as knowing about it. What we see, hear, touch, smell, and taste while in the place naturally spawns meanings, that is, interpretations of what we feel and think about the place. We bring to the place prior experiences that can influence and bias the framing. There are many ways we may value the place or not. And there are ways to explore, alter, and work the place into what we want or need it to be. But there usually are important elements to respect, preserve, and honor in the place. Finally, there are means to redesign and reconstruct its spaces. An arena of inquiry is comprised of the basic assumptions and ideas that define the decisions and practices leading to the architecture. As an arena, it influences the work and process of the inquirer, in this case, the architect who designs, the builder who constructs, and the human beings who occupy the space. When the architect adopts and works within one arena (paradigm), it is a way (frame) of thinking that influences and guides, but also limits thinking. But it is necessary to have to enable the discipline to exist. For the disciplined inquirer, in this case the architect, the frame (paradigm, arena) provides the rules, conceptual relations, principles, and accepted practices to make the architectural decisions required to compose and present the organization of space for human habitation. The paradigm scheme that I find informative is close to one published recently [8]. Paradigms are described to study effects of organized space, and I add a fifth (Systemic) to discuss paradigm for a more inclusive application to architecture. In brief, working within the Functional paradigm, we would be

58

A. Collen

preoccupied with whether the architecture is useful, efficient, and organizes space as intended. Does it work? To design within the Interpretive paradigm, we emphasize how people feel in the space, how they experience it. Is it reflective and enlightening? In the Emancipatory paradigm, we organize space to empower or subdue, liberate or imprison. Does the architecture free or control its occupants? To work in the Postmodern paradigm means to replicate and mimic the diversity and creativity of human beings who are to occupy the space. We would have a major interest in whether the architecture is heuristic and pluralistic, or delimiting and homogenizing. Finally, working within the Systemic paradigm, we would look for ways to combine, balance, configure, and complement the best features of the other paradigms when applied to a particular space. The broadest paradigm would be multi-methodological rather than restricted to one paradigmatic frame. The Systemic paradigm would be most akin to trans-disciplinary architecture, discussed later in this chapter. Given the variety of dwellings we see in our cities today, I find meaningful the following relations between paradigm and the kind of organized space: Functional affiliates with the factory to make a consumer product, Interpretive with the socializing place of a restaurant, Emancipatory with the health spa to promote human healing, Postmodern with the communal park square to support the social diversity of the community, and Systemic with combinations of the above. To illustrate this progression take the application of school architecture. During the industrialization of European and U.S. American continents, our public school systems rose for the populace as places to house our children while parents worked in factories. It is often argued that education then was more about control and socialization than learning and personal development. The design and construction of schools served former ends. Of course, these outdated functionalistic ideas cannot serve our present conditions and needs, even though the idea of containment in a space called school appears of enduring prevalence still. The architecture of schools has advanced extremely to explore the design and construction of more open environments [9,10], in fact to the extreme of considering the community the learning laboratory that once was the classroom. Learning is continuous, life-long, and increasingly augmented by the Internet. Places of learning are confined no longer to metaphors of the one-room schoolhouse, bricks-and-mortar campus, and local geography. To decide the inclusion and placement of a rectangular or oval window in a wall is a prime element and architectural decision. The decision is not divorced from the frames we bring to the act, but to the contrary, partly induced by them. To have familiarity with the arenas of inquiry in advance I contend invites more informed choices and a higher level of awareness to make the architectural

Emergence of Architectural Phenomena in the Human Habitation of Space

59

decisions required to design, construct, and alter human habitats to fulfill the range of human interests represented in the arenas.

8. Architectural Emergence The complexity of framing described in the two previous sections becomes even more profound when we take into consideration that the relations among the elements of the space we perceive changes continuously and multiple paradigms apply. Note the relations enrich and compound experience, for example, when we smell the changing odors walking through a garden (the passage of the body through space), and when sitting we see shadows moving on a wall through the day and feel rising and falling temperatures over days (occupying the same place through time). We are both instruments and recipients of change.

As we move through spaces, the body moves in a constant state of essential incompletion. A determinate point of view necessarily gives way to an indeterminate flow of perspectives. The spectacle of spatial flow is continuously alive . . . It creates an exhilaration, which nourishes the emergence of tentative meanings from the inside. Perception cognition balance the volumetrics of architectural spaces with the understanding of time itself. An ecstatic architecture of the immeasurable emerges. It is precisely at the level of spatial perception that the most architectural meanings come to the fore [11]. As every point of view gives way to the spatial flow of experience, an architecture emerges (Fig. 2). It is inherent in the existent manifest experience of the space occupied. It is a resultant architectural induction. There will likely be an architecture associated with the place one occupies, whether an office, town square, restaurant, or home. But we can also state that the idea of architecture is emergent from the personal experience of the place. That emergent phenomenon from the person is a valid phenomenon. Furthermore, it is justifiably legitimate to name the architecture of one’s experience and communicate it to others. This personal reference point and name of the experience are to be distinguished from the name architecture that is likely associated with the person and design used to construct and organize the space prior to human occupancy. The personal architecture has greatest relevance. From a phenomenological point of view, the totality of organized space experienced personally constitutes the experiential manifestations of consciousness. When lights, sounds, odors, and objects pervade a space, the space, as we experience it, is as much about what is there as what is not. The

60

A. Collen

Figure. 2. Multiple paradigms apply in organizing the spaces of this Parisian indoor emporium for the intended architectural induction to promote emergent behaviors expected in a haven of consumerism.

following are illustrative paired qualities of experience that may become descriptors of our experience of a particular place: Empty-full, present-absent, visible-invisible, loud-quiet, black/white-colored, soft-hard, hot-cold, and strongweak. They represent dimensions of experience, along which we use language to label and communicate experience to others. What is the sight, sound, smell, touch and taste of the space of the place? But descriptors need not be restricted to the sensorial. More complex constructions occupy our experience of space. Are the materials synthetic and artificial, or natural? What and who occupies the space? What interactions among the occupants of the space add to our experience of the place? Our perceptions and cognitions of sounds, lines, shapes, colors, odors and contacts become forces of influence. One may read, reap, interpret, and make meanings--the essential structures and contents of

Emergence of Architectural Phenomena in the Human Habitation of Space

61

consciousness of the place. But of great relevance is the relational nature of the space to our perceptions of the space and meaning attributions that constitute the experience we reflect upon, report, and discuss with others. The particular qualities that describe our experience in the most rudimentary and essential respects are emergent phenomena constituting the experience. They are examples of emergence. Regarding those aspects that stem from decisions determining the overall design and organization of a given space, we may use the phrase architectural emergence to refer to them. The phenomena of induction and emergence are complementary processes, like the two sides of the same coin. They are evocations of our existence in context. Which one to highlight is a matter of emphasis. We may focus on the inductive nature of experiencing space. The impact of the place is described in terms of induction. What flows from the habitat to the occupant, so to speak? What is the induction? Alternatively, we may focus on the emergent qualities of our experience of the place. When in the place, what comes forth to become the foreground of consciousness? What is emergent? Generally speaking, we may refer to the two phenomena as the architectural induction and architectural emergence of the organized space, respectively, when we can know the key architectural decisions involved to design and organize the space associated with the induction and emergence. To illustrate succinctly, placement of a stone arch at the entrance/exit joining two spaces (rooms, courts, passages) has an induction/emergence different from that of a heavy horizontal beam.

9. Systemics of Architecture, Emergence, and Attitude Put people together in a place. Organize the space by means of architecture via the architect, the occupants, or both. After some time, their interactions will likely induce a human activity system. In other words, a social system of some kind emerges, a human activity system defined not simply by the collective beings per se, but more definitively by their interactions. The nature and qualities of the interactions make the system what it is. But it is important to include in our thinking: The architecture of the space is part of the system. It induces to influence human interaction, there by participating in the emergence of properties that come to characterize the system. Given many interactive relations of the people with the environment and each other, concepts and principles applied to describe the designing and organizing of the space for the human beings who occupy it may be termed the systemics of its architecture, that is, those systemic concepts and principles applied to and active in that context.

62

A. Collen

Figure. 3. The office building skyscraper dominates the cityscape.

To illustrate, we may imagine a particular dimension of our experience of place (hot-cold, small-large, still-windy). If we select one element too extremely and focus on it, the whole may go out of balance with the other elements. In other words, a strong force or energy from one element can go so far as to obliterate the presence of others in the space. One element may overshadow the others, like one large tree blocks the sunlight that would nourish the other trees. We witness this spectacle entering a city square or living room of a home to immediately notice a towering building or large stoned floor-to-ceiling fireplace, respectively, with all others entities occupying the space organized around it. The size and intensity of the dominating entity (Fig. 3) tends to command and hold the attention, block out, or mask other entities. Whether the space is being organized in genesis, such as the design, plan, and construction of a new building, or the built space altered, such as remodeling the home, there are architectural decisions being made. The elements that dominant the space, the emergent qualities, may become particular inducements known to and characteristic of that architecture. The kiva (half egg-shaped oven-like fireplace), for example, has acquired this distinguishing status in the homes of Santa Fe, New Mexico. As to the systemic nature of architecture, we may wonder what overriding principle influences our thinking to make the architectural decisions by which the prominent qualities emerge. Is ideal architecture balance? Once we have knowledge of the emergent elements of a given architecture, is the task to find the balance of the most favorable inducements for human habitation? Similarly, we may ask: Is ideal architecture integration of those elements known to promote

Emergence of Architectural Phenomena in the Human Habitation of Space

63

well-being? Of particular relevance is that the emergence of any element to dominate the experience of the occupants of the place may lead further to concerns of human betterment at one extreme and human detriment at the other extreme. Which attitude (nature-for-humans or humans-for-nature) does the hallmark elements of an architecture support? What hallmarks a “green” ecologically sustainable architecture? The thesis developed in this chapter is that the spatial organization we impose through architectural decisions is an inducement in the emergence of the human social systems inhabiting the space. It merits testing to seek evidence for and against it, and whether it might be applied in constructive ways for human betterment. Given current concerns over survivability, it would also support shifts in consciousness from the presently dominant to the advisedly sustainable attitude. Our understanding of this relation seems both obvious and critical to the best of what architecture has to contribute. It should be generally known what inducements favor sustainability, well-being, productivity, and peaceful cohabitation. There is a powerful feedforward loop prominent in the systemics of architecture in its integral relation with design and technology [2]. Civilization progresses by accretion through novelty, diversity, and necessity [12]. We benefit from the discoveries and achievements of those who precede us. Through our immediate activities of design and construction involving feedback loops, we learn what works and what does not. The process is very pragmatic, requiring invention, innovation, and refinement; practical application; and extensive repetition by trial and error until efficacious action becomes reliable and sustainable. Thereby, we come up to the challenge of what is needed to solve the problems of our day. In the case of architecture, the performance, maintenance and endurance of the spaces we design and occupy come under our scrutiny. Ideally, our evaluations should lead over subsequent generations to increasingly superior dwellings in their construction [13], and our healthy living and experience of them [7,14]. As applied to the systemics of architecture, the myriad of feedback loops of human activity systems, coupled with the more macro feedforward loop linking generations are at the heart of second order systemics [15]. It is from the latter that architectures should emerge to apply to the present challenges we face.

10. Emergence of Trans-disciplinary Architecture One implication from the design, organization, and construction of the spaces we inhabit is that the emergent qualities bring preeminent importance to the trans-

64

A. Collen

disciplinary nature of architecture. It follows naturally from the systemics of architecture applied to a given space, because making an architectural decision increasingly has become a more complex endeavor. Some areas to consider are cultural elements; recognition of the unique qualities of indigenous materials; imaginative perspectives; knowing physical, physiological, psychological, social, and economic effects of the architecture on living beings; familiarity with current environmental conditions and fauna; knowing the perceiver’s angle of vision; the history of the place; and preconceptions of the inhabitants. All of these areas have a potential for inclusion in a particular architectural decision. Bringing a set of them together to define in part a given architecture recommends consultation with a range of experts, disciplines, and knowledge domains beyond the principal training and experience of the architect. Thus, to ensure successful completion of a project, the situation commands a systemic approach to organizing the space involved. A confluence of disciplines becomes important to consider and likely necessary, in order to design both conscientiously and consciously with the humans-for-nature attitude. This means a trans-disciplinary approach to making architectural decisions. This chapter has considered architectural phenomena and some aspects of architectural decision-making that would recommend organizing space for human habitation based on systemic and trans-disciplinary approaches. But articulation of the aspects often merely introduces key elements comprising the experience of those who made the original architectural decisions, and later those who occupy the place. From the relations among elements, specifically those that stem from various fields of study and disciplines of human experience and inquiry, we may see trans-disciplinarity emerge. Although matters of economics, physical design, perceptual cognitive relations, and engineering of structure are critical to applications of architecture, there are also psychological, socio-cultural, historical, and contextual influences to be included. For a particular place of human habitation, too much weight given to one aspect may have adverse consequences on the other aspects specifically and the entire space generally. Again, we must question the main principles driving the architectural process, such as balance or integration, mentioned earlier in this chapter.

11. Summary and Conclusion Our experience of space influences our state of being, relationships with others, home and work life, and connectedness to context. The name induction is given to label this phenomenon. Induction is a mediating construct to suggest critical relations between architectures and human activities. The importance of the

Emergence of Architectural Phenomena in the Human Habitation of Space

65

consequence of induction is termed emergence, another phenomenon defined as a quality, feature or characteristic of human interaction with the environment and others associated with and intentionally attributed to its inductive influences. Once the influences are known, their intentional confluence in making architectural decisions is termed convergence. When applied to developing human habitats architectural induction, emergence, and convergence may become advantageous to promoting mutually beneficial humans-for-nature relations. The three architectural phenomena can have strategic and explanatory value to detect and understand the consequences, respectively. The presumption is that our heightened awareness of these phenomena and the framing we apply to decision-making may better enable us to perceive acutely the influences of organized space on our well-being, human relations and activities; evidence the multiple systems of which we are part; and design more efficacious spaces for human beings and human activities. This chapter has been written with systemic and trans-disciplinary importance being given to the imposition of architecture in a place. Sensitivity is imperative to the phenomena of induction, emergence, and convergence. Well worth studying are the architectural decisions having relations to architectural designs and consequential evocations. If we are to become more appreciative of and caring for our environments, and thereby have a quality of life, it is paramount we understand and apply as wisely as possible these relations.

References 1. D. Lowenthal, J. of Environmental Psychology 7, 337 (1987). 2. A. Collen, Systemic Change Through Praxis and Inquiry (Transaction Publishers, New Brunswick, New Jersey, 2004).

3. P. Checkland, Systems Thinking, Systems Practice (Wiley, New York, 1981). 4. L. Fairweather and S. McConville, Prison Architecture (Architectural Press, New York, 2000). C. Day, Spirit and Place (Architectural Press, New York, 2002). V. Di Battista, Towards a systemic approach to architecture, in Ref. 15, p. 391. A. de Botton, The Architecture of Happiness (Pantheon, New York, 2006). M. Mobach, Systems Research and Behavioral Science 24, 69 (2007). M. Dudek, Architecture of Schools: The New Learning Environments (Architectural Press, New York, 2000). 10. A. Ford, Designing the Sustainable School (Images Publishing Group, Victoria, Australia, 2007). 11. S. Holl, Parallax, (Architectural Press, New York, 2000), p. 13. 12. G. Basalla, The Evolution of Technology (Cambridge, New York, 1988).

5. 6. 7. 8. 9.

66

A. Collen

13. A. Stamps, Psychology and the Aesthetics of the Built Environment (Springer, New York, 2000).

14. J. Hendrix, Architecture and Psychoanalysis: Peter Eisenman and Jacques Lacan (Peter Lang, New York, 2006).

15. G. Minati, Towards a second systemics in Systemics of Emergence: Research and Applications Eds. G. Minati, E. Pessa and M. Abram (Springer, New York, 2006), p. 667.

QUESTIONS OF METHOD ON INTEROPERABILITY IN ARCHITECTURE EZIO ARLATI (1), GIORGIO GIALLOCOSTA (2) (1) Building Environment Sciences and Technology, Politecnico di Milano Via Bonardi, 15 - 20133 Milan, Italy E-mail: [email protected] (2) Dipartimento di Progettazione e Costruzione dell’Architettura, Università di Genova Stradone S. Agostino, 37 - 16123 Genoa, Italy E-mail: [email protected] Interoperability in architecture illustrates contemporary instances of innovation. It aims, through the standardization of instruments and procedures (and especially through shared languages of/in IT tools and applications), at the optimization of interactions amongst agents and the work done. It requires, within a consistently non-reductionist systemic approach: (1) interactions and activities of conscious government in/amongst its fundamental component parts (politics, technical aspects, semantics); (2) development of shared languages and protocols, to verify technical, poietic, etc., innovations which do not destroy accumulative effects and peculiarities (axiological, fruitional, etc.). Keywords: systemics, architecture, innovation, sharing, interoperability

1. Introduction “Some might be filled with wonder watching a flock of birds, but such wonder derives from the impossibility of understanding their means of communication: wonder comes from a lack of comprehension, one can not understand because the communication codes are unknown or, if one prefers, because there is a lack of interoperability between their and our language” (Marescotti, in [1, p. 53], author's translation). In a similar way, in architecture, different languages and/or ineffectiveness between codes of communication in the transmission of data, information, etc., and in the operational instructions themselves, lead to interpretative difficulties: the latter often leading, at least, to inefficiencies and diseconomies in technical and management processes. In this way, interoperability in architecture aims at optimizing interactions amongst agents (as well as the work done), using shared common standards for processing/transmitting documents, information, etc.

67

68

E. Arlati and G. Giallocosta

Interoperability, however, if consistently intended in a non-reductionist sense [4, pp. 84-86], and [1, p. 23], should “(...) be developed in three modalities: technical interoperability, of which one can clearly state that although IT techniques and tools (hardware and software) present no problems, problems do exist and lie in the ability to plan adequate cognitive and thus cultural models; semantic interoperability, which takes us back to interdisciplinarity and the construction of dictionaries and thesauri; political interoperability, reflected in law, in the value and various aspects of the law, and in data property and certification. On these questions, standards are studied and developed (...) through the activities of agencies such as ISO (International Organization for Standardization) and OGC (Open Gis Consortium) ...” (Marescotti, in [1, pp. 56-57], author's translation). In the same manner, the sharing of standards and protocols (which constitute the instrumental apparatus for interoperability), when used coherently without detracting from the languages themselves, axiological particularities, etc., ensures: • development of cognitive and cultural models which are acquired (in the sense of improved relative synergies) or also tendencies towards new ones (more adequate and effective interactions); • validation of the former models as an interdisciplinary resource [4, pp. 4951]. Only in this sense can interoperability in architecture find real meaning in an advanced systemic approach: this poses the fundamental questions of method for its use in a non-reductionist key. 2. From the industrialization of building to interoperability in architecture From a rationalist point of view [3, pp. 30-40], the industrialization of building seals a relationship between architecture and industry in the sense of a possible mass production (and in terms of product standardization and interchangeability of products and components). This is the result of those instances of innovation, typical of post-artisan choices and logic, through approaches and operational praxis inspired by mass production (prefabricated parts and industrialization of on-site castings) which, however, carry important implications for the design and management stages. This leads, especially with the use of construction techniques with high levels of prefabrication and/or industrialization of on-site castings, to standardizations which are not always

Questions of Method on Interoperability in Architecture

69

compatible with project poiesis, nor with more consolidated aspects of construction culture or living culture. Over the past few decades, however, new needs and awareness regarding populating processes emerge; often, moreover, such new needs and awareness transform obsolete connotations of progress and development into evident dichotomies: from growth and experience from localized hyper-population, to the containment of urban sprawl, the renovation of the pre-existent, the sustainability of building activities. New assumptions develop regarding the architecture-industry relationship, which are deployed mainly: • according to greater compatibility of the structural peculiarities of the former (indications of the singular nature of buildings, the limited opportunities of product standardization, etc.); • consistently with the most important developments in the latter (series of moderate quantity, analogical series, etc.). With such an evolution of the scenario, the traditional connotations of the industrialization of building are modified. Thus do they appear, less and less characterized by the rigid assumptions of mass production [3, pp. 41-64]. More recent connotations of the industrialization of building, therefore, tend to follow the objectives of the standardization of instruments and procedures, minimizing mass production aspects: and with the latter, any implications of possible offsets in the design stage and in many of the operational and management aspects of architecture. Amongst these objectives, technical interoperability leads to, as mentioned, a need for optimized integration of agents (and their respective activities), through shared languages, currently developed by specifications (called IFC Standards - Industry Foundation Classes) for the production/application of interoperable softwarea. Clearly, the use of the latter: a

The IFC standardized internationally through ISO PAS - Publicly Available Standard 16739/2005, is an open source software, and thus freely available to competent and expert users, and is run by the IAI - International Alliance for Interoperability (an international institution comprising researchers, public sector managers, industrial organizations, academics and university teachers). IAI-IFC develops applications of BIM (Building Information Model) concepts; allows the representation in an explicit, shared and thus interoperable way, of objects and their spatial-functional interrelationships. The BIM model consists of a unified information system whose component parts are explicitly marked in order of represented entities, geometric and typological correlations, assigned characteristics: operation is regulated as a function of tasks and responsibility given to the various subjects holding given capabilities and decisional roles; each subject, while authorized to operate only on their own activities, can visualize the whole set of transactions in progress in the model. In this way the integration of the various decisions can benefit from the simultaneous

70

•

•

•

E. Arlati and G. Giallocosta

represents an optimized scenario of operational integration, supporting programming, design, life-cycle management of building organisms, plant and infrastructure networks, etc.; it becomes an important component of an instrumental apparatus for a systemic approach towards architecture, identifying, moreover, shared integrations, and is coherent with effectively non-reductionist approaches, of its various aspects (functional, technological, energetic, structural, etc.); develops, above all, shared and optimum opportunities, even though only operationally, for the management and control of some of those factors (design, technical-operational, etc.) inducing processes of emergence in architecture [4, pp. 37-42, 98-99).

Thus the dichotomy (which still exists) between the unpredictability of emergence and the need for the prefiguration of architecture can, for some aspects and to some extent, be reduced, also through the use of shared simulation and modeling for the preliminary management and control of possible interactions amongst those factors: design, operational, etc. More generally (and reiterating some previously mentioned aspects) interoperability in architecture, together with other instruments: • overcomes previous connotations of predominant product standardization in building industrialization, endorsing late-industrial assumptions of predominant procedure standardization; • requires/foresees mature connotations of a systemic approach to architecture, replacing traditional structural reductionism. In the same way, in the development of interoperability, the risks explicit in the previous praxis of building industrialization are still present, although in different forms and ways. As mentioned above, in fact, emphasizing product standardization often implies behavior which potentially ratifies operational approaches with the consequent outcomes of building projects; in a similar way standardization of procedures practices, using decision support systems, often wiping out cultural peculiarities through the way they are used, for carrying out and managing architectural activities, in the safeguarding of memory etc., may lead to:

checks regarding potential, undesired conflicts and/or interference, allowing adequate solutions to be found rapidly.

Questions of Method on Interoperability in Architecture

•

•

71

possible removal, with those peculiarities, of the premises regarding multiple languages, axiologies, distinctive traits, etc., of the various architectural contexts; unacceptable breaks in cultural expression and accumulation, and ratification and reductionism at such a level b.

In this sense technical interoperability, when correctly understood (together) validates: • languages and shared operational contributions (and the efficiency of the work done), • cultural premises and peculiarities regarding the many and varied architectural contexts, especially regarding effectiveness and flexibility of the work done (related to project poiesis, axiological acquisitions, model outcomes, etc.), requires the removal of any technicist drift in the set-up and the use of standards, protocols, etc. In the same way, as mentioned above, it also requires conscious government and interactions with modalities (according to Marescotti) of political interoperability and especially semantic interoperability: where one considers, in the validation and the more advanced developments in those cultural premises and peculiarities, roles and contributions ascribable to interdisciplinarity and the construction of dictionaries and thesauri (Marescotti, in [1, pp. 56-57]. Within this framework, interoperability, through consistent interaction between its modalities (technical, semantic and political), as mentioned above, combines advanced definitions from a systemic approach and prepares the ground for a non-reductionist development of architecture. Or rather: when accepted in a rigorously systemic manner, it acts at the operational levels of architecture while conditioning cultural settings and developments (and in this sense one is dealing with the validation of virtuous tendencies).

b

Here, it should be stressed, more generally, that there is a risk of technological determinism, typical of sophisticated IT structures when not suitably controlled especially regarding the manmachine interface: clearly, this risk also tends towards an uncontrolled process of technological proxy.

72

E. Arlati and G. Giallocosta

3. Methodologies of technical interoperability 3.1. Cultural origins of methodological equipment and contextual motivations for evolutionary developments Methodological apparatus for technical interoperability in the processes of design, production and management of architecture have so far essentially been limited by a conservative approach: an approach whose main aim, as for most software companies, was a rapid market success for IT applications. This bears clear witness to the non mature nature of supply for the building industry (especially in Italy), compared to other particularly competitive sectors in global markets (regarding optimization of efficacy/efficiency of the work done) such as electronics, avionics, high-precision engineering. It also bears witness to, and is an expression of, the separate nature and fragmentation of the multifarious markets for building products, which preside over and choose, mistakenly rooted in a given geographical site, the main points of their specificity: and so far have been able to condition the various actors in building activities and the nature of the initiatives, on the basis of effective requirements for the validation of local cultures, acquired values, etc., but also when faced with unmotivated unwillingness to develop procedures for the integration and sharing of technological and operational resources; it also follows from this (amongst the various doubts surrounding a persistent dualism identity/mass production) that the essential existence of systems of representation and processing on IT platforms are difficult to integrate, being aimed more at confining within the boundaries of specific product series the need for cooperation between the various actors involved c. But it is precisely this scenario which constitutes the object of a progressive change, typical of the current state of processes of production and management of architecture and is mainly due to the rise of two decisive factors: • the multiplication of the know-how needed to satisfy the set of requirements of increasing breadth and complexity, stimulated by the need to optimize the use of resources at continually higher levels of quality; • the development of regulatory aspects, aimed at providing guarantees in terms of security and the certainty of quality through the whole of the

c

Nevertheless (as previously mentioned) one may often have, due to other aspects (and as effects of mistaken ideas about innovation transfer and operational approaches towards interdisciplinarity): (1) better transference of procedures and equipment from other sectors towards the building industry, (2) instead of suitable translations in this (and thus coherent with its effective specificities) of practices and outcomes reached elsewhere.

Questions of Method on Interoperability in Architecture

73

building production line, faced with increased attention being paid to economic aspects and social strategies for the creation of the constructed environment (and whose sustainability crisis, as mentioned above, is now apparent). Within this context of rapid transformation, the progressive modification of the building initiative requires the production line of the city and of its buildings to undergo a global re-thinking of its meanings: • in relation to its economic and social points of references, • faced with a system of requisites which are no longer completely part of the traditional knowledge of the architectural disciplines. Thus, there is the need for the predisposition of a renewed thought scenario, before any other condition and capable of representing the wealth of interdependencies and interactions amongst the decisive factors in the design of architecture. In this way emerge the reasons for accepting technical, or technological (the better term given its meaning of cultural renewal) interoperability, and with its procedural and instrumental corollaries, as an environment of operational resources for reaching the objectives of sharing between knowledge and knowhow. The cognitive approach is the fundamental directing criterion for the adoption of the method of technological interoperability; its operational horizon is, in fact, that of a melting pot in which one can: • contaminate with insight the demands from the various universes of reference (traditional disciplines and their decision-making instruments, so far separate), • remodel the nature of the relationships and interactions amongst the factors determining the configuration of the processes and products. In this sense, the contribution of expert knowledge is fundamental, that is the acquired knowledge of those with experience of past effects and consequences of the behavior in operation of building objects faced with given design assumptions. 3.2. Technologies for modeling data-flows Techniques for the virtual representations, expressed (the latter) through reference to an explicit expert knowledge-base and to a declared asset of

74

E. Arlati and G. Giallocosta

requisites and qualitative objectivesd, allow one the render coherent the evaluation of a design model even during its initial conception, through the support of a powerful software simulation environment. Further: the availability of advanced processing and simulation equipment is leading to the progressive loss of the aspects (still significant) of an exclusive and separate nature, allowing one to observe and cooperate in the development of a project down to its most minute details. These aspects thus offer the operators shared opportunities of representations, modeling and checks on the flow of transactions of meanings and values during the development of the design solutions, simulating its concrete realization as well as the later stages of use and management. In this way in architecture, as mentioned, the unpredictability of emergence can, to a certain extent, be directed following positive propensities of the component parts and interactions which produce it: parts and interactions, naturally, which can effectively be managed through forecast probabilistic hypotheses. Thus, in the force field between political, and especially semantic and technical interoperability (Marescotti, in [1, pp. 56-57], the structure of the cognitive instances which circumstantiate the design of architecture can take shape. For this, the powerful instrumental support is still that virtual model which configures the cooperating objects in the formation of the project, identifying the main interactions (semantic, functional, materials, technicalconstructional, maintenance, etc.). The essential intrinsic novelty in any real project, and its natural, innovative content, lies in the order taken on by the set of significant factors in a given context, in a given cultural and productive point in time, and aiming at a specific scenario of present and future requirements: from this point one proceeds, through successive translations of knowledge in their respective appropriate languages and through progressive transformations of identity, to the construction of a virtual model of solutions to the programme of requisites to be satisfied. Experimental control of the outcome of the design stage (traditionally very late in the life-cycle, and extremely expensive in terms of time and work due to comparisons which have to be made with alternative solutions) so far, because of the continuing insufficient development of design support technologies, has limited experimental activities on alternative designs mainly to within the sphere of the expert knowledge of individual designers and executors: d

One of the fundamental aspects, presiding over the introduction of technological interoperability into the architectural project, consists precisely in maintaining the possibility that the motivation/necessity network be rendered explicit, described, feasible and controllable by a community sharing the whole ensemble of aims and interests (or at least significant parts of it).

Questions of Method on Interoperability in Architecture

75

expert knowledge, moreover, predominantly implicit (the tradition lacking a system of representation of design values beyond formal-linguistic, constructive or structural critique), and thus not sufficient (at least in terms of efficiency) when substituting interoperable although fundamental resources, as previously mentioned, to ensure governance. Properly, the binomial governance - technical interoperability, the former intended as an impeding factor (amongst others) of technological determinisme, allows, however, optimum (and efficient) experimental activities. Interoperability offers the various specialist involved in the design the possibility of actually seeing the same identical model through their own specific system of software instruments, in which they have imported, read and processed the original relational database, without the need for re-processing, re-coding, etc., and avoiding approximate interpretationsf. It supports the design stage throughout its development, from its birth, to the definition of the model (in all its spatial and technological components), through the executive and operational definition, to the updating of the data on the edifice as built, down to the management and maintenance stages. It is then possible to proceed with the processing of the component parts (structures, plant, etc.), properly integrated and with optimum control of mutual interferences and overlap, updating the original model of set with the grafting of new components, and also processing families of comparable and superimposable models (thus providing possible design stage alternatives about which choices to make). The project stage thus described generates original potentialities for reaching the prefixed qualitative objectives. It does, in fact, allow the implementation of a praxis of experimenting with design solutions during their actual formulation, and optimizing the acquisition of gained experience by grafting it onto new projects with much greater rapidity with respect to traditional methods.

e f

See note b. The BIM modeling environment assumes the presence of a family of specialist software instruments, conceived for cooperating in the definition of the various sets of characteristic values defining the project; each is an expression of specific domains of expert knowledge deriving from the diverse disciplinary traditions, all of them operating on the basis of the ability to recognize and process the objects defined in the relational database, together sharing: (1) the concept of object-oriented, (2) 3-dimensional vectorial representations in Euclidean space, (3) the qualitative and parametric characteristics defined in the attributes. A further fundamental assumption is the existence of a conventional code for the description, interpretation and recognition of objects in their effective nature (in the sense of represented entities) defined in the relational database. See note a.

76

E. Arlati and G. Giallocosta

The multitude of implemented objects (each with its attributes) in the relational database, moreover, its inherent expert knowledge contributions and especially explicable ad hoc for different project situations (cultural, contextual, poietic, etc.), ensure significant increases in possible variations in options and applications. Naturally, these do not exhaust, however, the number of theoretically possible alternatives. 3.3. Paradigms of experimental application Currently, interoperable technologies of architectural design (based on IAI-IFC standards) require experimental applications in pilot-projects or other, in order to use in the experiment the areas of know-how of the various agents who contribute to making design and actuative decisions, in order to verify the effectiveness of the modelingg. To evaluate this effectiveness, and with it the ability to offer performance advantages to both agents and end-users, the results must be compared with various paradigms. Mainly, the possibilities of: • increasing the levels of government of a project, • proceeding to its most important interfaces, • progressively understanding its main qualitative and economic aspects, • modeling a project, • sub-dividing it into its main phases, • acquiring experience, • facilitating successive monitoring of behavior during life-cycle. The increased possibilities of governing a project, facilitated by the IAI-IFC standard, clearly define prototypes optimized for positive correspondences with end-user requirements and expected quality levels as defined by the regulation: this means especially control of the efficacy of the experimental procedures, on the basis of their achievable advantages (cost-benefit). Feasibility and checking the important interfaces with ranges of further models on the building and microurban scales allow control over the suitability for habitation and services as required by the end-users, and/or the opportunity of implementing innovative aspects. The possibility of progressively understanding, during project modeling, its main qualitative and economic aspects are fundamentally inscribed in the checks retrospectively made with respect to the useful life-cycle of the end-product; in g

Applicative experiences of this type are already under way in the BEST Department at the Politecnico di Milano.

Questions of Method on Interoperability in Architecture

77

this sense the design stages are considered as strategic for logical and temporal priorities, also as a function of the interoperability with successive ones (quality and economies in construction, maintenance, etc.). The effectiveness of modeling a project also concerns the quality and the rapidity of interaction with evaluations of cost and of the technical-operational process (calculations, estimates, etc.). Articulation into its main phases also concerns the identification of building suppliers’ products, to be integrated into the construction, management and maintenance cycles (libraries of interoperable products). The acquisition of experience (implemented, coded and shared within the relational database) is also functional for suitable re-use in future activities. The monitoring of the behavior of manufactured during life-cycle can also be facilitated due to the implementation of a single relational database of the results of the initial simulations, and of those carried out during successive operations. 4. Conclusions It becomes particularly clear, partly reiterating some of the above aspects, how the theoretical-methodological approaches and the outcome of experience of technical interoperability carried out so far suggest, for architecture, the need for and the opportunity of developing its cultural statutes, the sense of its formal and symbolic languages which are still often reduced, confined within selfreferentiality and conserving supposed purifying objectives and presumed intangibility of their relevance. The innovative opportunities offered, and especially the reasons inferable from the current historical context (social, economic, environmental sustainability, etc.), can no longer be inscribed within single disciplinary outlooks, or the breaking up of disciplines, or separated expertise: those opportunities and reasons, on the contrary, require the fundamental integration of disciplines and expertise, optimized osmosis of specialist rigor, expert knowledge, etc., capable of shared syntheses at the highest levels of generality (and not vagueness) and prospectively a harbinger of positive outcomes. In this sense, interoperability in architecture, as seen above, acquires decidedly systemic connotations. In fact, it promotes forms of collective, shared, multidisciplinary and positively desecrating knowledge; it is aware of synergy and does not obliterate, on principle, the specificities and peculiarities of contributions (expert knowledge). On the contrary, it implements the sharing of a cognitive model amongst the agents.

E. Arlati and G. Giallocosta

78

Such an innovative tendency is even more striking, when one considers (and to a much greater extent with respect to other productive sectors): • the traditional (still physiological) heterogeneities of the operators in building activities, • the frequency of conflict amongst them, • the lack of willingness towards common codes of communication amongst various skills, • often the diversification of interests (also due to the existence of persistent disciplinary barriers), • etc. Moreover, the same main objective of past experience in the industrialization of building, shown in the development in (precisely) an industrial sense in the building sector, can now also be pursued, although with different strategies, and at least within the terms of adopting ideally structured (and shared) operational practices, through the contribution of interoperable resources. The latter, however, still require further experimentation and elaboration and especially, as far as is relevant here (and as observed above), regarding developments in a non-reductionist sense. Technical interoperability, in this case, certainly requires further development and improvements, for example in the optimum rearrangement of cognitive and cultural models (Marescotti, in [1, p. 56]), with more flexibility of protocols and instrumentation, etc. But, above all, and of fundamental importance, confirming once again what has been said above, is coherent interaction amongst the technical, political and semantic components of interoperability in architecture, which will render it completely suitable for: • encouraging advanced and shared developments, • favor positive tendencies of components and interactions which lead to emergence, • produce efficient practices of diversity management. And not only on principle. References 1. V. Di Battista, G. Giallocosta, G. Minati, Eds., Architettura e Approccio Sistemico (Polimetrica, Monza, 2006).

2. C.M. Eastman, Building Product Models: Computer Environments Supporting Design and Construction (CRC Press, Boca Raton, Florida, 1999).

3. G. Giallocosta, Riflessioni sull’innovazione (Alinea, Florence, 2004). 4. G. Minati, Teoria Generale dei Sistemi, Sistemica, Emergenza: un’introduzione (Polimetrica, Monza, 2004).

COMPREHENSIVE PLANS FOR A CULTURE-DRIVEN LOCAL DEVELOPMENT: EMERGENCE AS A TOOL FOR UNDERSTANDING SOCIAL IMPACTS OF PROJECTS ON BUILT CULTURAL HERITAGE

STEFANO DELLA TORRE, ANDREA CANZIANI Building Environment Science & Technology Department, Polytechnic of Milan Via Bonardi 3, 20133 Milano, Italy E-mail: [email protected], [email protected]

Cultural Heritage is comprehensible within an integrated vision, involving economic, cultural and ethic values, typical of not renewable resources. It is an open system that doesn’t correspond just to monuments but is made by the complex interactions of a built environment. The systemic relationships between cultural goods (object, building, landscape), and their environmental context have to be considered of the same importance of the systemic relations established with stakeholders/observers. A first partial answer to Cultural Heritage systemic nature has been the creation of “networks” of cultural institutions, that afterwards have been evolving in “cultural systems” and have been recently followed by “cultural districts”. The Cultural District model put forward a precise application for the theory of emergence. But its systemic nature presents also some problematical identifications. For Cultural Heritage the point is not any more limited to “direct” actions. We must consider stakeholders/observers, feedback circuits, emergence of activation of social/cultural/human capital, more than that linked to the architectural design process. a Keywords: local development, relation object - user, Heritage, network of relationships.

1. Cultural Heritage: between nature and history 1.1. Cultural “things” or cultural “heritage” A new vision of the role of Nature and Cultural Heritage might be find out in a well aware transdisciplinary vision of man and of his works within the ecosystems and the environment arose during the second half of the last century. It is enough to remember the 1972 first Club of Rome’s report, The Limits To Growth, or the UNESCO Convention concerning the Protection of the World

a

While the ideas expressed in the present paper derives from common analyses and reflections, the writing of the first part should be attributed to Stefano Della Torre (1.1, 1.2) and the second to Andrea Canziani (1.3, 2,3).

79

80

S. Della Torre and A. Canziani

Cultural and Natural Heritage (1972) and the ICOMOS Amsterdam Declaration (1975). The idea that Cultural Heritage is formed by “all the goods having a reference to civilization history” and that a Cultural Good is every “material evidence having value of civilisation” [1] is affirmed internationally in 1954 by The Hague Convention [2] and in Italy by the statements of the Commission for the Protection and Enhancement of the Historical, Archaeological, Artistic, and Natural Heritage (Commissione di indagine per la tutela e la valorizzazione del patrimonio storico, archeologico, artistico e del paesaggio), commonly known as the Franceschini Commission after its chairman, established in 1964. It is the overtaking of an aesthetic conception, which former laws have been based on, in favour of a wider idea of the cultural value. A value that includes every tangible and intangible evidence and is not limited to aesthetic or historic excellence. It is the expression of “a concept of culture that, against the imbalance caused by fast economic and social modifications [...], assumes a broad anthropological sense and put Nature in touch with History. Therefore artistic and historic heritage is considered as ‘concrete entity of site and landscape, of survival and work’ uninterruptedly placed over the territory and therefore not open to be considered separately from natural environment and in fact coincident at last with Ecology itself” [3, pp. 30-31]. The relevance of giving an open definition and bypassing a mere visual approach is due to the fact that persisting in a preservation of “things of great value” – handled one by one as a consequence of the aesthetic exceptionality that caused their condition of being under preservation laws – means persisting in the vision of Cultural Heritage as belonging to its own separate category, where the value is expressed by a “declaration of interest” that can not and shall not go further the merits of that particular “thing”. A first consequence of this state of things is to separate the goods from their territorial and social context. Just that same context that produced them. A second consequence is to divide a world where cultural value is absolutely prevailing, so that any restoration cost seems to be admissible, from another world where the cultural value is instead absolutely negligible. Third further consequence is the division between the object and its history: the “work of art” is just considered as materialization of a pure artistic value, forgetting its documentary evidence. In this way the preservation based on single value declarations singles out – i.e. divides, isolates – each object to preserve, that is indeed excluded from the evolution of its active context – social and economical –. Just a symbolic and aesthetic value is attributed to goods and they are used as touristy attraction at the most. It is

Comprehensive Plans for a Culture-Driven Local Development: …

81

unnecessary underlining the coherence between this preservation model and the idea of restoration una tantum, or better still, once and for all. And it is unnecessary to stress the distance from systemic visions and sustainable strategies. This terminological change should have been implying a change in general perspective [4]: either for shifting from a passive, static preservation to a proactive one, based on prevention and maintenance, and for a different attention to types and ways of enhancement and management activities. These ones directed to support an open and real common fruition of cultural value within the good. But nevertheless this has just partially happened. 1.2. A new idea of culture: heritage as open system Cultural goods indeed are an open system that does not correspond to monuments. The attention should overtake single elements, considering their inclusive interactions and “recognizing in the general environment the more important informative document” [3, p. 31]. The idea that the attention has not to be given to single goods or to their sum [3, pp. 44 ff.][6, pp.78 ff.] – the idea of the catalogue as taxonomic enumeration that should exhaust the global knowledge of a system – but to their interactions, opens to wider systemic visions. Cultural Heritage is only comprehensible within a global and integrated vision, involving economic, cultural and ethic values typical of not renewable resources. Therefore it does not makes any sense neither the restoration as sum of isolated interventions, nor the separation of protection processes from territorial context [7][15, p.13 ff.]. It is the basis of a conservation defined as coherent, coordinate and planned activity. From this perspective the value does not consist in the tangible object, but in its social function, “seen as intellectual development factor for a community and as historical element which defines the identity of local communities” [4,8]. Nowadays we know beyond doubt that Cultural Heritage is a source of exchange, innovation and creativity [9]. Therefore speaking of enhancement means referring to choices and acts that allow to use the potentialities of a cultural good to create mainly a social advantage. Of course, also if enhancement has to do with possibilities of use/fruition by a larger and larger public, either in a synchronic sense and in a diachronic sense [10], it cannot set aside protection. An object that becomes part of heritage must preserve the values that make of it a historical evidence, a symbol of cultural identification for a community. These values consist in its

82

S. Della Torre and A. Canziani

authenticity that is always tangible, related to its material body. The evidence value consists in the fabric because the memory depends on the signs of passing time on the fabric. Without preserving all the richness of the signs in their authenticity, the evidence itself is missed. And there is no emergence process of the value that originate from the meeting between the object and people histories. The systemic relationships between cultural goods and their environmental context has to be considered of the same importance of the systemic relations established with individuals and society. And perhaps these exercise an even deeper influence. This means that we have to give substance to enhancement not only at level of actions on each object/building/landscape (allowing fruition, minimizing carelessness, giving support services and rights management [11], ...), but also working on territorial context (transport system, accommodation facilities, ...), on environmental quality [12] and social context improvement (comprehension and recognition, involvement of users in care, planned conservation processes, ...). From this viewpoint the integration between different territorial elements is crucial to reach what we might call “environmental quality”. This is one of the main strategic aim for protection of cultural goods, not seen as separate entities whose aging has to be stopped, but as systemic elements whose co-evolutive potentialities we have to care [13,17]. The Cultural Heritage accessibility involves also the issues of mass democratic society and cultural citizenship rights [14][3, p.89, p.255], of social inclusion and cultural democracy [16]. The relationship between use/enhancement and socio-historic values [17, pp.10-13], with its accent on the user/observer, designs an idea of heritage that can have either a direct influence on development as innovation/education and also of course an indirect influence on the economic system. What may look like an inversion between direct and indirect benefits is just seeming: indeed if development is knowledge, economic benefits are the main positive externalities of knowledgeb [18]. What does it mean conserving cultural goods is wonderfully expressed by Andrea Emiliani when he writes: “Non era più possibile immaginare che un dipinto non facesse parte di una certa chiesa e che quella chiesa, a sua volta, non fosse parte integrante di una certa città, di un certo paesaggio, di una certa economia e di una certa società. Non era più possibile trascurare che, per b

“All that is spent during many years in opening the means of higher education to the masses would be well paid for if it called out one more Newton or Darwin, Shakespeare or Beethoven”. (A. Marshall, 1920 [5, IV.VI.26]).

Comprehensive Plans for a Culture-Driven Local Development: …

83

quanti fossero gli interventi di consolidamento e restauro, il risultato complessivo avrebbe aperto una nuova fase di precarietà e di rischio per edifici storici rimessi, si, in sesto ma destinati a subire le conseguenze del sempre più rapido e inarrestabile indebolimento dei relativi contesti territoriali. Chiunque avesse anche lontana pratica di certe faccende, doveva sapere ormai bene che restaurare un dipinto in una chiesa montana, sita in una zona afflitta da spopolamento, non aveva se non un significato interlocutorio o al massimo di mera conservazione fisica: una frana avrebbe poi colpito quella chiesa; il disboscamento avrebbe potuto facilitare quella frana; la fragilità socioeconomica di quel comprensorio avrebbe accelerato il disboscamento...e a cosa sarebbe servito allora aver restaurato quel dipinto, averlo tolto dal suo contesto culturale, averlo -in fondo- sottratto alla sua superstite funzione sociale, aver infine- con la sua assenza stessa- aggravato le condizioni di quella zona? E se a Raffaello è possibile sopravvivere anche nell'atmosfera indubbiamente rarefatta del museo, per il dipinto di minore interesse storico o qualitativo la vita nel contesto suo originario è tutto o quasi. Toglierlo di lì vuol dire operare con leggerezza quel temibile fenomeno di ‘dèracinement’ che è l'attentato più pericoloso che mai si possa organizzare sull'oggetto culturale” [19]. Considering the whole territory as a common good means that its control mechanisms have to deal with the heritage conservation and evolution. That means having to deal with participated management, with scenarios of shared design and confrontation with other disciplines studies, like politics, economy, social or biological sciences. These are the frameworks of governance, of systems like cultural districts, of transdisciplinary approach. 1.3. From integrated cultural systems to cultural districts The first partial answer to Cultural Heritage systemic nature has been the creation of “networks” of cultural institutions – such as museums or libraries – that afterwards have been evolving in “cultural systems”c [20]. The “cultural system” refers to ideas of programming and management rationalization, expressing a deeper generalization aim. The addition of an adjective like “integrated” expresses the awareness of the importance of connections with the territory and of resources diversification. But planned and controlled actions still c

“Molte delle considerazioni che accompagnano le iniziative di “messa a sistema” dei musei invece, come ad esempio quelle inerenti l’applicazione degli standard museali, hanno la tendenza a considerare il sistema una figura organizzativa in grado di supplire alle dimensioni, supposte insufficienti, dei musei locali, consentendogli di assumere comportamenti simili a quelli di maggiori dimensioni.” (Maggi Dondona, (2006), [15, p. 6 ff.]).

84

S. Della Torre and A. Canziani

remains central, as it was a predictable mechanism of networks and transmissions. The integrated cultural systems have been recently followed by “cultural districts”. As Trimarchi said, “Looking carefully the axiom was easy. The Italian industrial production [...] has developed in this complex unit, with a renaissance taste, in which the issues that for the big industry could be defects (the informality, the importance of family relations, the role of decoration, etc.). are instead unreplaceable virtuous elements” [21]. The district model at last seems to be the answer to all those enhancement problems of a Cultural Heritage that too often comes out as the Italian basic economic resource [22]. Moving from the success of many cultural districts within urban environments, it has been developed a set of models for new scenarios of culture-driven development, trying to deal also with situations really different from the urban ones. And then, while the model is still under the analysts’ lens, quite a few prototypes begin to be applied: usually cultural institutions combinations or development initiatives collected under some cultural symbol whose establishing as a system should produce profitable spinoffs, always not quite well defined at the moment but sure and wide in the future [23]. But drawing near districts, production, development and culture is not easy. A cultural district has been defined, from a theoretical standpoint, as the product of two key factors: “the presence of external agglomeration economies and the awareness of the idiosyncratic nature of culture [which is peculiar to a given place or community and to a specific time]. When these two factors join within a dynamic and creative economic environment, the conditions for having a potential cultural district are satisfied. Adding efficient institutions is the political factor that can transform a potential district in a real outcome” [24]. The attention is concentrated mainly in the creation of a value chain and in the main role played by the organization of the so called Marshallian capital: that “industrial atmosphere” with continuous and repeated transactions that cause the information to circulate. There are actually several cultural district connotations [24,25,26,32]. The word reminds an industrial/economic matrix and therefore the idea of incomes generated by the Cultural Heritage or even a culture commercialization. But a more detailed analysis makes clear that such a connotation has been deleted and the expression is used because of its relationship with local community participation, with the answering to government incentives, with the capability of such a system to produce and spread innovative cultural issues and external

Comprehensive Plans for a Culture-Driven Local Development: …

85

economies connected with innovation [23]. The idea of a district stresses the added value of concentration and localization, but also the emergence of these processes. The cultural district idea is linked to an inclusive vision that can re-discuss and understand on the one hand the role of Cultural Heritage within the development economies of a single territory, on the other hand “the deep change in the role of culture within the contemporary society and its present intellectual and emotional metabolisms” [27]. It is possible to recognize in people’s mental space the main infrastructure that has to be the aim of programming and planning. The quality of each action has the possibility to improve the cultural capital, i.e. the local community capability [28]. From the viewpoint of conservation studies, where we investigate cultural mechanisms and practical procedures that form the bases of architectonical heritage preservation, this model is pretty interesting. The systemic links among heritage, territory and society represent the cutting edge of preservation. Moreover the accent that the district model put on users participation is in tune with the most up-to-date visions of government/governance shifting and with conservation as activity made of studies, capability and prevention. 2. Emergencies between cultural districts and architectural heritage The cultural district model put forward a precise application for the theory of emergence. Moving from the systemic nature of Cultural Heritage we observe that preservation and enhancement procedures present complex interactions between elements and the properties of heritage as a whole are not deducible from single objects. We need strategies going further than the search of a local perfection. The same strategies need to be auto-controlled and adaptive [29] to answer to the evolutionary nature of Cultural Heritage. Moreover we have to consider that what is heritage and what is not is absolutely observer-dependent, with a crucial role of the observers’ awareness of the process. Within this framework the Dynamic Usage of Models (DYSAM) [30] is particularly suitable, having to deal with different stakeholders expectations and social systems modellisation [31]. But the systemic nature of a cultural district and of every single object, building or landscape, presents also some problematical identifications. A district is clearly an emerging system because of the nature of interaction and behaviour of its elements, but how interactions can act on a common property? and to which set these elements belong? It is not obvious that a church

86

S. Della Torre and A. Canziani

and a vineyard belong to the same set. A mere geographical link is too weak and the district link would be self-referential. The only real link is due to the sense of belonging to Cultural Heritage, and it requires a conscious acknowledgment act. Since a system is not behaving as a machine, where each component plays a role and can be replaced without the need to act on the others, any change influences the whole system. It is not possible any intervention on the system just acting on elements and the characteristics of a system can be influenced only by more complex interventions on components interactions over time. How do that reflect on the conservation of each single object when we deal with Cultural Heritage? “A building is considered a dynamic system, where technological interventions inspired by the concept of machine and repair are absolutely inadequate, based on reductionist simplifications. The task is to identify the level of description where occur those emergence processes that maintain the materiality of structure and to support them” [30, p. 92]. But what is the communication between a single building and territorial complex? The possibility to work on a single item without prejudice to the system may indeed be a useful property for the system conservation. But for the Cultural Heritage we must take into account the impossibility of support replication and the need for specific processes that require studies and a non-trivial -not standardizedknowledge. Therefore it is evident the point is not any more limited to the “direct” actions on the good, but we have to consider stakeholders –the observers- and feedback circuits. According to Crutchfield [33] there are three possible types of emergence: intuitive definition (something new appear), pattern formation (an observer identifies organization), intrinsic emergence (the system itself capitalizes on patterns that appear). If only the last one is a right and complete specification for systemic properties, few more doubts arise about the applicability to districts. How might we speak of systemic properties not reducible to single elements properties, but emerging from their interactions? Is it possible speak of unpredictable behaviours that lead to the need of a new cognitive model? At first glance economical interactions between i.e. agricultural and construction sectors do not seem to be unpredictable and the same might be for interactions between built objects and life quality. We should more properly speak of “not trivial consequences” of a model. However these consequences are predictable analyzing the model and it leads us to the second emergence type. Architecture is something coming from a design process, the basis is the prediction of use, coming from forecasts and willing. If an emerging behaviour is something that was not in the designer’s aims, then there are no such a

Comprehensive Plans for a Culture-Driven Local Development: …

87

behaviours for architectural objects. In architecture a real emergence is possible only when we recognize into a building some values that have not been foreseen in the design or construction process. That is exactly what usually happen to something recognized as cultural heritage. From this viewpoint it is much more interesting the emergence linked with the activation of social/cultural/human capital than that of architectural design process –that still remains unclear- [34]. 3.

From events to processes: integration of preservation and enhancement process with other territorial structures

The increase of human capital is recognized as a fundamental basis for development and culture can just address growing processes that otherwise might become a threat for heritage. The reference is the basic role of local community participation in the cultural district, the bright answering to government incentives, the ability to produce and spread innovative external economies connected with innovation: that is to say the emerging character of the district. The evolution of culture has always been the capability to interact building a knowledge network. Let us recall here the ideas of “acting/living local and thinking global” [35], at the opposite of predominant projects where the role of intelligence is restricted to immediate and circumscribed problems microscopically thinking- while consumption is global. That is not only a matter of levels of description. It is the case of the peripheral thinking at the basis of Internet: “With the Internet, decisions were made to allow the control and intelligence functions to reside largely with users at the “edges” of the network, rather than in the core of the network itself” [36,37]. Conservation and enhancement processes could act as a catalyst for quality, innovation, creativity, research also in peripheral zone of territory and society. You need to involve users, especially the local community, using a contextual approach that through narration and participation leads to new learning and awareness [18]. That is the recognition or even the construction of the identity connected to Cultural Heritage. Fixed identities must be re-think in the present time, where everyone lives a multiplicity of belongings, less and less linked to territorial definitions. We need to develop the idea of cultural value not as guardian of tradition, but as something emerging from the meeting between heritage elements and internal people’s space [38]. Within this framework if the meeting is not just one, but it is a history where each point of contact is a new deposit of value that is renewed by each event. It is the construction of a

88

S. Della Torre and A. Canziani

dynamic identity, not built just on consolidated values but on global and hybrid relationships. From this standpoint the cultural diversity is seen as necessary for humankind as biodiversity is for nature. In this sense, it is the common heritage of humanity and should be recognized and affirmed for the benefit of present and future generations” [39]. That is the reference for sharing values and for giving the right weight to the cultural performance [40]. Within this frame of reference the built heritage has a basic catalyst role because of its easy recognizable importance, of its use, of its high visibility. But the classical loop – investments, growth, profitability, investment – encounters difficulties when dealing with public goods, characterized by high interconnections between environment complexity and stakeholders. When investments for the Cultural Heritage conservation give rise to new knowledge and education a new loop is established: the heritage is better understood, the identity is enriched or reformulated, there is a new awareness of the importance of taking care, there are the premises for its participated conservation. References 1. “Report of the Franceschini Commission on the Protection and Use of Historical, 2. 3. 4. 5. 6. 7.

8. 9. 10. 11. 12.

Archaeological, Artistic and Natural Heritage”, Rivista trimestrale di diritto pubblico 16, 119-244 (1966). UNESCO, Convention for the Protection of Cultural Property in the Event of Armed Conflict with Regulations for the Execution of the Convention, (The Hague, 14 May 1954). M. Montella, Musei e beni culturali. Verso un modello di governance (Mondadori Electa, Milano, 2003). G. Pitruzzella, Aedon, 1, 2.6 (2000). A. Marshall, Principles of Economics (Macmillan and Co, London, 1920). S. Settis, Italia S.p.A. L’assalto al patrimonio culturale (Einaudi, Torino, 2002). P. Petraroia, “Alle origini della conservazione programmata: gli scritti di Giovanni Urbani”, TeMa, 3, (Milano, 2001). C. Fontana, in: L’intervento sul costruito. Problemi e orientamenti, Ed. E. Ginelli, (Franco Angeli, Milano, 2002), p.15 ff. S. Settis, “Le pietre dell'identità”, Il Sole 24 ore, (13 november, 2005), p. 29. G. Pastori, Aedon, 3, 1.6-8 (2004). UNESCO, Universal Declaration on Cultural Diversity (Paris, 2001). A. Cicerchia, Il bellissimo vecchio. Argomenti per una geografia del patrimonio culturale (Franco Angeli, Milano, 2002). G. Guerzoni, S. Stabile, I diritti dei musei. La valorizzazione dei beni culturali nella prospettiva del rights management (Etas, Milano, 2003). P.A. Valentino, “Strategie innovative per uno sviluppo economico locale fondato sui beni culturali”, in La storia al futuro: beni culturali, specializzazione del territorio e

Comprehensive Plans for a Culture-Driven Local Development: …

13.

14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35.

89

nuova occupazione, Ed. P.A. Valentino, A. Musacchio, F. Perego, (Associazione Civita, Giunti, Firenze, 1999), p. 3 ff. S. Della Torre, in Ripensare alla manutenzione. Ricerche, progettazione, materiali, tecniche per la cura del costruito, Ed. G. Biscontin, G. Driussi, (Venezia, 1999). S. Della Torre, G. Minati, Il Progetto sostenibile, 2, (2004). S. Della Torre, Arkos, 15, (2006). L. Fusco Girard, P. NijKamp, Eds., Energia, bellezza, partecipazione: la sfida della sensibilità – Valutazioni integrate tra conservazione e sviluppo (Franco Angeli Editore, Milano, 2004). M. Maggi, Ed., Museo e cittadinanza. Condividere il patrimonio culturale per promuovere la partecipazione e la formazione civica, Quaderni Ires, 108, (Torino, 2005). M. Maggi, C.A. Dondona, Macchine culturali reti e sistemi nell’organizzazione dei musei (Ires, Torino, 2006). Economia della Cultura, 14(4), (Il Mulino, Bologna, 2004). L. Fusco Girard, Risorse architettoniche e culturali: valutazioni e strategie di conservazione (Franco Angeli Editore, Milano, 1987). D. Schürch, Nomadismo cognitivo (Franco Angeli, Milano, 2006). A. Emiliani, Dal museo al territorio (Alfa Editoriale, Bologna, 1974), pp. 207-208. L. Zanetti, “Sistemi locali e investimenti culturali”, Aedon, 2, (2003). M. Trimarchi, Economia della cultura, 15(2), (Il Mulino, Bologna, 2005), p.137. “L'arte, 'petrolio d'Italia'”, in: Settis, (2002), p.30 ff. P.L. Sacco, S. Pedrini, Il Risparmio, 51(3), (2003). W. Santagata, Economia della cultura 15(2), (Il Mulino, Bologna, 2005), p.141. P.L. Sacco, G. Tavano Blessi, Global & Local Economic Review, 8(1), (Pescara, 2005). P.A. Valentino, Le trame del territorio. Politiche di sviluppo dei sistemi territoriali e distretti culturali (Sperling & Kupfer, Milano, 2003). M. Trimarchi, Economia della cultura, 15(2), (Il Mulino, Bologna, 2005), p.138. A. Sen, Rationality and Freedom (Harvard Belknap Press, 2002). S. Guberman, G. Minati, Dialogue about Systems (Polimetrica, Milano, 2007). G. Minati, Teoria Generale dei Sistemi. Sistemica. Emergenza: un’introduzione, progettare e processi emergenti: frattura o connubio per l’architettura? (Polimetrica, Milano, 2004). G. Becattini, in Il caleidoscopio dello sviluppo locale. Trasformazioni economiche nell'Italia contemporanea, Ed. G. Becattini, M. Bellandi, G. Dei Ottati, F. Sforzi, (Rosenberg & Sellier, Torino, 2001). A. Canziani, Beni culturali e governance: il modello dei distretti culturali, Ph.D. dissertation, (Politecnico di Milano, Milano, 2007). J.P. Crutchfield, in Physica D, special issue on the Proceedings of the Oji International Seminar Complex Systems - from Complex Dynamics to Artificial Reality, 5-9 April 1993, Numazu, Japan, (1994). V. Di Battista, G. Giallocosta, G. Minati, Architettura e approccio sistemico (Polimetrica , Milano, 2006). L. Sartorio, Vivere in nicchia, pensare globale (Bollati Boringhieri, Torino, 2005).

90

S. Della Torre and A. Canziani

36. V. Cerf, U.S. Senate Committee on Commerce, Science, and Transportation Hearing on “Network Neutrality”, (February 7. 2006).

37. F. Carlini, “Io ragiono solo in gruppo”, Il manifesto, 25 luglio 2004. 38. U. Morelli, Ed., Management delle istituzioni dell’arte e della cultura. Formazione,organizzazione e relazioni con le comunità di fruitori (Guerrini, Milano, 2002). 39. UNESCO, Universal Declaration on Cultural Diversity, (Paris,2001). 40. A. Canziani, M. Scaltritti, Il Progetto sostenibile, (2008, in printing).

SYSTEMIC AND ARCHITECTURE: CURRENT THEORETICAL ISSUES GIORGIO GIALLOCOSTA Dipartimento di Progettazione e Costruzione dell’Architettura, Università di Genova Stradone S. Agostino 37, 16123 Genoa, Italy E-mail: [email protected] Systemics approaches towards architecture, traditionally within a structuralist framework (especially within a technological environment), may evolve in a non-reductionist way through: - non-reductive considerations of the role of human requirements in the definition of inhabited spaces; - acceptance of the use-perception dialogical relationship, and more generally of the artscience nexus, as being characteristic of architecture. Likewise, there are theoretical issues in the development of systemic, particularly within the discipline of architecture, including: - the role of the observer, in the constructivist sense and within the exceptions of scientific realism; - the unpredictability of emergence, with its related limits (of purely ontological significance). Keywords: systemics, architecture, emergence, observer.

1. Introduction A great amount of experience with the systemic approach towards architecture distinguishes studies and applications in various disciplinary environments, which operate within that context. Sometimes by accepting the more important developments of systemic, in other cases reiterating the classical concepts of Systems Theory, such experience, however, does not appear to be significantly projected toward the objectives of disciplinary recomposition. In effect, there remains, especially in Italy, the anti-historical counterposition of scientific culture against artistic culture which still characterises all the relationships between the diverse disciplinary aspects in architecture, and any project of an interdisciplinary nature. For example, in Italian Faculties of Architecture, within the different environments operational, professional, etc., there are clear separations between project approaches and project culture (based on requirements, poietic, morphogenetic, etc.).

91

92

G. Giallocosta

The architectural project, on the other hand, when oriented towards allowing the optimum use and management of its many implications (social-technical, economic, perceptive, etc.), requires suitable interdisciplinary elaboration/ applications to be governed, given the mutual interactions and emergent effects, through transdisciplinarity (Minati, 2004 [12, pp. 37-42 and 49-52]). Similarly, the importance of infradisciplinary research should not be underestimated (regarding epistemological foundations and methodologies which are fundamental in obtaining specialistic rigour). The related problems are not only of an operative or applicational nature, but also concern (with these) the need/opportunity to put architecture on trial in its multiplicity of components (compositive, technological, social, economic, etc.) and in the interactions produced. These issues, identified here above all as being of a theoretical nature, can lead to just as many problems regarding architecture and the systemic approach. In the former, it is useful to work with conceptually shared approaches of definitions of architecture, inferring from these, suitable directions/motivations for scenarios of a systemic nature. In the latter, one needs to recognize, within the developments of systemic itself, those problems of major importance for architecture. 2. Possible shared definitions of Architecture Numerous definitions of architecture can be found in recent contributions, and in the historiography of the sector. For example, amongst those ascribable to various Masters (whose citations are given by Di Battista, in Di Battista et al., Eds., 2006 [5, p. 39]): • Architecture “(...) can be seen as the twin of agriculture; since hunger, against which men dedicated themselves to agriculture, is coupled to the need for shelter, from which architecture was born ...” (Milizia, author's translation); • “(...) construire, pour l’architecte, c’est employer les matériaux en raison de leur qualités et de leur nature prope, avec l’idèe préconcue de satisfaire à un besoin par les moyens les plus simplex et les plus solides ...” (Violletle-Duc); • “l’architecture est le jeu savant, correct et magnifique des volumes assemblès sous le soleil (...)”, and also, “(...) the Parthenon is a selected product applied to a standard. Architecture acts upon standards. Standards are a fact of logic, analysis, scrupulous study, and derive from a well-

Systemic and Architecture: Current Theoretical Issues

93

defined problem. Experimentation fixes, in a definitive manner, the standard ...” (Le Corbusier, author's translation). Amongst these, a definition by William Morris in 1881 describes architecture as the moulding and altering to human needs of the very face of the earth itself, except in the outermost desert (Morris, 1947, cit. in Benevolo, 1992 [2, p. 2]). One can agree with Di Battista in considering this definition as a compendium of many of the preceding and later ones; “(...) it takes architecture back to the concept of inhabited environment (...) where human activities have changed, utilised, controlled natural situations to historically establish the built-up environment (...) Architecture could therefore be defined as a set of devices and signs (not ephemeral - author's note) of man which establish and indicate his system of settlement (...) Architecture is applied to a system of settlement as a system of systems: ecosphere (geological, biological, climatic, etc.) and anthroposphere (agricultural/urban, social, economic). Within these systems there are simultaneous actions of: • observed systems (physical, economic, social, convergent with/in settlement system, according to the schematisations of Di Battista - author's note) which present different structures and a multiplicity of exchange interactions; • observing systems as subjects or collectives with multiple identities and values, but also as cognitive models (philosophical, religious, scientific, etc.) which explain and offer multiple intentions and judgement criteria” (Di Battista, in Di Battista et al., 2006 [5, pp. 40-41], author's translation). Morris's definition (Morris, 1947 [13]), especially within the most explicit definition of architecture as a built-up environment (or founding and denoting settlement systems) for satisfying human needs, hence provides unitary and (at least a tendency towards) shared conceptions where, for example, one considers the sphere of human needs in a non-reductive sense of material needs, but also of axiology, representation, poiesis, amongst others. Another definition (most agreeable, and useful here for our declared purposes) is that of Benjamin who, in his most famous essay (Benjamin, 1936 [3]), stresses the particular nature of architecture as a work of art and from which one benefits in two ways, through use and perception: here we also see that dialogical, and highly interactive, relationship between artistic and scientific

94

G. Giallocosta

culturea. The aim of overcoming the art-science dualism was one of the declared objectives of the birth of the Faculties of Architecture, as a disciplinary synthesis of traditional contributions from the Fine Arts Academies (Accademie di Belle Arti) and the Engineering schools. One must, therefore, develop that relationship in a non-reductionist way (Minati, 2004 [12, pp. 84-86], and Minati, in Di Battista et al., 2006 [5, p. 23]), precisely as a synthesis of the twofold fruitional modes of architecture (Benjamin, 1936 [3]) and the effective indivisibility of its diverse aspects: not only representation, or communication, or use, etc., but dynamic interactions involving the multi-dimensional complex of those modifications and alterations produced (and which produce themselves) in the moulding and altering to human needs of the very face of the earth itself (Morris, 1947 [13]). The dialogical recovery of that relationship (art-science), moreover, becomes a necessary requirement, above all when faced with the objectives of optimised management of multiple complex relationships which distinguish contemporary processes in the production of architecture: transformation and conservation, innovation and recollection, sustainability and technological exaltation, etc. In current practices (and reiterating some of what has been said above) there is, however, a dichotomy which can be schematically attributed to: • on one hand, activities which are especially regulatory in the formal intentions of the architectural design, • on the other, the emphasis upon technical virtuosities which stress the role of saviour of modern technology (Di Battista, in Di Battista et al., 2006 [5, p. 38]). Nor can one ignore, particularly in the classical approach of applying systemic to architecture (and emphasised, especially in technological environments), the existence of an essentially structuralist conception, precisely in the sense described by Mounin (Mounin, 1972 [2], cit. in Sassoli, in Giorello, 2006 [8, p. 216])b: in this conception, for example, the leg of a table would be characterized,

a

b

The art-science nexus has been well clarified, moreover, since ancient times. It is referred to, for example, and with its own conceptual specificities, in Plato's Philebus, Vitruvio's De Architectura, the Augustinian interpretation of architecture as a science based upon the laws of geometry, Le Corbusier's Le Modulor, etc. (Ungers, in Centi and Lotti, 1999 [4, pp. 85-93]). When a structure (or for others, a system) can essentially be considered as a construction, in the current sense of the word. In this formulation, analysing a structure means identifying the parts which effectively define the construction being considered (Mounin, 1972 [14], cit. in Sassoli, in Giorello, 2006 [8, p. 216]): and such a definition exists because “(...) a choice is made in the

Systemic and Architecture: Current Theoretical Issues

95

in an analogous manner to the constituent parts of the current concept of building system, “(…) neither by the form nor by the substance, because I could indifferently put an iron or a wooden leg (…) the functions of this part in relation to the structure will remain (in fact - author’s note) invariant (…) LéviStrauss (…) defined the method (…) used in his anthropological research in a way clearly inspired by structural linguistics (defining the phenomenon studied as a relationship between real or virtual terms, build the framework of possible permutations amongst them, consider the latter as a general object of an analysis which only at this level can be made, representing the empirical phenomenon as a possible combination amongst others and whose total system must first of all recognize itself - author’s note) …” (Sassoli, in Giorello, 2006 [8, pp. 216-217], author's translation). In 1964, Lévi-Strauss identified as virtual terms some empirical categories (raw and cooked, fresh and putrid, etc.) showing how, once the possible modes of interaction had been established, such elements can “(...) function as conceptual instruments to make emergent certain abstract notions and concatenate them into propositions ...” (Lévi-Strauss, 1964 [11], cit. in Sassoli, in Giorello, 2006 [8, p. 217], author's translation) which “(...) explain the structures which regulate the production of myths. In that work, Lévi-Strauss tried to show how ‘if even myths (i.e., apparently the most free and capricious human cultural elaboration) obey a given logic, then it will be proven that the whole mental universe of man is subjected [...] to given norms’. But if all the cultural production of man (...) can be traced back to unconscious structures possessing an internal logic independent of the subject, then structuralist ‘philosophy’ will result in a radical anti-humanism. We can interpret in this way the famous statement by Foucault, according to which the structuralist formulation decrees the ‘death of man’ (Foucault, 1966 [7], author's note) ...” (Sassoli, in Giorello, 2006 [8, pp. 217-218], author's translation). It thus becomes clear how the declared radical anti-humanism of the structuralist approach leads to approaches towards systemic which would obliterate (apparently?) one of its most important current paradigms: the role of the observer as an integral part, a generator of processes (Minati, in Di Battista et al., 2006 [5, p. 154]). But the systemic concept, as applied classically to architecture by Italian technological schools, does contemplate an observer in the role of user; the latter, in fact, with its own systems of requisites, functions as a referent for the requisites-performance approach, despite underlining its role attribution in a reductionist sense (as it fundamentally expresses the requisites arrangement of the various parts. And the main criterion for that choice is the function which they have ...” (Mounin, 1972 [14], cit. in Sassoli, in Giorello, 2006 [8, p. 216], author's translation).

96

G. Giallocosta

deriving from the use value of the end-product) and lacking significant correlations with other agents (and carriers of their own interests, culture, etc.)c. More generally, therefore, although definitions of architecture which tend to be shared can trigger mature trial scenarios of a systemic nature (Di Battista et al., 2006 [5]), and within an interdisciplinary, transdisciplinary (and infradisciplinary) framework, unsolved problems, however, still exist. Amongst others: which observer (or better, which system of observers) is best for activities which put on trial architecture in its multiplicity of components and interactions? But this is still an open question even in the most recent developments in systemic. Similarly, there is the problem of the unpredictability of emergent phenomena (especially in the sense of intrinsic emergence)d, when faced with objectives/requisites, typical of the architectural disciplines, of prefiguring new arrangements and scenarios. 3. Specific problems in systemic and implications for Architecture The role of the observer, “(...) an integral part of the process being studied, combines with constructivism (...) This states that reality can not be effectively considered as objective, independently from the observer detecting it, as it is the observer itself which creates, constructs, invents that which is identified as reality (...) Essentially one passes from the strategy of trying to discover what something is really like to how it is best to think of it” (Minati, in Di Battista et al., 2006 [5, p. 21], author's translation). Moreover, the connection between the role of the observer and the Copenhagen interpretation of quantum theory is well-known from 20th century scientific research; amongst others, Heinz Pagels refers to it explicitly (as far as this is of interest here), where one can consider as senseless the objective existence of an electron at a certain point in space independently from its concrete observation: thus reality is, at least in part, created by the observer (Pagels, 1982 [15], cit. in Gribbin, 1998 [9]).

c

d

In a similar way, the structuralist formulation (Mounin, 1972 [14]), at least in Mounin's idea (which, in postulating a choice in the arrangement of the various parts, identifies the criteria through their functions, and the latter will, in some way, have to wait), would seem in any case to assume implicitly the existence of a referent with such a waiting function, even though in a tautologically reductionist sense. Di Battista, moreover, develops a proposal for the evolvement of the classical requisites-performance approach, which integrates use-values with cultural, economic, values, etc. (Di Battista, in Di Battista et al., 2006 [5, pp. 85-90]). Where not only the establishment of a given behavior (even though compatible with the cognitive model adopted) can not be foreseen, but where its establishment gives rise to profound changes in the structure of the system, requiring a new modelling process (Pessa, 1998 [16], cit. in Minati, 2004 [12, p. 40]).

Systemic and Architecture: Current Theoretical Issues

97

Clearly, there is also great reserve regarding the role of the observer in the terms described above. Prigogine for example, although dealing with more general problems (including unstable systems and the notions of irreversibility, probability, etc.), states that: “(...) the need for introducing an ‘observer’ (much more significant in quantum mechanics than in classical mechanics - author's note) necessarily leads to having to tackle some difficulties. Is there an ‘unobserved’ nature different from ‘observed’ nature? (...) Effectively, in the universe we observe equilibrium situations, such as, for example, the famous background radiation at 3°K, evidence of the beginning of the universe. But the idea that this radiation is the result of measurements is absurd: who, in fact could have or should have measured it? There should, therefore, be an intrinsic mechanism in quantum mechanics leading to the statistical aspects observed” (Prigogine, 2003 [17, p. 61], author's translation). Moreover: the role of the observer would lead to the presence “(...) of a subjective element, the main cause of the lack of satisfaction which Einstein had always expressed regarding quantum mechanics” (Prigogine 2003 [17, p. 74], author's translation). Thus, the presence of a subjective element brings with it risks of anthropocentrism. Nor can one avoid similar implications of constructivism with architecture (in which exist, tautologically, all anthropic processes); if the role of the observer, in fact, especially regarding decision-making and managerial activities, entails taking responsibility for safeguarding common interests, it becomes clear from other points of view that dichotomic attitudes can arise from this responsibility, and above all can be theoretically justified through considerations of subjective presences. But such problems regarding the observer in current developments in systemic also allude to the dichotomy between scientific realism and anti-realist approaches, whose developments (especially regarding logical-linguistic aspects of science) are efficaciously discussed by Di Francescoe. Particularly meaningful in that examination (Di Francesco, in Giorello, 2006 [8, pp. 127-137]), who, moreover, explains the position according to second Putnam (converted to antirealism, according to a form of so-called internal realism), Wiggins, and Hackingf, is the strategy suggested by the latter regarding the dual dimension of e

f

Roughly, scientific realism (as opposed to anti-realist approaches) contemplate a reality without conceptual schemes, languages, etc. (Di Francesco, in Giorello, 2006 [8, p. 127]). More precisely: Hacking, 1983 [10]; Putnam, 1981 [18]; Putnam, 1987 [19]; Wiggins, 1980 [20]. Roughly speaking, the internal realism of the according to second Putnam contemplates, for example, how to “(...) ask oneself: of which objects does the world exist? only makes sense within a given theory or description” (Putnam, 1981 [18], cit. in Di Francesco, in Giorello, 2006 [8, p. 133]).

98

G. Giallocosta

scientific activity (representation and intervention) and therefore effectively translatable into architectural processes: reasoning “(...) of scientific realism at the level of theory, control, explanation, predictive success, convergence of theories and so on, means being confined within a world of representations. [...] And thus engineering, and not theorizing, is the best test of scientific realism upon entities. [...] Theoretical entities which do not end up being manipulated are often shown up as stunning errors” (Hacking, 1983 [10], cit. in Di Francesco, in Giorello, 2006 [8, p. 137], author's translation). Naturally, in architecture, there are still critical implications regarding the role of the observer (and in further aspects, beyond those described here). If it could be accepted regarding that interaction between observed systems and observer systems, and as already mentioned (Di Battista, in Di Battista et al., 2006 [5, pp. 40-41]), for the formulation of the latter, the non-reductionist evidence of the multiple (and often different) interests, values, culture, etc., characteristic of the agents in construction projects, is also necessary. Neither is this a question of predominantly computational importance (how to formalise the observer systems); in fact, it also involves, and especially within the framework of managing and governing predominantly social interests/values/etc., of defining and following systemics, organised and self-organise, collectivities (Minati, in Di Battista et al., 2006 [5, p. 21]), avoiding problems: • from shared leadership to unacceptable dirigism, • from self-organisation to spontaneity. Then again: who observes the observer systems? Further more detailed considerations, amongst (and beyond) the problems and hypotheses mentioned so far, are therefore necessary. The role of the observer, moreover, is considered to be of fundamental importance also for the detection of emergent properties (Minati, in Di Battista et al., 2006 [5, pp. 21-22]). But the unpredictability of the latter (as mentioned above) leads to further problems in systemics approaches to architecture, effectively persisting in every outcome of anthropic processes: including those ex-post arrangements of the moulding and altering to human needs of the very face of the earth itself (Morris, 1947 [13]). This unpredictability, however, can to a certain extent be resolved dialogically with respect to the requisites of the prefiguration of scenarios (typical of architectural disciplines): • taking those scenarios to be consistently probabilistic (or as systems of probabilities);

Systemic and Architecture: Current Theoretical Issues

•

99

optimising them through validation of positive propensities of and amongst its component parts (and minimising negative potentialities), also through suitable formalisations and ex-ante simulations.

What is more, it is not unusual in architecture to resort to probabilistic formalizations. On the basis of the Bayes theorem, for example, when there is usually a number of experimental data (b1, b2, ..., bn) which can then be formulated in an appropriate manner, suitable networks can be formalized to support, amongst other things, the evaluation of technical risks (intended, even in building, as being connected to reaching the quality and the performance planned in the project), where, for example, the experimental evidence represents symptoms of the outbreak of a pathology (or hypothesis a)g. In the Bayesian approach, however, there is still the problem of the definition of the a priori probability, even though this can be reduced, according to some, on the basis of its own aspects of subjective variability through suitable inductive principles (besides those usually considered in the calculation of the probabilities), which would lead to corresponding variabilities in the attribution of a posteriori probabilitiesh. Emphasis should therefore be placed, in general, upon those theoretical issues regarding the dichotomy between the unpredictability of emergence and the necessity for ex-ante prefigurations in architecture. The systemic approach in this sense, given the state of current cognitive models of observer systems (Di Battista, in Di Battista et al., 2006 [5, p. 40]), can simply direct it towards an appropriate reduction of that dichotomy. But even this question takes on a purely ontological importance. g

h

See, for instance, the Bayesian approach to risk management in the building industry, see, for example, Giretti and Minnucci, in Argiolas, Ed., 2004, pp. 71-102. As is well known, the Bayes theorem (expressed here in its most simple form) allows calculation of the probability (a posteriori) p(a\b) of a hypothesis a on the basis of its probability (a priori) p(a) and the experimental evidence b: p(a\b) = (p(b\a)p(a))/p(b) The a priori probability “(...) can be interpreted as the degree of credibility which a given individual assigns to a proposition a in the case where no empirical evidence is possessed (...) Whereas p(a\b), which denotes the epistemic probability assigned to a in the light of b, is said to be the relative probability of a with respect to b. In the case where a is a hypothesis and b describes the available experimental evidence, p(a\b) is the a posteriori probability ...” (Festa, in Giorello, 2006 [8, pp. 297-298], author's translation). Besides mentioning (even problematically) some of the inductive principles for minimising subjective variability, Festa recalls the subjectivist conception (de Finetti et al.): according to which, notwithstanding the subjective variability in the choice of the a priori probability, “(...) as the experimental evidence (...) available to the scientists grows, the disagreement (amongst the latter regarding the different evaluations of the a posteriori probabilities - author's note) tends to decrease ...” (Festa, in Giorello, 2006 [8, p. 305], author's translation).

100

G. Giallocosta

References 1. C. Argiolas, Ed., Dalla Risk Analysis al Fault Tolerant Design and Management (LITHOSgrafiche, Cagliari, 2004).

2. L. Benevolo, Storia dell’architettura moderna, 1960 (Laterza, Bari, 1992). 3. W. Benjamin, L’opera d’arte nell’epoca della sua riproducibilità tecnica, 1936 (Einaudi, Turin, 2000).

4. L. Centi, G. Lotti, Eds., Le schegge di Vitruvio (Edicom, Monfalcone, 1999). 5. V. Di Battista, G. Giallocosta, G. Minati, Eds., Architettura e Approccio Sistemico (Polimetrica, Monza, 2006).

6. H. von Foerster, Sistemi che osservano (Astrolabio, Rome, 1987). 7. M. Foucault, Les mots et les choses (Gallimard, Paris, 1966). 8. G. Giorello, Introduzione alla filosofia della scienza (1994) (Bompiani, Milan, 2006).

9. J. Gribbin, Q is for Quantum (Phoenix Press, London, 1998). 10. I. Hacking, Representing and Intervening (Cambridge University Press, Cambridge, 1983).

11. C. Lévi-Strauss, Le cru et le cuit (Plon, Paris, 1964). 12. G. Minati, Teoria Generale dei Sistemi, Sistemica, Emergenza: un’introduzione 13. 14. 15. 16. 17. 18. 19. 20.

(Polimetrica, Monza, 2004). W. Morris, in On Art and Socialism (London, 1947). G. Mounin, Clef pour la linguistique (Seghers, Paris, 1972). H. Pagels, The Cosmic Code (Simon and Schuster, New York, 1982). E. Pessa, in Proceedings of the First Italian Systems Conference, Ed. G. Minati, (Apogeo, Milan, 1998). I. Prigogine, Le leggi del caos, 1993 (Laterza, Bari, 2003). H. Putnam, Reason, Truth and History (Cambridge University Press, Cambridge, 1981). H. Putnam, The Many Faces of Realism (Open Court, La Salle, 1987). D. Wiggins, Sameness and Substance (Blackwell, Oxford, 1980).

PROCESSES OF EMERGENCE IN ECONOMICS AND MANAGEMENT

This page intentionally left blank

MODELING THE 360° INNOVATING FIRM AS A MULTIPLE SYSTEM OR COLLECTIVE BEING VÉRONIQUE BOUCHARD EM LYON, Strategy and Organization Dpt. 23 av. Guy de Collongue, 69132 Ecully Cedex, France Email: [email protected] Confronted with fast changing technologies and markets and with increasing competitive pressures, firms are now required to innovate fast and continuously. In order to do so, several firms superpose an intrapreneurial layer (IL) to their formal organization (FO). The two systems are in complex relations: the IL is embedded in the FO, sharing human, financial and technical components, but strongly diverges from it when it comes to representation, structure, values and behavior of the shared components. Furthermore, the two systems simultaneously cooperate and compete. In the long run, the organizational dynamics usually end to the detriment of the intrapreneurial layer, which remains marginal or regresses after an initial period of boom. The concepts of Multiple Systems and Collective Beings, proposed by Minati and Pessa, can help students of the firm adopt a different viewpoint on this issue. These concepts can help them move away from a rigid, Manichean view of the two systems’ respective functions and roles towards a more fluid and elaborate vision of their relations, allowing for greater flexibility and coherence. Keywords: innovation, organization, intrapreneurship, models, multiple systems, collective beings.

1. Introduction Confronted with fast changing technologies and markets and with increasing competitive pressures, business firms are now required to innovate fast and continuously [1,2,3]. Conventional innovation processes led by R&D and Marketing departments are not sufficient to meet these requirements. In effect, conventional innovation processes tend to be rigid, slow and focused on technology and product development whereas what firms need now is flexible, rapid and broad scope innovation, encompassing all the key elements of their offer, management and organization [4,2,5,3]. Firms have to improve and transform the way they produce, manage client relations, ensure quality, configure their value chain, manage employees, develop competencies, generate revenues, etc. They have to innovate on all fronts and become “360° innovating

103

104

V. Bouchard

firms”. To this end, more nimble innovation processes are required and, above all, innovation must take place in every department and division of the firm. The 360° innovating firm has to rely on the creativity, talent, energy and informal network of its employees. In the 360° innovating firm, employees must be able to autonomously identify opportunities and re-combine the resources and competences that are spread throughout the various departments and sites of the firm to seize these opportunities. Sales and service persons, in frequent contact with clients can identify emerging needs and business opportunities, computer experts can grasp the value creation potential of new IT developments, experts in manufacturing and logistics can propose new solutions to concrete problems, finance experts can help assess costs and benefits, idle machines can be used to produce prototypes, foreign subsidiaries can come up with low-cost solutions, etc. 2. Intrapreneurs and the 360° innovating firm Opportunities and inventions that are identified and developed outside the conventional innovation track cannot succeed without a champion, someone who strongly believes in the project and is personally committed to its success. 360° innovation relies, therefore, on the emergence of internal entrepreneurs or “intrapreneurs” from the pool of employees [6,7,8,9,10]. Internal entrepreneurs or “intrapreneurs” are employees who identify internal or external value creation opportunities and seize the opportunity relying first and foremost on their own talent, motivation and network. Intrapreneurs can take advantage of the financial and technical resources as well as the wide array of expertise and competencies the firm detains. However the life of intrapreneurs is far from easy: they often cumulate the difficulties faced by entrepreneurs (understanding the market, improving the offer, creating a sound economic model, managing a team, making the first sale, etc.) to the difficulties that arise when one pursues an original project within a rigid and risk adverse environment. 3. The intrapreneurial process as a superposed organizational layer In their quest for 360° innovation, a number of firms try to encourage the emergence of intrapreneurs. To do so, they set up structures, systems and procedures whose goal is to encourage, identify, support and select intrapreneurial initiatives [11,12,13].

Modeling the 360° Innovating Firm as a Multiple System or Collective Being

The firm IE

105

The intrapreneurial proces s

Formal processes and established circuits

The formal organization

Slack resources

The environment Figure 1. Two interacting systems, the formal organization (FO) and the intrapreneurial layer (IL).

By doing so, firms de facto superpose a different and potentially conflicting organizational layer (the intrapreneurial process) over the formal organization [14,15,12,11,16,17]. The two layers can be seen as two systems interacting in a complex way (see Figure 1). 3.1. Two different but highly interdependent systems The formal organization (FO) performs well-defined tasks using well-identified procedures, people and resources, while the intrapreneurial layer (IL) assembles people and resources located anywhere in the organization (even outside the organization) on a ad hoc basis, relying extensively on informal networks (see Table 1). The two systems, however, have numerous contact points since most people and resources involved in the IL “belong” to the FO. Most of the time, the intrapreneur herself is a member of the FO, where she continues to fulfill her tasks, at least in the initial phases of her project. The relations between the two systems are complex: 1. The formal organization enables the emergence and unfolding of the intrapreneurial process by 1) granting autonomy to the internal entrepreneur, 2) providing most of the resources he uses, and 3) giving legitimacy to his

106

V. Bouchard Table 1. Two very different systems.

The formal organization (FO)

The intrapreneurial layer (IL)

Well defined set of elements and interactions

Fuzzy, constantly evolving set of elements

Relatively stable over time

Temporary

Planned (top down)

Emergent (bottom up)

A priori resources and legitimacy

Resources and legitimacy are acquired on the way

2. 3. 4.

project. In other words, system IL is embedded in system FO on which it depends for its survival and success. However system FO is also dependent on system IL. In effect, the intrapreneurial layer allows the formal organization to 1) overcome some of its structural limitations and 2) reach its objective of fast 360° innovation. The intrapreneurial layer is often competing for resource and visibility with some parts of the formal organization and often enters in conflict with it. (IL competes with a subsystem of FO). Finally, the intrapreneur and more generally all those who contribute significantly to the IL, can be rejected or envied by formal organization members because their values, work styles, status are different. (The culture – norms, values and behaviors – of system IL and that of system FO are conflicting).

3.2. Managing the intrapreneurial process The single intrapreneurial initiative is managed – primarily – by the intrapreneur himself. However the intrapreneurial process as an organizational dynamic, a sustained flow of intrapreneurial initiatives, has to be managed by the top management of the firm. Let us review what goals these actors pursue, the levers they control and some of their main strategic options. 3.2.1. Top management Top managers pursue several objectives. Among them: • Multiply the number of intrapreneurial initiatives; • Improve their success rate; • Contain the risks and costs; • Leave the formal organization (FO) “undisturbed”; • Provide concrete examples of the desired behavior to members of the formal organization.

Modeling the 360° Innovating Firm as a Multiple System or Collective Being

107

Some of their most significant control variables are: • The level and type of support granted to internal entrepreneurs; • The conditions at which support is granted; • The definition of desired/ undesired, licit/illicit intrapreneurial initiatives; • The creation of formal links between the two layers. Their strategic options can be positioned along two continua: • Granting high autonomy to employees vs. granting moderate autonomy to employees • Providing strong formal support to intrapreneurs vs. providing minimal formal support to intrapreneurs. • Relying essentially on informal links between the IL and the FO or relying on both informal and formal linkages. 3.2.2. Intrapreneurs Internal entrepreneurs generally seek to maximize their chances of success by: • Securing access to needed resources and competencies; • Minimizing conflict with the formal organization; • Getting the support of members of the leading coalition. Some of their most significant control variables are: • The level of strategic alignment of their project; • Their level of self-sufficiency/autonomy vis-à-vis the formal organization (FO); • The degree of visibility of their project. Here again their strategic options can be positioned along various continua: • Pursuing a strategically aligned project vs. pursuing a not so strategically aligned project; • Being highly self sufficient vs. trying to get formal help and support early on; • Keeping the visibility of the project low vs. giving the project high visibility. 4. A recurrent and bothering problem There are numerous empirical evidences that, over time, systems dynamics play strongly against the Intrapreneurial Layer, which remains marginal or shrinks after an initial period of boom [12,18,11,16,13,17].

108

V. Bouchard

In spite of the declarations and measures taken by the top management to encourage intrapreneurial initiatives, many intrapreneurs face so many difficulties that they renounce to their project. Some fail for reasons that can be attributed to the weakness of their project or their lack of skills but many fail because of the insurmountable organizational or political obstacles they face. And without a small but growing number of visible successes, the intrapreneurial dynamic soon comes to a halt. Some recurrent problems faced by intrapreneurs: • Parts of the formal organization actively or passively oppose the intrapreneur (including the boss of the intrapreneur); • Excessive work load, no team, no help; • The intrapreneur cannot obtain the needed financial resources; • The intrapreneur cannot secure solid and lasting top management support; • The intrapreneur is isolated and does not benefit from the advice of mentors or fellow intrapreneurs; • The intrapreneur is not able to simultaneously face external (market) and internal (political) challenges. A critical issue for firms interested in promoting 360° innovation, therefore, is to realize that such a negative dynamic is at play and find ways to counteract it. If we understand better the complex interactions between the two systems (FO and IL) and their main agents (top management, intrapreneurs, other managers), we might be able to find ways to reduce the pressures experienced by intrapreneurs thus favoring innovation and the creative re-deployment of resources within the firm. New concepts in system modeling such as multiple systems (MS) and collective beings (CB) could help us in this endeavor. 5.

New concepts in system modeling: multiple systems (MS) and collective beings (CB)

We propose to try and apply to the FO-IL systems dynamics depicted above the concepts of Multiple Systems (MS) and Collective Beings (CB) developed by Minati and Pessa [20]: • A MS is a set of systems established by the same elements interacting in different ways i.e., having multiple simultaneous or dynamic roles. Examples of MS include networked interacting computer systems performing cooperative tasks, as well as the Internet, where different systems play different roles in continuously new, emerging usages.

Modeling the 360° Innovating Firm as a Multiple System or Collective Being

•

109

A CB is a particular MS, established by agents possessing the same cognitive (natural or artificial) system. Passengers on a bus and queues are examples of CB established dynamically by agents without considering multiple belonging. Workplaces, families and consumers are examples of CB established by agents simultaneously and considering their multiple belonging.

These new concepts can help us reframe the challenges faced by the “360° innovating firm” which could be approached as a problem of increasing the degrees of freedom of various systems simultaneously involved in innovation i.e., increasing the number of representations simultaneous available to the various agents. For instance, we may design the Intrapreneurial Layer not only in opposition to the Formal Organization, but also considering the possibility of: • redefining problems by distinguishing between conventionally established differences and intentionally established differences between the two systems, for the purpose of systems modeling; • distinguishing between subsystems and systems of the multiple system; • not only establishing a distinction between functional relations and emergent relations but also mixing and managing the two. The proposed approach can help us move away from a rigid, Manichean view of the systems’ respective functionalities and roles towards a more fluid and elaborate vision of their relations, allowing for greater flexibility and coherence when tackling the organizational and managerial issues facing the 360° innovating firm. Let us illustrate these new concepts by applying them to the system “firm” in its productive function. Aspects such as production, organization, cost effectiveness, reliability and availability can be viewed: • as different properties of the firm viewed as a single system or as a set of subsystems, or • as different elements of the MS “firm”, constituted by different systems established by the same elements interacting in different ways. In the second eventuality, production will be considered as an autonomous system possessing its own independent representation and dynamics and not only a property of the system “firm”, itself dependant on organization. In the same way, quality is an autonomous system and not only an effect of production, and so on. The different dimensions are not only viewed as functionally related aspects of the system or of different subsystems, but also as different

110

V. Bouchard

combinations of the same elements (e.g., human resources, machines, energy, rules and facilities) forming different systems (e.g., production, organization and quality). What difference does it make ? In this case, we may act on a given system of the MS not only in a functional way, but also via the complex web of interactions that emerge from its elements’ multiple belonging. From a functional standpoint, the increasing of production may reduce quality and cost-effectiveness affect organization. In an MS perspective, quality is not an effect of production, but an autonomous property of elements also involved in production. Quality, in this case, will derive from design rather than production procedures. It becomes possible to consider properties laterally rather than functionally. Properties as quality, reliability and cost effectiveness are not properties of a single system, but properties of the whole. In the same way, human resources will be considered as agents able to pursue multiple roles in producing, organizing, communicating, marketing, developing new ideas, controlling quality and so on. In the “360° innovating firm”, no agent has a single specific role but rather multiple, dynamic, contextdependent roles. 6. Applicability of DYSAM The Dynamic Usage of Models (DYSAM) has been introduced in Minati and Brahms [19] and Minati and Pessa [20] to deal with dynamic entities such a MS and CB. The dynamic aspect of DYSAM relates to the dynamic multiple belonging of components rather than to the dynamic aspect related to change over time. DYSAM is based on simultaneously or dynamically model a MS or CB by using different non-equivalent models depending on the context. For instance, a workplace may be modeled in a functional way by considering the processing of input and the production of output; as a sociological unit by only considering interactions among human agents; as a source of waste, pollution and energy consumption; and as a source of information used for innovation. The concept of DYSAM applies when considering a property in the different systems of a MS or CB. Moreover, in this case, the models must take into account the fact that different systems are composed of same elements. In this way, dealing with quality in a system affects the other aspects not only in a functional way, but also because the same elements are involved in both. Depending on effectiveness, a firm may be modeled as a system of subsystems and/or as an MS or CB. For instance, the profitability of a firm cannot be modeled by using a single optimization function, linear composition of single different optimization functions, but rather by using a population (i.e., a system)

Modeling the 360° Innovating Firm as a Multiple System or Collective Being

111

of optimization functions continuously and dynamically established by considering context-sensitive parameters. DYSAM allows considering different, non-equivalent models, such as the ones related to profitability, reliability, availability, flexibility and innovation, as autonomous systems of a MS. 7. Conclusion Firms are nowadays required 1) to maximize return on assets, which implies strict performance control and efficient use of resources and 2) to innovate on all fronts (360° innovation), which implies local autonomy, trials and errors and patient money. In order to face these simultaneous and apparently contradictory requirements, several firms superpose an intrapreneurial layer to their formal organization. While the formal organization (FO) performs well-defined tasks using well-identified procedures, people and resources, the intrapreneurial layer (IL) assembles people and resources located anywhere in the organization (even outside the organization) on a ad hoc basis, relying extensively on informal networks to develop innovative projects. The two systems are in complex relations: if the IL is, to a large extent, embedded in the FO, sharing its human, financial and technical components, it also strongly diverges from it when it comes to representation, structure, values and behavior of some shared components. Furthermore, the two systems simultaneously cooperate and compete and frequently enter in conflict. In the long run, one observes that the organizational dynamic set forth usually ends to the detriment of intrapreneurial processes, which remain marginal or regress after an initial period of boom. The concepts of Multiple Systems and Collective Beings, proposed by Minati and Pessa, can help students of the firm adopt another viewpoint on the issues just described and tackle them differently. These concepts can help them move away from a rigid, Manichean view of the two systems’ respective functionalities and roles towards a more fluid and elaborate vision of their relations, allowing for greater flexibility and coherence when tackling the organizational and managerial issues facing the 360° innovating firm. The application of these concepts together with the related DYSAM techniques, could help students of the firm come to term with the multiple contradictions that arise from the mandatory adoption of multiple, non additive roles by the managers of 360° innovating firms. Acknowledgments I wish to express my gratitude to Professor Gianfranco Minati for his help and feedback on the paper.

112

V. Bouchard

References 1. P. Drucker, Innovation and Entrepreneurship (Harper Business, 1993). 2. G. Hamel, Harvard Business Review 77(5), 70-85 (1999). 3. J.P. Andrew, H.L. Sirkin, and J. Butman, Payback: Reaping the Rewards of Innovation (Harvard Business School Press, Cambridge, 2007).

4. P.S. Adler, A. Mandelbaum et al., Harvard Business Review, March-April, 134-152 (1996).

5. R.M. Kanter, Executive Excellence 17(8); 10-11 (2000). 6. G. Pinchot III, Intrapreneuring: why you don’t have to leave the corporation to become an entrepreneur (Harper and Row, New York, 1985).

7. R.A. Burgelman, Administrative Science Quarterly 28(2), 223-244 (1983). 8. D. Dougherty, C. Hardy, Academy of Management Journal 39(5), 1120-1153 (1996).

9. A.L. Frohman, Organizational Dynamics 25(3), 39-53 (1997). 10. P.G. Greene, C.G. Brush and M.M. Hart, Entrepreneurship Theory and Practice 23(3), 103-122 (1999).

11. Z. Block, I.C. Macmillan, Corporate venturing : creating new businesses within the firm (Harvard Business School Press, Boston, 1993).

12. R.M. Kanter, J. North et al., Journal of Business Venturing 5(6), 415-430 (1990). 13. V. Bouchard, Cahiers de la recherche EM LYON, N. 2002-08 (2002). 14. N. Fast, The rise and fall of corporate new venture divisions (UMI Research Press, Ann Arbor, 1978).

15. R.A. Burgelman, L.R. Sayles, Inside corporate innovation: strategy, structure and managerial skills (Free Press, New York, 1986).

16. P. Gompers, J. Lerner, in R.K. Morck, Ed., Concentrated Corporate Ownership (University of Chicago Press, Chicago, 2000).

17. V. Bouchard, Cahiers de la recherche EM LYON, N. 2001-12 (2001). 18. R.M. Kanter, L. Richardson, J. North and E. Morgan, Journal of Business Venturing 6(1), 63-82 (1991).

19. G. Minati, S. Brahms, in: Emergence in Complex, Cognitive, Social and Biological Systems, G. Minati and E. Pessa, Eds., (Kluwer, New York, 2002), pp. 41-52.

20. G. Minati, E. Pessa, Collective Beings (Springer, New York, 2006).

THE COD MODEL: SIMULATING WORKGROUP PERFORMANCE

LUCIO BIGGIERO (1), ENRICO SEVI (2) (1) University of L’Aquila, Piazza del Santuario 19, Roio Poggio, 67040, Italy, E-mail: [email protected], [email protected] (2) LIUC University of Castellanza and University of L’Aquila, Piazza del Santuario 19, Roio Poggio, 67040, Italy E-mail: [email protected]

Though the question of the determinants of workgroup performance is one of the most central in organization science, precise theoretical frameworks and formal demonstrations are still missing. In order to fill in this gap the COD agent-based simulation model is here presented and used to study the effects of task interdependence and bounded rationality on workgroup performance. The first relevant finding is an algorithmic demonstration of the ordering of interdependencies in terms of complexity, showing that the parallel mode is the most simplex, followed by the sequential and then by the reciprocal. This result is far from being new in organization science, but what is remarkable is that now it has the strength of an algorithmic demonstration instead of being based on the authoritativeness of some scholar or on some episodic empirical finding. The second important result is that the progressive introduction of realistic limits to agents’ rationality dramatically reduces workgroup performance and addresses to a rather interesting result: when agents’ rationality is severely bounded simple norms work better than complex norms. The third main finding is that when the complexity of interdependence is high, then the appropriate coordination mechanism is agents’ direct and active collaboration, which means teamwork. Keywords: agent-based models, bounded rationality, law of requisite variety, task interdependence, workgroup performance.

1. Introduction By means of the COD (Computational Organization Design) simulation model, our main goal is to study the effects of the fundamental modes of connection and bounded rationality on workgroup performance. Therefore, we are at a very micro-level of analysis of a theory of interdependence and coordination. Technological interdependence is one of five types of interdependence,1 the others being the behavioral, informational, economic, and juridical. Technological interdependence coincides with task (or component) interdependence, when it is referred at the micro-level of small sets of technologically separable elementary activities. Task interdependence is

113

114

L. Biggiero and E. Sevi

Figure 1. Modes of connection.

determined by several factors, which occur at network, dyad, and node levels.1 One of the most important factors is just the mode of connection, that is the parallel, sequential or reciprocal ways in which tasks and/or agents interactions can take place. Two (or more) tasks can be connected by means of one (or more) of these three modes of connection (Fig. 1): (1) parallel connection, when tasks are connected only through its inputs and/or outputs; (2) sequential connection, when the output of one task is the input of the following; (3) reciprocal connection, when the output of a task is the input of the other and vice versa. This categorization coincides with those that, in various forms and languages, has been proposed by systems science, calling them systemic coupling [2,3]. It’s noteworthy to remind that they exhaust any type of coupling and that, as underlined by cybernetics, only the reciprocal mode refers to cybernetic systems, because only in that case there is a feedback. Indeed, into the systems science the reciprocal connection is usually called structural coupling, while into the organization science it is called reciprocal [4,5]. According to Ashby [6] the elementary and formally rigorous definition of organization is the existence of a functional relationship between two elements. For some links are more complex then others, the degree of complexity resides into the form and degree of constraint connections establish between elements. In fact, in parallel connection systems are almost independent (Fig. 1), because they are linked just through resources (input) sharing and/or through the contribution to the same output. These are very weak constraints indeed. The strength of the constraint increases moving to the sequential connection because the following system depends on the output of the preceding one. It is not just a “temporal” sequence, but rather a sequence implied by the fact that the following operation acts on the output of the previous one. Thus, according to

The COD Model: Simulating Workgroup Performance

115

Figure 2. The model structure.

the definition of complexity in terms of the degree of constraint, the sequential is more complex than the parallel connection. Finally, the reciprocal connection has the highest degree of constraint because it operates in both directions: system B depends on the input coming from A’s output, and vice versa. Here we see a double constraint, and then the reciprocal is the most complex connection. Moreover, the double constraint makes a radical difference, because it constitutes the essence of feedback, and therefore the essence of the cybernetic quality. Lacking the feedback relationship, parallel and sequential connections are not cybernetic interdependencies. This reasoning leads to argue that the ranking of the three basic types of interdependencies in ascending order of complexity is the following: parallel, sequential, reciprocal. This way Thompson’s [4] and Mintzberg’s [5] arguments are supported and clarified by cybernetics. Biggiero and Sevi [7] formalize these concepts, and link them to organization and cybernetics literature. Moreover, they analyze the issue of time ordering, which expands the number of the fundamental modes of connection from three to seven. However, notwithstanding these developments and a vast literature no any operative and conclusive demonstration has been supplied, neither through empirical data or algorithmically. The first aim of the COD model, therefore, is just to do it in the virtual reality. In section three it is shown that, in order to achieve a satisfying performance, a workgroup executing tasks characterized by sequential or, even more, reciprocal connections should employ progressively more and more complex coordination mechanisms. Indeed, without them performance is very low in regime of parallel connection, and near zero in regime of sequential or

116

L. Biggiero and E. Sevi

reciprocal connection. Moreover, in section four limits to agents’ computational capacity are introduced, and it is evidenced that they sharply decrease group performance. Finally, it is shown that when such limits are severe, weaker coordination mechanisms perform better. 2. The COD model architecture and the methodology of experiments 2.1. The general structure The COD modela has a hierarchical structure, which sees on the top two “objects”: TaskCreator and WorkGroup. The former generates modules while the latter manages agents’ behavior. Through a frequency chosen by the model user, TaskCreator gives the quantity and quality of modules to be executed. Thus, by simulating the external environment, it operates as an input system for the group of workers. By creating modules with more or less tasks, it defines also structural complexity (Fig. 2). In this simple version of the COD model we suppose that both the structure of modules and agents’ behavior will not change. A module is constituted by tasks, which are made by components. In each interval, that is in each step of the simulation, one component per each task is examined, and eventually executed. Here we make the simplified assumption that each single module is characterized by only one mode of connection, and that between modules, whatever is its inner interdependence, there is only parallel interdependence. Thus, the mode of connection refers to the relationships between the tasks of a single module. Finally, it is assumed that the components of each single task, regardless of its number, are always connected by a sequential interdependence. It configures a workgroup producing independent modules, which are not given at once at the beginning, but instead supplied progressively at each interval according to a given frequency. Apparently, this is a situation rather different from that usually described in the literature on technology, modularity or production management. There, few complex products and its parts are planned and available as a stock at the beginning. Therefore, in the language of the COD model, many modules are connected in various ways to build a complex output. Conversely, here, many (relatively simple) products (modules) a

The program running the model is available on the URL of Knownetlab, the research center where the authors work: www.knownetlab.it. To run the program is needed the platform and language LSD (Laboratory on Simulation Development), available at www.business.auc.dk/lsd. With the program even some indications on its concrete handling are given. Anyway, the authors are available to give any support to use it and to questions concerning the topics here addressed.

The COD Model: Simulating Workgroup Performance

117

Table 1. Agents’ behaviors. Behaviors

Description

Search

Looking and engaging for a component (of a module’s task)

Execution

Working on that component

Inactivity

Being locked into a component

are supplied alongside the simulation. Rather than a car or an electric appliance, the situation simulated by the COD model resembles more a library small workgroup. Books are modules, and cataloguing, indexing, and placing in the right place of shelves are tasks, with its own components, that is elementary operations like checking, writing, etc. A task can be in one of the following states: • not-executable, because – in the case of reciprocal connection – it is waiting for a feedback, or because – in the case of sequential connection – its preceding task has not been executed; • executable, because there is parallel connection or – in the case of sequential connection – the preceding task has been executed or – in the case of reciprocal connection – the feedback is ready. 2.2. Agents’ behavior At any interval each agent can do one of three things (Tab. 1): searching, that is looking for a component (of a task) to work on; working in the component where she is currently eventually engaged; being inactive, because she cannot do any of the previous two things. In this basic version of the model we suppose that all agents are motivated to work and that they have all the same competencies. The latter assumption will be implicitly modified by the introduction of norms, which force agents to follow a specific behavior. These norms can be interpreted as coordination mechanisms, and have been set up in a way to improve agents’ behavior, so to increase group performance. Searching is performed by an agent who is not already engaged, and therefore looking for a (component of a task of a) module. Such a searching consists in checking all components of all tasks of all modules existing in that specific interval in TaskCreator, and then in randomly choosing one of them. If she finds a component, then she engages in that same interval, and in the following interval she works it out. If she doesn’t find any free component, then she waits for the next step to start a new search. Hence, searching and engaging activity takes at

118

L. Biggiero and E. Sevi

least one interval. Of course, only one agent can engage into the same component. The agents executing a certain component will finalize the corresponding task by moving to the next component until the whole task is completely executed. In this case, while moving from a component to the next one into the same task there are no intervals spent for searching. Once ended up the last component of the task, the agent will be free to search a new component in a new task of the same module or of another module. The third possible behavior is the inactivity. It means that, temporarily or definitely, she cannot work out the component she has chosen and has been engaged. This situation occurs in one of two cases: (i) she doesn’t find any free task to be engaged in; (ii) she engages in a component of a sequential task whose preceding task has not been executed; (iii) she chooses a component of a task which is connected in reciprocal interdependence with other tasks, whose feedback is missing or delaying. In other words, she needs the collaboration of other agents, who at the moment (or definitively) are not available. It is supposed that in each step an agent can operate (work out) no more than one component, while two or all (three) agents can work on the same module. 2.3. The formalization of the modes of connection Let’s consider the two tasks X and Y, and the following formalisms: • xt and yt represent the state at time t respectively for X and Y. They indicate the state of advancement of the work of the two tasks; • α and β represent the number of components constituting respectively the tasks X and Y. In our model we consider component length of 1 step; • px and py indicate the number of components ended up into a task. A task is considered ended up when its last component has been completed. In our case X task is executed when px = α and Y task when py = β; • Ca,t indicates the contribution that the agent a provides at time t. Once engaged into a task, at each step an agent increases the degree of advancement of the task and reduces the number of remaining components. Generally, the value of C is not a parameter because it depends on the characteristics of an agent in each specific step. However, in this basic version of the model, C is assumed as a stationary value equal to 1. 2.3.1. Parallel connection This mode of connection is characterised by indirect connection through the common dependence (complementary or competitive) from the same inputs or

The COD Model: Simulating Workgroup Performance

119

through the contribution to the same module’s output. Tasks are interdependent because they are organised (interested) to achieve (contribute to) the same output. Agents engage into a specific task and execute it needless any input or feedback from other tasks. Once engaged into a task, the agent proceeds until its end. In formal terms:

for the task X

if p x < α then xt = xt −1 + Ca , t else xt = xt −1

for thetask Y

if p y < β then yt = yt −1 + Ca , t else yt = yt −1.

At each step the state of advancement increases by the value corresponding to the contribution supplied by the agent who works that task. Once all the components are completed, that is when p x = α for X task and p y = β for Y , the state of advancement stops in a stationary value. Tasks indirect dependence on inputs is represented by the fact that once agent ai is engaged into X she cannot engage into Y . It’s clear also the indirect dependence on outputs because the two components contribute to the outcome of the whole module ( xt + yt ). Let’s suppose tasks X and Y were made respectively by three and four components ( α = 3 and β = 4 ). Let’s suppose that agent a1 engages into the former task at time t1 while agent a 2 engages into the latter task at time t3 . Each agent employs one step into searching the tasks before starting working in the next step. Agent a1 starts the execution of the X task at time t 2 and, ending up a component in each step, completes the task after 3 steps at time t 4 , when the number of completed components reaches the number of components of the task ( p x ≥ α = 3 ). In the same way the Y task is ended up after four steps at time t 7 , when p y ≥ β = 4 . The whole module is considered completed at the major time ( t 7 ), with a state of advancement equals the sum of the state of the two tasks ( x7 + y7 = α + β ). 2.3.2. Sequential connection It is characterised by the fact that the output of a system -a task, in the present model- enters as input into the following system. This is a direct asymmetric dependence relationship. In formal terms:

120

for the task X

L. Biggiero and E. Sevi

if p x < α then xt = xt −1 + Ca ,t

for the task Y

else xt = xt −1 ; if p y < β and

px ≥ α

then yt = yt −1 + xt −1 + Ca , t else yt = yt −1 ; As in the parallel connection there is also an indirect interdependence related to resources sharing, because if agent ai is engaged in X she cannot work in Y . The task Y depends entirely on X either because it takes into account the state of X in the previous step ( yt = yt −1 + xt −1 + Ca ,t ) or because if all components of X are not executed ( p x < α ), then Y cannot start ( yt = yt −1 ). Workflow crosses sequentially both tasks and the final output is obtained only with the completion of task Y . It is clear the asymmetry of the relationship: while task X acts autonomously, task Y depends on (adapts to) X ’s behaviour. Let’s suppose the task X and Y were made by three components ( α = 3 and β = 3 ). Let’s suppose that agent a2 engages into Y task at time t1 while agent a1 engages into X at time t3 . Since the starting of the Y task needs the output of X , at time t 2 agent a2 will not be able to start working on Y . In fact, because a1 engages into the former task at time t3 , execution of X task is started yet on time t 4 and, as in the parallel case, it is ended up after three steps at time t 6 (when p x ≥ α = 3 ). Only from the next time t 7 , agent a2 can start the execution of Y that is completed after three steps at time t9 (when p y ≥ β = 3 ). The whole module is considered executed at the end of the latter task at time t9 , with a state of advancement of works ( yt ) equals to 6. 2.3.3. Reciprocal connection The reciprocal interdependence is characterised by a situation like the sequential connection plus at least one feedback from the latter to the former taskb. The output of a task enters as input into the other, and vice versa. Therefore this connection can be truly considered a kind of interdependence, because the dependency relationship acts in both directions and employs the output of the

b

It can be hypothesised (and indeed it is rather common) a double feedback from the former to the latter component. The question concerns from which component comes out the final outcome. This double or single loop becomes rather important and complex when considering more than two systems and when dealing with learning processes.

The COD Model: Simulating Workgroup Performance

121

connected systemc. The major complexity of this connection respect to the previous ones mirrors in the formalisation too. The formalisation of the two tasks is symmetric.

for the task X

if ( p x = 0 AND p y = 0) OR ( p x ≤ p y AND p y < α ) then xt = xt −1 + yt −1 + Ca , t

for the task Y

else xt = xt −1 ; if ( p x = 0 AND p y = 0) OR ( p y ≤ p x AND p y < α ) then yt = yt −1 + xt −1 + Ca , t else yt = yt −1 ;

The function of dependence is formalised in the same way of the sequential connection, but now it occurs in both tasks. The execution of a task should take into account what happened in the other ( xt −1 + yt −1 ). They should exchange its outputs at the end of each component, so that workflow crosses over time both tasks. For instance, if the task X worked out component 1, in order to execute component 2 it needs to get from task Y the output of its component 1. The work on a task cannot take place until in the other at least the same number of components is executed. In the formalisation this is represented by the following conditions: if ( p x ≤ p y ), and if ( p y ≤ p x ). Thus, tasks exchange feedback as many as the number of its components. To illustrate the model, let’s suppose a module made by two tasks, both constituted by three components ( α = 3 and β = 3 ). Let’s suppose that agent a1 engage into the second task at time t1 , and that in the next step t 2 she works out the first component, while agent a3 , still at time t 2 , engages into the first task. At next time t3 , in order to proceed with the second component ( yt = yt −1 + xt −1 + Ca ,t ), the Y task needs to work out the output of the first component of the X task. The execution of the second component of the Y task cannot start until on the X task a number of components at least equals to the Y task have not been worked out. In formal terms when ( p y ≤ p x ). This way the second task results temporarily locked-in and the agent a1 cannot do anything else than being inactive in time t3 and try to work again in the next time. At time t3 agent a3 executes the first component of first task and, by giving its output to the second task, allows the execution of the second component by agent a1 at time t 4 . Remind that a module is considered completed only when all its tasks have been worked out. c

Actually both the conditions must hold.

122

L. Biggiero and E. Sevi

It is important to underlie that, even just from looking carefully at the formalisation, the reciprocal interdependence is much more sensitive to the risk of inactivity and/or delay. Tasks are really interdependent both on the exchanged output and on the time at which the transfer happens. Though in principle a module characterised by reciprocal connection can be worked out by a single agent moving between tasks, this implies delays for searching and engaging. Therefore, supposing modules of the same length, the best performance occurs when all tasks are taken at the same moment by an agent, because simultaneity eliminates delays.

2.4. Norms and coordination mechanisms Norms and coordination mechanisms would not be necessary if the pure chance were sufficient to assure a satisfying, even if not maximum group performance. However, as we will discuss in next section, through our model we experimented that, without some elementary norm, the performance is unsatisfying in regime of parallel connections and nearly zero for the other two types of connection. We have hypothesized six norms (Tab. 2), and its corresponding coordination mechanisms [5]. Cooperation Norm guarantees that every agent is willing to cooperate: nobody defects its job or voluntarily defeats the colleagues purposes. Finalizing Norm drives agents to complete the task in where they are engaged by moving from the current to the next component. As we have shown in 2.2 and 2.3 paragraphs, agents can be engaged in not-executable tasks. The Anti-inactivity Norm works out this inactivity by prescribing that agents leave locked task and search for another one. Since this norm works out the situation of inactivity but doesn’t prevent it, we introduce the Outplacement Norm able to prevent choosing locked components. This norm works on sequential connection by forbidding agents to pick tasks following tasks not yet executed, while in reciprocal connection by driving agents to avoid tasks that are waiting for feedback. The Focusing Norm prescribes that agents give precedence to incomplete tasks by choosing tasks of modules in progress. More complex is the Norm of Collaboration, since it recommends that agents choose with priority tasks currently under working, that is, incomplete tasks on which other agents are currently engaged. The first five norms are forms of weak planning focused on tasks, because agents are told how to search and cope with tasks, overlooking other agents. However, they are weak forms of planning, because they don’t specialize agents on a specific kind of task or module, and neither they are directed by a

The COD Model: Simulating Workgroup Performance

123

Table 2. Norms and coordination mechanisms. Type of norm 1. Cooperation Norm 2. Finalizing Norm 3. Anti-inactivity Norm (1 + 2 +3)

4. Anti-trap Norm (1+2+3+4) 5. Focusing Norm (1+2+3+4+5) 6. Collaboration Norm (1+2+3+4+5+6)

Description Every agent does work (nobody defeats, free-rides, defects or loafs) Once started a task, agents must end it up moving from the current to the next component. Agents forced to inactivity because engaged in a locked task leave it immediately and move to search another task. Agents avoid to be engaged in locked tasks. In sequential connection they avoid tasks following tasks not yet executed, while in reciprocal connection they avoid tasks that are waiting for feedback. Agents give priority to choose tasks of modules in progress. Agents give priority to choose tasks of modules under working by other agents.

Corresponding coordination mechanisms Planning agents’ behavior Planning agents’ behavior

Planning agents’ behavior

Planning agents’ behavior

Planning agents’ behavior Favoring reciprocal adaptation

supervisor. In fact, the corresponding configuration of the workgroup is not hierarchical: it is a group of peers who do not coordinate directly one another. The sixth norm is qualitatively different, because it addresses precisely to agents’ collaboration, and thus it configure the group as a teamwork. These norms have been applied in a cumulative way, increasing complexity at each level. Some of them can be seen as a sub-set of the previous one. Higher complexity means that each norm implies more constraints than the previous one. This way to measure norm complexity equals that used to measure the complexity of the modes of connection. These constraints limit agents’ behavior, by addressing their efforts in a more effective and efficient way. By constraining behaviors many wrong choices are prevented, and thus, group performance increased. The issue of the role played by norms and coordination mechanisms pertains to the theory of coordination and not to the theory of interdependence: while the former deals with concrete or abstract objects, like systems, tasks, activities, etc., the latter deals with and refers to agents. Tasks (systems) are connected in a certain way, and that way is defined by the technology. Agents are coordinated in a certain way, and that way is defined by the norms that somebody (or the agents themselves) sets up and applies. The rationale for the need of norms and coordination mechanisms is more complex and cannot be extensively discussed here. We can just say simply that without them group performance results unsatisfying or just null. As we will argue in next section, the need for norms progressively more complex can be carried on as the

124

L. Biggiero and E. Sevi

demonstration that some type of connection is more complex than others. Moreover, norm complexity can be measured in the same way as the complexity of the modes of connection, that is in terms of the degree of constraint they put on agents’ behavior. The more restrictive, that is the more limiting agents’ choices, the more complex they are. Constraints limit agents’ behavior by addressing their efforts to a more effective and efficient way, so that wrong choices are prevented and (consequently) performance increased. Notice that the COD model does not deal with the issue of how norms are set up or emerge or eventually change. Moreover, respect to Mintzberg’s categorization [5] of coordination mechanisms, here managers’ supervision is not considered. Finally, components are supposed to be totally standardized, task eventually differ only for the number of components, and modules for the modes of connection among its tasks.

2.5. The methodology and working of the model Our model analyzes the effects of task interdependence by showing how, in order to get a satisfying performance, more complex connections require more complex norms. Group size is fixed at 3 agents, whose performance is measured by the following two indexes: • effectiveness: number of executed modules divided by the maximum number of executable modules. This index varies between 0 and 1; • efficiency: number of working steps divided by the maximum number of steps that can be employed in working. This index refers to the degree of use of inputs, which here is constituted by the agents’ time actually employed for working divided by the maximum number of steps that the group can employ for working, thus excluding the steps for engagements. Two aspects should be underlined: (i) these indexes are normalized on group size and structural complexity, so that effectiveness and efficiency are independent on them; (ii) maximum efficiency doesn’t necessary correspond to maximum effectiveness because only through an adequate coordination agents efforts can be addressed on the right tasks and resources can be minimized. Experiments are conducted on specialized groups, that is groups executing modules characterized by only one of the three modes of connections. Thus, performance of workgroups specialized on parallel modules (henceforward labeled as (P), sequential modules (S) and reciprocal modules (R) are analyzed separately. We run 10 simulations per each experiment, using different values for the random generator, so to prevent its possible influence on results. Data

The COD Model: Simulating Workgroup Performance

125

record the mean of performance indexes given by each series of 10 simulations. Each simulation lasts 900 intervals, and the module creation frequency is fixed on 0,50 and kept constant during the experiments, so that in each simulation are respectively generated 450 modules. Each module is composed by three tasks and each task by three components, that is, each task needs three working steps and one engaging step. Given these environment parameters, a group of three members has a production capacity of 675 tasks resulting in 225 modules. It means that, when the group performs at maximum, it completes 225 modules corresponding to its maximum productive capacity. According to a satisfying, and hence non-maximizing approach to social sciences [8,9,10], it is important to underlie two types of group performance: the conditions under which is reached respectively the maximum and the satisfying performance. In the former case maximum effectiveness and efficiency are achieved, while in the latter a satisfying performance could be enough. The rationale is that groups whose members work efficiently –that is, they don’t waste time in searching in vain or remain tapped in blocked components- and whose effectiveness is acceptable can be judged positivelyd. Our library small workgroup is therefore supposed at maximum to complete the storing of 225 books (modules) per year, that is 75 books per worker, which means 9 working days per book. At first sight this is not a hard goal to reach if books are simple, and if this would be the only activity of librarians. Saying that books are simple means, as we will show with our simulation results, that tasks (cataloguing, indexing, and placing) can be executed independently, that is in a parallel regime. In other words, each worker could independently work on one of the tasks related to the same book. The situation would change slightly whether task connections were sequential and dramatically whether were reciprocal. The difficulty would further increase whether agents’ ability to search among incompletely stored books were limited. In both cases, and of course much hardly when these conditions occur simultaneously, it would be really difficult to reach, if not the maximum, at least a satisfying performance without employing a set of complex coordination mechanisms. This is one of the things we are going to discuss in next section with the results of our experiments. d

Though it is not treated in this basic version of the model, a crucial point is that each norm has its own cost, and that more complex norms are more costly. Adding more and/or more complex (costly) norms increases effectiveness maybe up to the maximum, but it should be checked in each specific case whether the advantages coming from the maximum effectiveness do compensate the disadvantages coming from managing more numerous and eventually more complex norms.

126

L. Biggiero and E. Sevi

3. The effects of task interdependence We base on the following argument our ordering of modes of connection in terms of complexity: a mode is more complex than another one if, ceteris paribus, a workgroup operating in a certain mode of connection requires that, in order to reach the same performance, more numerous and complex coordination mechanisms should be employed. This is an indirect demonstration, based on computational experiments. Although this demonstration results a coherent ordering of modes of connection, it is different to that used above (see the introduction section) and in other works [5] where complexity is defined in terms of degree of connection constraint. Workgroups facing with parallel interdependence are not complex, and don’t need special devices to be effective. Conversely, in regime of sequential or reciprocal interdependence complexity grows up, and consequently coordination becomes more complex too. In spite of model simplifications, our analysis confirms the main suggestions coming from consolidated literature on this subject. The radical difference is that now such statements are based not on the “ipse dixit”, that is on the reputation of some scholar, but rather on an algorithmic demonstration. Incidentally, in our case the “scholars” were perfectly right in the main arguments. The results of our simulation model (Tab. 3) show that the Cooperation Norm is actually unable to help groups achieve adequate performance. Effectiveness is low whatever the mode of connection, while efficiency reaches a satisfactory level only in the parallel connection. The Finalizing Norm guarantees an almost satisfying performance only to the group working tasks connected with parallel interdependence, while in the other two cases agents are locked into tasks that cannot be executed. In the sequential regime too many agents engage into tasks successive to those not yet completed, and in the reciprocal interdependence they wait too long for a feedback from other tasks. In most simulations agents are almost all yet locked in the early steps, so that the group soon enters in an irreversible paralysis. The Anti-inactivity Norm prescribes that agents locked into a task leave it immediately (during the same interval when they engage in it) and search for another task. Hence, this norm works out the situation of inactivity but doesn’t prevent it, because it intervenes on the effects and not on the causes of inactivity. This norm leaves untangled the performance of the group working in parallel regime, because there is no inactivity to be worked out, and it improves a little bit the group working with the sequential mode. The performance of the

127

The COD Model: Simulating Workgroup Performance Table 3. The effects of task interdependence, main results from the simulation model.

1. Cooperation Norm

2. Finalizing Norm (1 + 2)

3. Anti-inactivity Norm (1 + 2 +3)

4. Anti-trap Norm (1 + 2 +3+4)

5. Focusing Norm (1+2+3+4+5)

6. Collaboration Norm (1+2+3+4+5+6)

Effectiveness

Efficiency

P

0.16

0.53

S

0.03

0.19

R

0.01

0.23

P

0.66

1.00

S

0

0.01

R

0

0

P

0.66

1.00

S

0.54

0.76

R

0.16

0.58

P

0.66

1.00

S

0.65

1.00

R

0.29

0.74

P

1.00

1.00

S

1.00

1.00

R

0.79

0.79

P

1.00

1.00

S

1.00

1.00

R

1.00

1.00

reciprocal group remains definitely unsatisfactory. Indeed agents consume a lot of time in searching. In order to substantially improve the performance another norm becomes necessary. The Anti-trap Norm prevents choosing locked tasks. Its action requires that agents know the right sequence of execution of each task. While the group working in the regime of reciprocal tasks remains into its respective quasi-satisfying and bad performance, the group facing with sequential tasks reaches the same outcomes of the parallel regime. Through the Focusing Norm, which prescribes that agents choose prior incomplete modules, a sharp increase of performance is realized. It brings groups in both parallel and sequential to the maximum. Once focusing agents on the same modules, their efficiency pushes effectiveness. Even the group working with reciprocal tasks benefits substantially from this norm, but it doesn’t yet reach the maximum performance. To this aim it is necessary the (final) Norm of Collaboration, which forces agents choosing firstly modules currently under working, that is, incomplete modules on which agents are engaged in. This norm is more restrictive than the

128

L. Biggiero and E. Sevi

previous one, because, in order to get priority, it is not enough that a module is incomplete. It is even necessary that in that module other agents are currently working on. By adding this norm all the three types of interdependence reach the maximum performance. Notice that this norm is qualitatively different and more complex than the previous ones: it establishes coordination between agents, while the others intervene on the relationships between agents and tasks.

4. The effects of bounded rationality Our model is truly non-neoclassical, because: a) agents are rule followers 8, 9 and not utility maximizers; b) agents’ rationality is bounded. 10, 11, 12, 13, 14 Currently there is a hot debate concerning the ways to operationalize bounded rationality so to give it a sound and effective scientific status. Among the many ways in which this could be done and the many facets it presents, in our model we chose one of the simplest: agents’ computational capacity. The idea is that agents cannot look at and compute all executable modules, because they should be checked and computed in order to decide which one is better to be executed, and in which component of its tasks. Fig. 3 shows that the challenge to agents’ rationality sharply increases over time, at the half of group working life there are at least 112 incomplete circulating modules to be computed by each agent. In particular, the problem is generated by the progressive proliferation of incomplete modules and tasks. Let’s consider the best situation of efficiency and effectiveness: a group working in regime of parallel connections, where agents are coordinated by the Collaboration Norm, that is the most effective coordination mechanisms. And further, let’s suppose they have no computational limits in searching, checking, and computing modules and tasks (Fig. 3). Well, even in this most favorable case already after early 20% of group working life, that is after 180 intervals, around 45 incomplete modules do circulate (Fig. 3). In the best conditions – easiest regime and most effective norms- in a single interval each agent after 180 intervals should be able to compute 45 books, that is 135 tasks (405 components). The size of the decision space becomes too large very soon. Tables and shelves of our small library workgroup progressively and soon become filled in by incompletely stored books, and hence the degree of disorder grows accordingly. Every day becomes much harder to compute all the incomplete modules in order to choose the right one. Even in the best conditions the goal which at first sight appeared so easy to achieve becomes unreachable. If the yearly duration of working time of an average American worker were supposed

The COD Model: Simulating Workgroup Performance

129

Figure 3. Complete and incomplete modules circulating in a group working in parallel regime and coordinated by the Collaboration Norm and with unboundedly rational agents. NMod: number of created modules; NModCompl: number of executed modules; NModToCompl: number of uncompleted modules.

to be 1800 hours, in our 900 steps simulations each interval would correspond to 2 working hours. Now, the problem of agent computational capacity sets up in the following way: how many modules can be looked at and computed (analyzed) in one interval (2 hours)? This problem is crucial, because group effectiveness and efficiency depends essentially on this ability. An agent, in fact, has to “open” and check all incomplete modules in order to choose the task to be worked out. Let remind that each module is made by 3 tasks, each constituted by 3 components. At the very end, an agent will choose a specific component of a specific task of a specific module. She will do that taking into account the specific interdependence regime and the norms eventually ruling that group. Even assuming that librarians work of storing is effectively supported by computer programs, it could be reasonably supposed that in a 2-hour-length standard interval a highly competent (strongly rational or efficient) agent can check 40 modules (equals to 120 tasks, equals to 360 components), while a lowly competent or motivated just 2. Tab. 4 results show that if the group is in the parallel or sequential regime and it is coordinated through the most effective norm, then the achievement of a satisfying performance requires a computational capacity of at least 20 modules per each agent. Consequently, only with a very high rationality joined with the best coordination mechanism it is possible that a group deals with complex task to achieve a satisfying performance. If the regime is in the reciprocal mode, then it is requested the double capacity. Let say that this regime needs librarians with double competence respect to the other two regimes. If the group is coordinated by a less effective norm, then in the reciprocal regime the performance will

130

L. Biggiero and E. Sevi

Table 4. Agents’ computational capacity effects on group performance. Group 1: Coordination through Anti-trap Norm. Group 2: Coordination through Norm of Collaboration. Group 1 Agents’ computational capacity in terms of modules 2

Effectiveness P

S

0.66

0.66

0.28

R

Efficiency P

S

R

1.00

1.00

0.74

5

0.67

0.66

0.28

1.00

1.00

0.74

10

0.65

0.64

0.28

1.00

1.00

0.74

20

0.65

0.65

0.29

1.00

1.00

0.74

40

0.66

0.66

0.30

1.00

1.00

0.74

80

0.66

0.65

0.29

1.00

1.00

0.74

Group 2 Agents’ computational capacity in terms of modules

Effectiveness

Efficiency

P

S

R

P

S

R

2

0.37

0.43

0.17

0.37

0.44

0.32

5

0.52

0.58

0.28

0.52

0.59

0.41

10

0.63

0.69

0.40

0.63

0.70

0.51

20

0.75

0.79

0.58

0.75

0.80

0.64

40

0.85

0.88

0.75

0.85

0.89

0.78

80

0.93

0.95

0.89

0.94

0.95

0.90

never overcome 30% even supposing a computational capacity of 80 modules per agent. It means that in presence of complex task interdependence and without collaborating at their best there is no way to reach a satisfying performance. Tab. 4 shows also another interesting result: at the lowest degrees of computational capacity it is more effective a simple coordination. In fact, when the regime when connection is parallel and computational capacity is less than 20 modules, Group 1 performs better than Group 2. Similarly, when the mode of connection is sequential or reciprocal, group 2 performs better than group 1 only if computational capacity exceeds 10. This is due to the fact that once rationality is really bounded, it is so also as concerning goal seeking behaviors. In fact, the norms of focalization and collaboration tell agents searching for two specific module categories: in progress and under the execution of other agents. However, the problem is that, if computational capacity is under a certain threshold, then the time consumed for searching those specific module categories is high. In particular, it becomes so high to vanish the advantages of being more goal seeking. The effectiveness of goal seeking behavior is more

The COD Model: Simulating Workgroup Performance

131

Figure 4. Effective combinations among bounded rationality and coordination complexity.

than compensated by the ineffectiveness of spending a lot of time in the searching activity. In other words, leading to less specific goals, that is allowing for a wider range of choices, less complex norms reduces the effort of searching and increases effectiveness of less rational agents. This explanation is confirmed by the analysis of efficiency, which actually is inversely dependent on the time spent in searching activity. When mode of connection is parallel or sequential, whatever the computational capacity of agents, the efficiency of group 1 is much higher than the efficiency of group 2. Similarly, when tasks are connected in a reciprocal way, group 2 scores a higher efficiency only if agents have a high computational capacity. Figure 4 summarizes the effective combinations among bounded rationality and coordination complexity. If the workflow arriving to the workgroup from external environment were better regulated than a flow of 0,5 module per step, then the performance would be, ceteris paribus, much higher, and in particular there would be less incomplete modules. In other words, in order to reach satisfying performances, workgroups would need less rationality or less (complex) coordination norms. At the extreme of a perfect regulation, the number of modules arrived from the external environment would coincide with those completed, and goal achievement would request no high rationality (just one module per agent per interval) and not all the norms. On the other hand, it is likely (but left as well to future research agenda) that a workflow more uncertain than a constant rate of module creation would require ceteris paribus more rationality or more complex coordination mechanisms. Such an increase of complexity associated with unstable workflow could be more than compensated by introducing agents’ learning in one or more of these three forms: i) better searching ability after completing tasks; ii) higher

132

L. Biggiero and E. Sevi

productivity when working on the same task; iii) major ability to collaborate as the number of successful collaborations grow over time. Actually, in this basic version of the model agents do not learn and therefore the corresponding forms of nonlinearity do not take place.

5. Conclusions Our simulation model tells us that groups working on complex interdependencies can reach an acceptable performance only by means of complex norms. Reciprocal interdependence can be managed satisfactory only through the Focusing Norm, and reaches the maximum only through the Norm of Collaboration, which actually includes five simpler norms. Sequential interdependence can be satisfactorily managed by applying the Anti-trap Norm, which includes three norms, and the parallel interdependence already with the Finalizing Norm. These results have also a normative side: it is redundant to employ complex norms to coordinate groups working on tasks connected by simple interdependencies. Further, and quite surprisingly, when agents’ rationality is severely bounded, the Collaboration Norm becomes not simply redundant, but indeed disadvantageous. In other words, coordination between agents does not work well when agents’ computational capacity is very low. Well focused taskbased coordination would perform better. Of course these results, and especially this normative side should be taken with prudence, because our model is still extremely simple. The introduction of knowledge exchange, competencies, personal conflicts, learning processes, and task specificity could change them significantly. However, by now we obtained four relevant findings: 1) an algorithmic demonstration of the ordering of interdependencies in terms of complexity; 2) an operationalization of bounded rationality in terms of computational capacity; and 3) an algorithmic analysis of the effects of bounded rationality on workgroup performance, which takes into account also task interdependence as a moderating factor; 4) an explanation of why and under what circumstances teamwork is a superior organization. This latter result confirms suggestions proposed, but left theoretically and empirically unproved in organization science. This version of the COD model is simple, because it supposes that agents have the same competencies and motivations to work, that they don’t learn, that they don’t make mistakes, that there are no behavioral issues (personal conflicts, leadership problems, etc.), and that there are no differences between tasks. Moreover, there are no externalities, neither other forms of nonlinear

The COD Model: Simulating Workgroup Performance

133

phenomena. However, despite its simplicity, this model is very helpful either because it is the ground on which building more complex and realistic models or because it already shows many interesting effects. Moreover, by the inner logic of simulation models, in order to be able to explain results coming from rich (complex) models it is necessary to know the behavior of the variables in simple (controllable) models.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

L. Biggiero, Towards a new theory of interdependence, (2008), www.ssrn.com . F. Varela, Principles of Biological Autonomy (Elsevier, NY, 1984). C.W. Churchman, The Systems Approach (Dell, NY, 1979). D.J. Thompson, Organizations in Action (Mc-Graw, New York, 1967). H. Mintzberg, The Structuring of Organization (Prentice-Hall, Inc. NJ, 1979). W.R. Ashby, in Principes of Self-organization, Ed. H. von Foerster and G.W. Zopf, (Pergamon Press, New York, 1962), p. 255. (Reprinted in Modern Systems Research for Behavioral Sciences, Ed. W. Buckley (Aldine Pub, Chicago, 1968)). L. Biggiero and E. Sevi, Modes of connection and time ordering definitions and formalisation of the fundamental types, (2008), www.ssrn.com . J.C. March, A Primer of Decision Making. How decisions happen (The Free Press, NY, 1994). J.C. March, in Organizational Decision Making, Ed. Z. Shapira (Cambridge UP, Cambridge, 1997), p. 9. J.C. March and H.A. Simon, Organizations (Blackwell Publishers, Cambridge, 1958). J. Conlinsk, Journal of Economic Literature 34, 669 (1996). B.D. Jones, Annual Review of Political Science 2, 297 (1999). D. Kahneman, American Economic Review 93, 1449 (2003). H.A. Simon, Organization Science 2, 125 (1991).

This page intentionally left blank

IMPORTANCE OF THE INFRADISCIPLINARY AREAS IN THE SYSTEMIC APPROACH TOWARDS NEW COMPANY ORGANISATIONAL MODELS: THE BUILDING INDUSTRY GIORGIO GIALLOCOSTA Dipartimento di Progettazione e Costruzione dell’Architettura, Università di Genova Stradone S. Agostino 37, 16123 Genoa, Italy E-mail: [email protected] Infradisciplinary, besides interdisciplinary and transdisciplinary, applications, forms part of the definition of new company organizational models and, in particular, for networked-companies. Their related systemic connotations characterize them as collective beings, especially regarding the optimization of interactions between agents as well as context-specific interference. Networked-companies in the building industry (chosen to illustrate the infradisciplinary values of the systemic approach towards company organizational models) require, due to their nature and particularities of context, certain specifications: behavioral microrules of an informal nature, suitable governance of their sector, etc. Their nature and particular context thus determine, especially in the systemic view, the need not only for an interdisciplinary and transdisciplinary approach, but also an infradisciplinary one. Keywords: systemics, infradisciplinarity, building, company, organization.

1. Introduction The Discorso preliminare of Diderot and d’Alembert's Enciclopedia states: “(...) there is not a single academic who would not willingly place the theme of his own study at the centre of all the sciences, in a way similar to primitive human beings who placed themselves at the centre of the world, convinced that the world had been made for them” (Diderot and d’Alembert, 1772 [3], cit. in Dioguardi, 2005 [5, p. 45], author's translation). Even today, within various disciplines, this tendency persists: • sometimes with each academic emphasising those features, carriers of assumed generalist evidence, of his/her own area of interest; • sometimes insisting tout court upon the particular collocation of their own expertise; • in other cases claiming the irreducibility of that science towards more general laws of interpretation and conduction of the phenomena, or stressing assumed structural peculiarities, etc. 135

136

G. Giallocosta

From the same Discorso, the role assigned to philosophy in the “(...) encyclopaedic order of our knowledge ...” (Diderot and d’Alembert, 1772 [3], cit. in Dioguardi, 2005, [5, p. 44], author's translation) also emerges. This encyclopaedic order, in fact, “(...) consists of collecting knowledge within the smallest possible space and putting the philosopher, so to speak, over and above this vast labyrinth, at quite an elevated observation point, from which he can completely embrace all the main arts and sciences ...” (Diderot and d’Alembert, 1772 [3], cit. in Dioguardi, 2005 [5, p. 44], author's translation). This approach leads to the fragmentation of knowledge into various disciplines. This disciplinary fragmentation remains common practice in many areas of study and research. Nor usually can one be completely free of mistaken interpretations of: • generalism, where the aim is to recompose (and/or construct general theories) but with unacceptable simplifications; • specialism, whenever means and ends inspired with scientific rigour in the interpretation and management of peculiarities (and in the relative operational activities) lead to artificial sectorialisms. In this sense systemic, especially through interdisciplinary and transdisciplinary processes (thus producing interactions, correspondences and theories at higher levels of generalisation), also leads to recomposition amongst the various disciplines. And it does so not by replacing but by integrating the specialistic knowledge of the latter: for this reason, and to avoid mistaken assumptions about centrality in all the sciences (Diderot and d’Alembert, 1772 [3], cit. in Dioguardi, 2005 [5, p. 45]), infradisciplinarity is associated with interdisciplinarity and transdisciplinarity. It is well-known that: • interdisciplinarity occurs when problems and approaches of one discipline are used in another; • transdisciplinarity occurs when systemic properties are discussed and studied in a general way, as properties of models and representations (without reference to cases in specific disciplines). Infradisciplinarity has, in its turn, already been defined epistemologically as being, fundamentally, a set of prerequisites necessary for any scientific activity and as a method of investigation regarding intrinsic aspects of disciplines (Lorenzen, 1974 [11, pp. 133-146]), and here is taken, above all, as resources and assumptions activating and validating specialistic rigour. It is thus important that it be considered in research activities, to avoid genericism.

Importance of the Infradisciplinary Areas in the Systemic Approach …

137

A example of the risks of mistaken generalism, and of insufficient attention towards infradisciplinary aspects (even though it is not strictly ascribable to the scientific disciplines), is the case of company organisational systems applied to the construction sector. The latter shows, in fact, significant and substantial peculiarities, as will be seen below: peculiarities which, moreover, are expressed to the same extent in the nature of the construction companies (especially Italian ones), leading to an indication of their particular behavior as collective beingsa, and sometimes precursors (as will be seen) of completely unexpected operational effectiveness. Current theories of company organisational models, converging, above all, towards the concept of networked-company (Dioguardi, 2007 [6]), provide a general reference framework for applications/elaborations over widely differing sectors. In this way, the latter (through the use of models, behavioral rules, etc.) can manage and coordinate innovational leanings and inherent multiple phenomenologies of various scenarios: local industrial estates, virtual industrial estates, etc.b. Also this does not exclude any specificity of such phenomenologies.

a

b

The concept of collective being expresses, above all, situations in which the system which emerges from the interactions amongst the component parts may show behaviour very different from that of its individual components: so different, in fact, as to require the dynamic use of multiple systems (interacting and emerging from the same components). This concept, when applied to the reality of companies, allows the description of new models, and thus novel possibilities of intervening in processes involving the ability to (Minati and Pessa, 2006 [12, pp. 64, 70-75, 89-113, 365-368]): decide, store information, learn, act intelligently, etc. Collective beings also refers to collective behaviour emerging from that of autonomous agents which share a similar cognitive model, or at least a set of common behavioural micro-rules (Minati and Pessa, 2006 [12, pp. 110-111]). Virtual can mean potential. In the thinking of St. Thomas Aquinas and other scholastics: - an effect is formally contained within its cause, if the nature of the former is present in the latter; - an effect is virtually present within its cause if, while not containing the nature of the former, it may produce it (Minati and Pessa, 2006 [12, p. 362]). The concept of virtual company usually refers to an electronic entity, put together by selecting and combining organisational resources of various companies (Minati and Pessa, 2006 [12, pp. 365-368]). This concept also expresses an opportunity for active cooperation amongst several companies, often having the same target. In this sense, the constitution of a virtual company implies the development of a suitable network of relationships and interactions amongst those companies, developed on the basis of customer requirements (Minati and Pessa, 2006 [12, p. 366]). More generally, the concept of virtual district comprises simultaneous meaning of: - potential (and specific) organisational development, where the constituent members appear to have significant proximity only from an IT point of view (Dioguardi, 2005 [5, p. 127]), - quasi-stability (Garaventa et al., 2000 [8, p. 90]).

138

G. Giallocosta

In this sense, therefore, when coherently adopted within a systemic view, new company organization theories avoid the risks of any pretext regarding the centrality of individual areas of application: thereby reducing the possible effects of self-referentiality. 2. Systemics connotations of networked-companies Although prior to his later contributions, Dioguardi defines the networkedcompany as “(...) a series of laboratories (...) expressed as functional areas which overall provide a network of internal operational nodes. Amongst these (...) economic transactions develop almost leading to an internal quasi-market. The company is also open to external cooperation from other companies, through transactions with suppliers, and these (...) produce a network of supplier companies which are nevertheless independent and able to activate transactions in a real external market which remains, however, an expression of the supplier network of the general company (...) The company is thus structured in a reticular way allowing the coexistence of a hierarchical order together with the efficiency of the market within an organisational harmony like that in Goethe's web of thought: “The web of thought, I'd have you know / Is like a weaver's masterpiece: / The restless shuttles never cease, / The yarn invisibly runs to and fro, / A single treadle governs many a thread, / And at a stroke a thousand strands are wed ” (Goethe, 1975 [10, p. 94]; Italian citation: “In realtà, la fabbrica dei pensieri / va come un telaio: / pigi il pedale, mille fili si agitano / le spole volano di qua e di là, / i fili corrono invisibili, / un colpo lega mille maglie”, cit. in Dioguardi, 1995 [4, p. 171], author's note) ... This company model, however, entails centrifugal freedom of movement capable of disaggregating the component companies (but also involving critical elements for the individual supplier companies - author's note)c. It is thus necessary (regarding the risks of activating such centrifugal autonomies author's note) to search for elements of aggregation and homogeneity. And these can be found precisely within the concepts of culture and quality, intended both as internal organizational requirements as well as external manifestations capable of expressing a competitive nature” (Dioguardi, 1995 [4, p. 171], author's translation). Especially in Italy, in the building sector (in the more advanced cases) models of networked-companies, or of general company, can be defined through connections “(...) at three fundamental levels: c

See, for example, particularly regarding the building sector, Giallocosta and Maccolini, in Campagnac, 1992 [2, pp. 131-133].

Importance of the Infradisciplinary Areas in the Systemic Approach …

• • •

139

the general company itself, which essentially takes on the role of managing and orchestrating (...) the multinodal company (the operational nodes of the general company) ... responsible for managing production, finance and plant and machinery; the macrocompany, consisting of the specialist external companies (...) involved, by the general company, through the multinodal company, in production and supplier activities, by means of quasi-stable relationships ...” (Garaventa et al., 2000 [8, p. 90], author's translation).

More recent conceptual developments, ascribable to modern aspects of company networksd, could lead to possible developments of traditional districts towards innovative forms having their own technological identity, able to compete in global markets. For the genesis and optimum development of these innovative structures, the themes of governance and the need for associationism amongst companies are very important; particularly the latter aspect, for which: “(...) associationism amongst companies should be promoted, with the objective of a more qualified presence in the markets, and thus one should promote the formation of networks (...) comprising not only small and medium-sized companies in similar markets, but also companies in other districts having analogous or complementary characteristics, interested in presenting themselves in an adequate manner to large-scale markets ...” (Dioguardi, 2007 [6, pp. 143144], author's translation). Clearly, the theme of governance becomes central for company networks (and networked-companies), especially where: • their formation occurs through spontaneous processes (company networks); • whenever criticality occurs (or there is a risk of it occurring) during useful life-cycles, but also due to more general requirements regarding the definition and realisation of goals (mission) and related management issues; in this sense, the existence of a visible hand (a coherent substitute for the invisible one of the market evoked by Adam Smith), deriving from the professional competence of the managers and from suitable regulatory strategies for that sector, ensures governance: thus acting as an observer, in the systemic sense, and thus active, being an integral part of the processes occurring (Minati and Pessa, 2006 [12, pp. 50-55]).

d

Such company networks “(...) lead to novel aspects of economic analysis which form units at a third level, in addition to individuals representing first level and companies second level elements” (Dioguardi, 2007 [6, p. 138], author's translation).

140

G. Giallocosta

Further systemic connotations of networked-companies lie in maximizing and, at least in the really advanced cases, optimization of the interactions amongst component companies (agents); these latter, moreover, are typical of a company collective being, expressing the ability to learn, accumulate know-how, follow a strategy, possess style, leadership, etc.: it follows that it possesses intelligence (or better, collective intelligence) in terms of, for example, the ability to make choices on the basis of information, accumulated know-how, elaborating, strategies, etc., and also when faced with peculiarities of context (Minati and Pessa, 2006 [12, pp. 110-134, 372-374])e. The explicit role played by the latter already alludes to the significant effects thus produced (but also, as will be seen, to the synergies which arise) regarding the ability/possibilities of the companies to make choices, and thus to assume suitable behavior, to follow suitable strategies, etc: the set of peculiarities factors are thus considered as agents (and in a dialogical sense with rules and general aspects) regarding the optimum development of behaviors, company strategies, etc., and coherently with innovative theories; these do in fact acquire, precisely because of this (and of the infradisciplinary aspects which it carries), continuous refinement and specification. In resolving the make or buy dichotomy, prevalently by way of orientation towards productive decentralisations (whilst maintaining strategic internal activities, avoiding a drift towards hole corporations), the networked-company also stresses its own behavior as an open system: and precisely, at least on the basis of its productive performance, in the sense of being able to continually decide amongst various levels of openness or closeness with respect to its own context (Minati and Pessa, 2006 [12, pp. 91-124]). The nature of the latter, and the decisions regarding the ways in which one can relate to it, also induce within the company, the possibility of adaptive flexibility (minimal where tendencies toward systemic closure prevail). Such strategies then evolve towards possible forms of dynamic flexibility, such that the company, even though it is suitably prepared in this sense (Tangerini, in Nicoletti, 1994 [13, pp. 387-392]), not only receives market input but modifies and develops it (Garaventa et al., 2000 [8, pp. 125-137]): for example, by anticipating unexpressed requirements, satisfying e

The concept of intelligence in company collective beings, coherently with a simplistic definition of the former, may be considered as corresponding to, for example, the ability to find the right answers to the questions, and assuming (or considering) that the right answers are not so much the real ones but the more useful ones (or rather, those that work). In this sense one can attribute intelligence to collective beings: the intelligence of flocks, swarms, companies, etc., are manifest in the specificity of their collective behavior, where only collectively (as opposed to the inability of the individual members) are they capable of solving problems (Minati and Pessa, 2006 [12, pp. 116-125]).

Importance of the Infradisciplinary Areas in the Systemic Approach …

141

latent needs, etc., and while sustaining in an evident manner, unacceptable risks of manipulating the processes of the formation of demand, the development of induced needs, etc., and for which inhibiting measures of such risks even through governance and sharing ethical codes is necessary (Minati and Pessa, 2006 [12, pp. 336-346]). Thus, there are mutual company-context influences, synergic modifications between them, following non-linear and recursive processesf. Above all, the existence of such interactions, their attributes, the implicit nature of their own connotations (closely related to the character and peculiarities of the company and of the other actors involved, and to the context in which they occur), require the use of infradisciplinary applications and improvements for: • an effective interpretation of such emergent phenomena (Minati and Pessa, 2006 [12, pp. 98-110]), • the most efficient management possible of the latter. 3. Networked-company in the Building Industry Specific aspects of the building sector (illustrating in particular its distinctive character compared to other areas of industrial activity, and for its direct interference with company organizational models) can be summarized, amongst others, by (Garaventa et al., 2000 [8, pp. 27-40] and Sinopoli, 1997 [16, pp. 4665]): • relationships with particularities of contexts (environmental, terrain, etc.)g; • technical and operational activities carried out always in different places; f

g

Non-linear processes, typical of complex systems (distinguished, moreover, by exchanges with the external environment), show behaviors which can not be formulated in terms of a (linear) function f (x) such that: f (x+y) = f (x) + f (y) and f (a*x) = a* f (x). A formulation of recursive processes, typical of autopoietic organisations (which produce themselves), can occur by means of a program (p) expressed in terms of itself, so that its execution leads to the application of the same algorithm to the output of the previous stage. A recursive program recalls itself generating a sequence of calls which end on reaching a given condition, a terminating condition. “Due to the fact of being a building, which occupies a significant portion of land over very long periods of time (...) the product of the construction process has to face up to the problem (...) of relating to the characteristics of its own context: the physical ones (climate, exposure, geology, meteorology), the environmental and historical ones (...) The relationship with its context ensures that the product from the building process adds to its economic role a series of cultural and symbolic meanings and that the agents involved in this process have to come to terms with a discipline (which industrialists almost never have to face) which deals precisely with these specific meanings, that is, architecture ...” (Sinopoli, 1997 [16, p. 48], author's translation).

142

• • • • •

G. Giallocosta

unique nature of the building; tendency towards high costs of construction; maintenance, and often increases in the economic value of the end products over the years; presence of fractionated leadership in the management of single initiatives; existence of temporary multi-organisations during the management of activities (as for other events and areas of activity, such as theatre, football or rugby matches, etc.).

The significant impact of the building industry upon regulations at the urbanistic and territorial planning level, and upon the multiple particularities of the various contexts, tends to place it in a special position within the macroeconomic scenario: there emerges, for example, the need for extremely detailed institutional regulations (Garaventa et al., 2000 [8, pp. 37-41]), and related interests and multi-disciplinary values (social, economic, cultural, etc.). In the building sector, the technical and operational activities are always carried out in a discontinuous manner and in different places (building sites) and burdened with significant risks (unforeseeable nature of climatic and environmental factors, etc.), supplying unique end-products: thus excluding rigorous programming of operational activities, any meaningful production standardizations, etc. The tendency towards high costs of the finished product, and the relatively long periods necessary for the design and production, often lead to burdensome outlays of economic resources by the construction companies, with consequent heavy negative cash-flow for the latter which usually persists over the whole period of the contract. The maintenance (and often the increase) over time of the economic value of the end-products also explains the lack of interest of those involved in this sector (compared to other sectors of economic activity) regarding aspects of productivity, technical innovation, etc.h: emphasis is, in fact, placed upon factors more ascribable to rent income rather than industrial profit (or the optimization of productive activities). h

The building industry, in fact, “(...) is ‘a strange world of suspicious people (...) who often reject (...) innovations which upset (...) behaviour and (...) habits’ (Sinopoli, 1992 [15, p. 12], author's note and translation) ... The mistrust is (...) deeply rooted since this community has accumulated millenia of experience, producing great works using almost all types of material available in nature (...) This ‘strange world’, thus, takes from experience the criteria of evaluation of possible innovations accepting external stimuli only through a long and not very clear process of ‘metabolism’ through which novel proposals are compared with the order of a system dominated by the persistence of intrinsic conditions, such as (...) the specific nature of the product and in the ways of producing it ...” (Maccolini, in AA. VV., 1996 [1], author's translation).

Importance of the Infradisciplinary Areas in the Systemic Approach …

143

Distinct from other industrial activities, where in almost all cases a single agent (the managing director) guarantees through her/his own staff the management and control of the various sub-processes (analysis of demand and of the market, product design, construction, marketing, etc.), in the building industry the functions of leadership are fractionated and assigned to a number of agents (designer, builder, etc.); the latter, moreover, are only formally coordinated by the client (who usually does not possess any significant professional competence). Often this situation, even when some of those agents (often the construction company, especially in Italy and notwithstanding initiatives taken to follow European Directives) take on a central role in the productive processes (Garaventa et al., 2000 [8, pp. 74-76, 120]), leads to conflict, legal wrangling, etc., heightened, moreover, by the existence of temporary multi-organizations. The latter (Sinopoli, 1997 [16, pp. 60-65]): • are formed for the period necessary for managing a given activity (clearly excluding the possibility of accumulating any common experience, as it is unlikely to be usable on successive occasions), • are composed of organizations (or agents, such as designers, companies, suppliers of materials and components, etc.) which, although each is independent, make decisions which depend upon (or interact with) those taken by the others. Situations emerge, endogenous to the building sector, which lead to, well beyond the peculiarities ascribable to the various social, local or other aspects, structural dissimilarities of processes and production in the building sector with respect to other sectors of industrial activity. With other so-called non-Fordist sectors (not in the historical sense), the building sector certainly possesses original modes of production and accumulation (Garaventa et al., 2000 [8, p. 28]). Symptomatic of this, for example, especially during its phases of greatest development, are the significant capital earnings ability but also the low productivity of the work done (which contributes much less to profit formation, as is well known, with respect to other industrial sectors). The character and the specific nature of the building sector thus become the aspects and questions to face up to in order to develop networked-companies in this industry. Above all, a governance as a consequence of public policies besides, naturally, the visible hand of a management possessing a suitable business culture for the building industry, will ensure sufficient compatibility with: • the significant implications for urbanistic and territorial layouts, • the processes of the formation and satisfaction of demand,

144

•

G. Giallocosta

harmonic developments (Giallocosta and Maccolini, in Campagnac, 1992 [2, pp. 131-133]) of supply and of the markets (general companies having responsibility for production and orchestration, specialist companies, independent companies active in maintenance, micro-retraining, renovation, etc.).

Naturally, the importance of the existence of business ethical codes is also clear, especially: • in dynamic flexibility strategies, • in make-buy optimizations which do not damage the competitiveness nor the independence of the supplier network (Giallocosta and Maccolini, in Campagnac, 1992 [2, pp. 131-133]). More generally, the requirements of governance seem to take on the aspects of an active observer (as recalled above in a precisely systemic sense), but with leading roles, nature and attributes, in the face of the peculiarities of the sector. The latter also concern, as mentioned above, particular methods relating to models of operational organization, and planning activities, standardization (only as a tendency and of limited reliability), etc. Above all, informal procedures become particularly important within the organizational structures of the sector. This phenomenon is of particular importance in the Italian situation: the significant “(...) presence of informal procedures, which often spill over (...) into aspects which are, at least, problematic within a normal framework of productive efficiency (...), do not, however, hinder the realization of buildings of a good qualitative level, especially for small and medium-sized buildings without particularly complex plant (...) One often observes, for instance, the high quality of the finishing of buildings in Italy (work in which precisely those informal procedures are the most widespread - author's note), with respect to the situation in France or in Britain” (Garaventa et al., 2000 [8, p. 93], author's translation). In this country, the “(...) companies work mainly by using techniques and rules of the trade learnt at the individual level (...) This is personal individual knowledge, rather than know-how or operational procedures developed by the company (...) The lack of any formalized rules of the trade lead to processes of professional training during the work itself (...) which produces certain homogeneity of the agents at all levels (...) Thus the development of responsibility of the operatives, (...) their common hands-on training, end up generating a significant understanding amongst the various operatives. This allows a good product from a largely non-formalized context (...) The operatives seem to have a unity of intention (...) which (...) renders possible the realization

Importance of the Infradisciplinary Areas in the Systemic Approach …

145

of buildings of acceptable quality” (Garaventa and Pirovano, 1994 [7], cit. in Garaventa et al., 2000 [8, p. 94], author's translation). In this sense, that particular behavior as a collective being derives from the sharing of a cognitive model taken up through training activities which are often not formalized (especially in Italy) but which can establish common behavioral micro-rules (Minati and Pessa, 2006 [12, pp. 110-111]): • notwithstanding discontinuous and heterogeneous experience, • provided that customary forms of intervention existi. Thus, in networked-companies in the building industry and, more generally, for the multiple forms of business aggregation to be found there, that dual organizational order, formal and informal, typical of socio-technical systems, develops: • where the latter translates the set of unwritten rules through the former, but originating from the distinct personalities of the operators and thus decisive in reaching successful conclusions or in determining the unsuccessful ones (Dioguardi, 2005 [5, pp. 87-89]), • with particular emphasis upon the peculiarity and the importance of the informal organization, faced with the analogous peculiarities of the sector. Here, therefore, amongst other factors distinguishing the sector from other areas of economic activity, any coherent development of networked-companies requires: • validation and optimization of the work done, even as the effects of informal procedures (as far as they can be ascribed to compatible type of intervention); • maximizing operational flexibility (for the work done and/or differentiated activities).

i

“The unresolved problem of the Italian building sector is that the unity of intention under these conditions is only possible for traditional working practices and for relatively simple buildings. In large and complex buildings (...) the operatives, especially at the operational level on the building site, lose the overall view of the job (...) and, with that, the unity of intention (...) In an analogous manner, the technical quality required for innovative building work can not be reached using the unwritten rules of the trade (...) On the other hand, the efforts being made to formalize processes and technical rules have the effect of destroying the context which generates artisan know-how and the unity of intention: (...) the Italian building industry finds itself in a difficult and contradictory situation ...” (Garaventa and Pirovano, 1994 [7], cit. in Garaventa et al., 2000 [8, p. 94], author's translation).

146

G. Giallocosta

Similarly, the diseconomies which still burden the processes of producing buildings, and which to different extents penalize the operators involved (negative cash-flows for the companies, high costs and prices for the customers, users/clients, etc.), demand innovation capable of significant reductions in these phenomena, notwithstanding the emphasis still placed upon rent earnings. In this sense, and also for questions of a more general nature (Giallocosta and Maccolini, in Campagnac, 1992 [2, pp. 131-133]), networked-company models above all, being validating procedures of productive decentralization in the building industry, require (together) regulatory activities regarding: • appropriate limits and well-defined suitability of such procedures; • policies of containment of the costs of intermediaries, often exorbitant and inherent in such activities. Clearly, there is also need for more innovative tendencies which: • optimise quality-cost ratios of the work done, • put into place shared rules and formal procedures, especially for complex and technologically advanced activities (where competition is ensured through interchangeability of know-how and operators). For the latter, naturally, there are aspects common to other industrial sectors, but, for the reasons outlined above, they acquire particular importance in the building industry and, thus, require appropriate governance. 4. Conclusions Networked-company models are emblematic of the infradisciplinary aspects of the systemic approach. Within these models one can, in fact, verify the effectiveness of the most recent theories of business organization, typical of the late-industrial era. At the same time, however, given the current specific aspects of the building sector, there are other peculiarities ascribable to: • networked-company in the building industry, • consistent developments. Thus, the infradisciplinary contributions to systemics (as adjuvant activities which process phenomena having multiple components within a system) do not

Importance of the Infradisciplinary Areas in the Systemic Approach …

147

lead to reductionismj, as long as there are no mistaken assumptions of centrality (Diderot and d’Alembert, 1772 [3], cit. in Dioguardi, 2005 [5, p. 45]). Moreover, within the terms mentioned above, the harmonious deployment of transdisciplinarity, interdisciplinarity and infradisciplinarity become essential. As observed above, a networked-company in the building sector provides an exemplary case of this. References 1. AA. VV., Nuove strategie per nuovi scenari (Bema, Milan, 1996). 2. E. Campagnac, Ed., Les grands groupes de la construction: de nouveaux acteurs urbains? (L’Harmattan, Paris, 1992).

3. D. Diderot, J.B. d’Alembert, in Enciclopedia o dizionario ragionato delle scienze, 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

j

delle arti e dei mestieri, 1772 (Laterza, Bari, 1968). G. Dioguardi, L’impresa nella società di Terzo millennio (Laterza, Bari, 1995). G. Dioguardi, I sistemi organizzativi (Mondadori, Milan, 2005). G. Dioguardi, Le imprese rete (Bollati Boringhieri, Turin, 2007). S. Garaventa, A. Pirovano, L’Europa dei progettisti e dei costruttori (Masson, Milan, 1994). S. Garaventa, G. Giallocosta, M. Scanu, G. Syben, C. du Tertre, Organizzazione e flessibilità dell’impresa edile (Alinea, Florence, 2000). P. Gianfaldoni, B. Guilhon, P. Trinquet, La firme-réseau dans le BTP (Plan Construction et Architecture, Paris, 1997). J.W. Goethe, Faust (Penguin Classics, Middlesex, 1975). P, Lorenzen, Ed., Konstruktive Wissenschaftstheorie (Suhrkamp, Frankfurt, 1974). G. Minati, E. Pessa, Collective Beings (Springer, New York, 2006). B. Nicoletti, Ed., Management per l’edilizia (Dei, Rome, 1994). R. Pietroforte, E. De Angelis, F. Polverino, Eds., Construction in the XXI Century: Local and global challenges (Edizioni Scientifiche Italiane, Naples, 2006). N. Sinopoli, L’innovazione tecnica nelle costruzioni, in Sinopie, 6 (1992). N. Sinopoli, La tecnologia invisibile (Angeli, Milan, 1997).

Reductionism intended above all as the unmanageability of emergent phenomena caused by unsuitable convictions about the exhaustive nature of praxis and cognitive models centered only upon specific details and particularities.

This page intentionally left blank

SYSTEMIC OPENNESS OF THE ECONOMY AND NORMATIVE ANALYSIS

PAOLO RAMAZZOTTI Dipartimento di Istituzioni Economiche e Finanziarie, Università di Macerata via Crescimbeni 20, Macerata, Italy E-mail: [email protected]

The paper discusses economic analysis as a normative – as opposed to positive – science. Contrary to conventional economics, it argues that: the economy does not consist of markets alone; both the economy and markets are open systems. The organization of markets and other economic activities therefore depends on the interaction between the economy and the rest of society. What configuration holds in practice is a matter of public policy. In this perspective, public policy is an intrinsic part of economic analysis, not something that follows once the economy has been investigated. The paper also argues that markets have a rationale of their own. As a consequence, public policy must define – or co-determine – the appropriate economic configuration not only by acting upon the institutional setup of markets but also by identifying those sections of the economy that should be coordinated by markets and those that should resort to other economic institutions. Keywords: openness of economy, markets as open systems, public policy.

1. Introduction This paper discusses economic analysis as a normative science. Contrary to conventional economics, it argues that since the economy does not consist of markets alone and both markets and the economy as a whole are open systems, the organization of markets and other economic activities depends on the interaction between the economy and the society they are a part of. What configuration holds in practice is a matter of public policy. In this perspective, public policy is an intrinsic part of economic analysis, not something that follows once the economy has been investigated. The paper also argues that markets have a rationale of their own. As a consequence, public policy must define – or co-determine – an appropriate economic configuration not only by acting upon the institutional setup of markets but also by identifying those sections of the economy that have to be coordinated by markets and those that have to resort to other economic institutions.

149

150

P. Ramazzotti

The paper is arranged as follows. The next section argues that, even in a very stylised model of the market, some political decisions are necessary concerning factor endowments and, in more general terms, property rights. This implies that, depending on which decision is actually taken, a whole range of market configurations is possible. Section 3 argues that the choice of configuration depends on the relation between the economy and the way it is perceived and understood by people. To this end, the section focuses on the characteristics of knowledge. It stresses its irreducibility to a consistent system and how this feature may affect how people assess the economy. More specifically, the multiple facets of knowledge reinforce the possibility of a significant variety of economic setups. Section 4 contends that, how the economy is organized ultimately is a matter of public action. This implies that economics cannot be viewed other than as a normative science. Economic inquiries that neglect the role played by the policy maker either rely on a semiclosed system view of the economy or implicitly assume that the only economy to be taken into account is the status quo. Section 5 provides the conclusions. 2. Capitalist markets as open systems The conventional notion of a self-regulating market can be traced back to conventional economic theory. Walrasian general equilibrium is based on the assumption that when such “exogenous” variables as technology and preferences are given and known and when resources are assigned to economic agents, a properly functioning price system provides all the information that is required in order to allocate those resources. Since technology, preferences and endowments are believed to be independent of how the market functions, the market itself can be viewed as a semi-closed systema: although it may be altered by exogenous shocks, it is a self-regulating systemb. A (Walrasian) market is one where prices are determined by preferences, endowments and technology alone. Prices, however, reflect the assignments of property rights, which simply means that someone is assigned the right to use something independently of the possibly negative consequences that this use may have on third parties: if individual A eats her apple, individual B will not be able to eat it. The rule whereby A rather than B has a right to eating the apple – a b

See Auyang (1988) [1] for a definition of semi-closed system. This view has been criticized on a number of accounts by a great many authors (see, for instance, Boulding 1968 [2], Georgescu-Roegen 1976 [7], Kapp 1976 [9]; see also Dow 1996 [6]). It is nonetheless appropriate to reassess it in order to appreciate its relation with the Polanyian themes that are discussed below.

Systemic Openness of the Economy and Normative Analysis

151

even if B is starving and A is not – is all but natural. It is determined according to some explicit or implicit decision. The assignment of the right – the related decision – is a political, surely not an economic, issue (Schmid 1987 [21]; Bromley 1989 [3]; Samuels, Schmid 1997 [20]; Medema, Samuels 2002 [15])c. The implication of the above is twofold. First, even at the highly abstract level of a Walrasian economy, the market has a political dimension, which obviously contrasts with its claimed independence from other societal instancesd. Second, depending on how property rights are assigned, a range of possible sets of relative prices is possible. In order for the price allocation mechanism to work, a decision has to be made concerning what the interests to be defended are, i.e. what the social priorities aree. Individuals make their economic choices under path dependent circumstances that are associated to political factors. These circumstances lead to a price set which is only one out of many possible ones. It is the price set that reflects past and present social priorities as they emerge from the existing system of power. A different system of power would not constrain a given market. Subject to the profit constraint, it would simply make the market function according to different priorities. Insofar as someone is entitled to charge a price for something, someone else is obliged to pay if she wants that something. Different price sets may be viewed, therefore, as leading to the different possible payoffs of a zero-sum game. There are instances where some sets of payoffs may be deemed superior to others, however. In terms of per capita income, for instance, some distributions may be preferable than others in that they favor a higher rate of income growth. Thus, economic policy – including the assignment of property rights – need not merely reflect the balance of power among conflicting interests. It may also reflect a choice related to some notion of social welfare. The problem is how this social welfare should be defined, i.e. what metric ought to be used to assess social efficiency and the performance of the economy. The above considerations on the variety of price sets suggest that it is rather inappropriate to assess economic outcomes in terms of a price-based indicator. c

d

e

Efficiency – i.e. finding the best way to achieve some goal such as allocation or growth – is not distinct of that political decision. Reducing slack, for instance, involves a decision over the right that a worker has in taking her time when she carries out a task. Markets are characterized by other institutions, e.g. those that affect the conduct of individuals and organizations. We shall not deal with these here. “The issue is not government versus not government, but the interests to which government is to lend its protection and the change of interests to which it gives protection” (Medema, Samuels 2002 [15, p. 153]).

152

P. Ramazzotti

Any trade off would reflect the specific set of relative prices it is based on. So, in general terms, before choosing, one would have to preliminarily choose which set of relative prices is appropriate. Physical output may be just as misleading as a price-valued output: decisions concerning what to produce and how to produce it are based on relative prices, thus on those same circumstances that undermine the uniqueness and objectivity of price-based indicators. The information that a given market provides is based on the political decisions that underlie its institutions. Priorities based only on that information would be biased by the status quo. In other terms, any attempt to change a given market according to the priorities set out by that same market would be selfreferential. The choice of the priorities that the economy should pursue, therefore, requires some value judgement. A benchmark that transcends those priorities is necessary. Independently of how a specific market is arranged, however, market transactions are unlikely to occur if they do not meet the requirement of profitability. Despite differences in how a capitalist market is arranged, it is always based on the profit motive. The name of the “game” is profit. Market institutions should be distinguished, in this regard. On the one hand, the profit goal is a key institutional feature of any capitalist market. On the other, this feature needs to be qualified by a range of other institutions, which we have discussed above. These institutions are not given once and for all but depend on explicit or implicit choices as to what priorities or interests should prevail. The profit game may be either expanded to the point that it encompasses all of the economy or it may be restricted. The same political decisions that assign property rights may choose not to assign them, i.e. they may choose that some good should not be treated as a commodity: this is the case when all members of a community are entitled to medical assistance, which is eventually paid through taxes. In such a case, medical assistance is not a commodity, it is an entitlement, i.e. a right that derives from being a member of a community. Political priorities underlie not only how the market works but also its boundaries. From this perspective, reliance on a profit-centred benchmark would imply the subsumption of society to the market rather than the other way round. Summing up, institutions determine property rights and entitlements, so they involve a value judgement concerning justice. From this point of view, the real obstacles to change would seem to be related to the political establishment – e.g. whether representative democracy works properly or not. The section that follows will focus on how a benchmark that transcends profit may emerge. This

Systemic Openness of the Economy and Normative Analysis

153

will provide some insights on whether it is actually possible to separate the economic domain from the political one. 3. Knowledge as an open system The previous section pointed out that the choice of the priorities that the economy must pursue transcends the market. It has to do with the relation between the society and the economy as well as with the role of the market within the economy. It therefore has to do with what the society values. Contrary to conventional economic theory, individuals are not able perfectly to process all the information required nor is that information generally available. This means that, whether they have to assess a good they wish to consume or the general performance of the economy, they must avail themselves of some assessment criterion. This requires knowledge. The definition of knowledge is definitely controversialf. Drawing on Loasby (1991, 1999, 2005) [10-12], I refer to knowledge as a set of connections – a pattern of relationships – among concepts that is required to make sense of (sections of) realityg. Since nobody can take everything into account at the same time, it is part of the learning process to select what is supposed to be relevant, i.e. to trace boundaries between what needs further inquiry and what has to be discarded. How to do this depends on the goals and the aspiration level of the learning actor (Simon 1976) [25]. An aspiration level reflects individual idiosyncrasies as well as the cultural environment of the learning actor, i.e. the range of shared beliefs, interpretative frameworks and learning procedures that other actors in that environment accept. It ultimately is a value judgement concerning relevance. Although everything is connected to everything else, so that one might conceive of a unique learning environment, in practice actors must adapt to their limited cognitive abilities by learning within specific sub-environmentsh: family, school, religious congregation, workplace, trade union, etc.. Specific knowledge f

g

h

The variety of approaches to the topic emerges in a recent “Symposium On Information And Knowledge In Economics” in the April 2005 issue of the Econ Journal Watch. A discussion of knowledge and public policy is in Rooney et al. (2003) [19]. The “A specific report can provide information only if it can be connected to something else, and it is unlikely to provide much information unless this ‘something else’ is a pattern of relationships—how some things fit together. Such patterns constitute what I call knowledge. Knowledge is a set of connections; information is a single element which becomes information only if it can be linked into such a set.” (Loasby 2005 [12, p. 57]). These environments are the subsystems of what Simon (1981) [26] referred to as a semidecomposable system.

154

P. Ramazzotti

depends on the specific problems that arise in each environment and, possibly, in those that are contiguous to iti. How those problems are framed – i.e. how problem solving activities are carried out – depends on the requirements and the priorities that arise within those environments: how you describe a brain depends on whether you are a butcher, an anatomopathologist, etc. (Delorme 1997, 2001 [4,5]). Market-related learning is constrained by profit in that it would be useless for a businessman – in his capacity as a businessman – to learn something that does not generate an economic gain. Obviously, he may wish to read Shakespeare independently of business considerations – possibly to make sense of life – but, in so doing, he will be pursuing a different type of knowledge, which is unrelated to profit and presumably unconstrained other than by his background knowledge and by the characteristics of the learning process itselfj: it could be associated to what Veblen referred to as idle curiosity. In his attempt to make sense of life, an actor may distinguish preferences, which are associated to egoistic goalsk, from commitments, which are associated to non-egoistic goals – be they those of another individual, of another group or of an entire communityl – or simply to ethical rules. What is important about this distinction is that there may be no common denominator between the two domains. As long as preferences and commitments do not interfere with each other, there may be no problem. Indeed, they may produce positive feedbacks, as may be the case when actors rely on non-egoistic rules in order to find a solution to the Prisoner’s Dilemma. When the two domains do interfere, the actor may face a conflict which is much like a moral dilemmam. An example might be an individual who carries out a specific economic activity – e.g. the production of armaments – that clashes with her ethical values – the non-

i j

k

l

m

The importance of contiguity is stressed by Nooteboom (1999) [16]. See the distinction that M. Polanyi (1962) [18] provides of different learning processes and of how they can be more or less restricted by the bounds that characterize them. Preferences may also include sympathy, which occurs when A’s well being depends on B’s well being. See Sen (1982) [23]. “Non-egoistic reasons for choosing an action may be based on ‘the possibility of altruism’. They can also be based on specific loyalties or perceived obligations, related to, say, family ties, class relations, caste solidarity, communal demands, religious values, or political commitment.” (Sen 1986 [24, p. 344]). The typical example of a moral dilemma is when Agamemnon was forced to choose between losing his army or losing his daughter. A similar concept in psychology is cognitive dissonance, which arises when an individual is unable to cope with information that is inconsistent with her strongly held views.

Systemic Openness of the Economy and Normative Analysis

155

acceptance of military conflicts as the solution to international or domestic disputes. Owing to bounded rationality, knowledge may well involve coexisting yet potentially inconsistent views of reality. In a capitalist economy, the profit motive provides the rationale for most economic transactions. It therefore underlies views of how to conduct business and of economic activity in general. These views may turn out to be inconsistent with what actors view as appropriate from other perspectives, e.g. ethical, religious, etc.n. Preferences and values associated to potentially inconsistent domains may coexist for quite a long time, without interfering with each other. Consequently, each domain is likely to lead to the insurgence of domain-specific institutions. Markets may therefore coexist with institutions that transcend them: clubs, churches, political parties, etc.. Institutional setups are not merely instrumental to the solution of specific problems. Once they are established they easily become a part of the reality that actors take for granted: they are internalized. This cognitive dimension of institutions (Zucker 1991 [29]; Scott 1995 [22]) suggests that it may not be easy to conceive of their dismantling as well as to envisage an alternative. One implication of the above discussion is that there is no single “game” being played, and quite a few sets of rules may coexist, interact and sometimes clash. The institutional setup that underlies the market may or may not be consistent with ethical values, as well as with institutions that are associated to those values. Thus, sweatshops may be consistent with profit and its related institutions – e.g. firms, stock markets, etc. – but may be deemed unacceptable from a range of ethical perspectives and inconsistent with the rules associated to their related – e.g. religious, human rights and political – institutions. From a policy perspective, this suggests that making sense of, and somehow dealing with, these inconsistencies should be a priority. Thus, although complexity requires that we devise simulation models that take account of multiple interactions and non linearities (Louie, Carley 2007 [13]; Law, Kelton 1999 [14]), so as to appreciate the dynamics of a system, the key point of the paper is that these should not be viewed as ever more sophisticated deterministic models. Inconsistencies involve choices and degrees of freedom within the models. They stress that the distinction between positive and normative economics is generally misleading. n

Ethics exists in business as well as in other domains of life. It is part of the market-related institutional setup. However, when I refer to ethics, in what follows, I refer to values that are independent of market-related activities.

156

P. Ramazzotti

A second implication is that knowledge about one’s reality may consist in a set of autonomous sub-systems but the boundaries between these sub-systems are never given once and for all. They may be reassessed. So, although actors may accept a distinction between the economic and other domains, thereby adapting to, and situating themselves within, a given economic and societal setup, they may also valuate those very setups and desire to change them. This involves confronting the profit motive with other values. More specifically, it involves confronting alternative ways that society can resort to in order to materially reproduce itselfo. The conclusion the above discussion leads to is that, although economic performance is usually assessed in terms of how profit-centered transactions within markets provide for the allocation of resources, for the rate of growth, or for accumulation, this type of assessment may be misleading for two reasons. First, societal values that clash with market-related values may eventually undermine social cohesion, thereby disrupting economic, as well as social, relations. Second, in so far as the economy is a sub-system of society, its performance should not be expected to prevail a priori over society’s overall performance, however assessed. At the very least they should be on the same standing. More generally, one would expect that the economy’s performance should be assessed in terms of society’s value system rather than in terms of its own criteria. Taking account of societal dimensions such as justice and carep, however, may be problematic, as the next section will argue. 4. Systemic openness and public policy Section 2 argued that political choices determine rights, which – together with other institutions – determine the structure of prices and the composition and amount of output and investment. The resulting institutional structure acts upon the choice sets of economic actors. This is the context where actors learn about the economy and learn to interact in compliance with the constraints that the extant market provides. As the discussion of knowledge in the previous section argued, however, learning actors generally transcend the economy and pursue a knowledge that is independent of market constraints. The interaction between the societal value system that this knowledge leads to and the economy allows for a great variety of potential economic and societal setups. Which one occurs o

p

Polanyi (1957) [17] specifically refers to contracted exchange, redistribution and reciprocity as the three available alternatives. Van Staveren (2001) [28] links the value domains of care and justice to giving – which corresponds to Polanyi’s notion of reciprocity – and distribution.

Systemic Openness of the Economy and Normative Analysis

157

in practice depends on the values that eventually prevail, either explicitly or implicitly. The above discussion on how markets are structured stressed that the very assignment of property rights affects distribution. It is therefore reasonable that different stakeholders within the economy will try to defend their vested interests or shift the balance of economic power to their advantage. Any economic analysis that acknowledges the openness of a market economy must take into account how different interests may affect the overall performance of the economy. The assumption that the economy should not be interfered with is tantamount to implicitly accepting the balance of economic power that is determined by the status quo. A more active policy generally changes such a balance. An appropriate policy procedure would require the explicit formulation of choices. Any policy reasonably has to decide what weight it must assign to each type of economic activity, thus what boundaries there should be between the market, the welfare state and a broadly defined non-profit sector (including families). A strictly interrelated task is to define the characteristics of these economic sectors, thus how they are expected to interact among each other. It is not enough, however, to argue in favor of principles such as redistribution and reciprocity as if they were alternatives to the market. Depending on a variety of circumstances, they may be either alternative or complementary. The relation between market and non-market values and priorities may vary. In some instances, it may be a positive one. Thus, a rise in employment may be functional to a rise in output, quite independently of any value judgement in favor of full employment per se, and redistribution and reciprocity may be functional to the profitability of the market independently of any reference to social justice or care. Post Keynesian economics, for instance, has stressed the relation between distribution and growth, especially emphasizing that a more balanced distribution of income generally has a positive effect on the level of income itself; welfare provisions such as schooling and public health typically create positive externalities that favor economic growth; as for reciprocity, while charities prevent the effects of the market economy from occurring in their most dramatic form, they also help the market in that they prevent social unrest, which would undermine economic activity. On the other hand, distribution may also affect profits – thus investment decisions and growth – negatively, as Kalecki (1943) [8] pointed out. Restrictions on polluting industries may reduce profitability. Under some circumstances – which mainstream economic thought tends to view as permanent – public expenditure

158

P. Ramazzotti

may displace private investment. Any help to the poor may be claimed to provide a disincentive to work, as the defenders of workfare policies contend. The three forms of integration and their related institutions may, therefore, be mutually consistent or inconsistent. What is more, inconsistency may exert its negative consequences even when it does not occur. In a capitalist market economy beliefs affect economic decisions, especially investment, in a significant way. So it may suffice for business just to believe that accumulation is precluded for that expectation to fulfil itself. Uncertainty may cause economic disruption quite independently of action by business to defend its vested interests. Change in how markets are structured may affect those perceptions, leading to reactions that range from uncertainty to cognitive dissonance. The implication is that, although the priorities underlying the inception of change may be generally accepted, the process may lead to the perception of unexpected and unwanted institutional inconsistencies. This is a critical issue in the light of the question: who is to choose? Owing to complexity, actors may change their mind in the process. The bandwagon effects of uncertainty may reinforce these effects. Policy must co-evolve with the parties involved. It must induce institutional change but unless it allows actors to change their perception of the economy as institutions change, it may determine two types of negative reactions. First, change may not be perceived, so that actors continue behaving as if nothing happened: for instance, provisions in favor of the weaker sections of society (e.g. consulting rooms provided by the welfare state) may remain underemployed, to the advantage of the market for private, costly and often inappropriate services (e.g. abortions). Second, opposition to what is perceived as institutional disruption might lead to reactions that recall luddist opposition to technological change. While a utility maximizer who learns only in order to achieve her goal would not be concerned about the general effects of the policy that is being carried out – she would merely focus on (micro) adaptation – a learning actor who can abstract from specific circumstances of time and place may judge policy at the macro level and either support it or oppose it. Thus, a bi-directional relation must occur between social actors and policy makers. The former must be aware of what change is occurring and how it may impinge on their life. The latter must achieve change through consent, which involves that they must avoid actors from perceiving change exclusively in terms of social disruption but also that they must be aware of what change is most important in the eyes of the social actors.

Systemic Openness of the Economy and Normative Analysis

159

Along with the bi-directional relation between social actors and policy makers, social actors must interact with each other. Precisely because they may wish to change their economic and societal environment, inconsistencies may arise among the metrics adopted by each one. In order to overcome these inconsistencies actors must be able to carry out appropriate search processes, that is, learn – a policy implication from section 3. In doing so they must also interact with others in order to achieve a generally shared view of what the appropriate metric should be. 5. Concluding remarks Systemic openness characterizes all markets: they could never work properly if they were not embedded in a broader (institutional) environment. Markets and institutions are interrelated. Not all institutions, however, are functional to the market, because some arise within extra-economic domains and may well be inconsistent with the institutions underlying the market, as well as with the profit motive that characterizes modern capitalism. The issue society has to deal with, therefore, is how to avoid institutional inconsistency. This involves choosing what relation must exist between the market, the economy and other societal institutions. The above choice requires a view of how things are and of how they ought to be: it requires knowledge of the reality people are a part of. Knowledge, however, is also an open system: people cannot separate, once and for all, their economic lives from their ethical lives. At the same time, they cannot keep everything together because they are boundedly rational. They cannot have allencompassing and consistent views of what is appropriate. Quite to the contrary, inconsistencies may occur within individuals as well as among them. A priori there is no reason to believe that economic constraints, which are not technically neutral but ultimately depend on discretionary decisions, should prevail over other societal requirements. Similarly, there is no reason to believe that the status quo is preferable to other situations. Economic analysis must, therefore, investigate how direct the economy towards society’s ends. It must deal with normative issues. If public policy is concerned with the overall quality of life of the members of society, it must allow them to overcome the inconsistencies discussed above. It must therefore take into account that, along with the values underlying the functioning of the market, a range of different values exists, and only members of society can judge what the priorities are. But, in order to choose, these members must be provided with the preliminary requirements for free choice.

P. Ramazzotti

160

The issue is not figuring out what people want but giving rise to a process that will lead people to learn how and what to choose. The process of change that such a policy determines may well lead actors to perceive new inconsistencies. Actors who initially favor one type of change may eventually change their views about what the priorities are. This is consistent with the assumption that actors are not substantively rational and that they learn as their environment evolves. It implies that the choice of priorities is an ongoing process that requires interactive learning and dialogue between policy makers and the actors involved, as well as among the latter. The general conclusion this discussion leads to is that democracy matters for normative economics. Democracy may be a means for an ongoing learning process by social actors – that eventually leads to appropriate choices – or it may be a mere counting of votes. Similarly, when institutional inconsistencies prevent governments from choosing, the solution may consist in allowing society to deal with those inconsistencies – at the risk of some social instability – or in restricting the action of minorities and dissenters. The type of action that governments take eventually affects the subsequent ability of society to actually choose the relation between its general values and economic ones as well as between the status quo and other alternatives. References 1. S.Y. Auyang, Foundations of complex-system theories (Cambridge University Press, Cambridge, 1988).

2. K.E. Boulding, in Management Science 2(3), 197-208 (1956); also published in: 3. 4. 5. 6. 7. 8. 9.

K.E. Boulding, Beyond Economics. Essays on Society, Religion and Ethics (University of Michigan Press, Ann Arbor, 1968). D.W. Bromley, Economic Interests and Institutions – The conceptual foundations of public policy (Blackwell, New York, 1989). R. Delorme, in Beyond Market and Hierarchy, Ed. A. Amin and J. Hausner, (Cheltenham, Elgar, 1997). R. Delorme, in Frontiers of Evolutionary Economics. Competition, SelfOrganization and Innovative Policy, Ed. J. Foster and J.S. Metcalfe, (Cheltenham, Elgar, 2001). S.C. Dow, The methodology of macroeconomic thought: a conceptual analysis of schools of thought in economics (Elgar, Cheltenham, 1996). N. Georgescu-Roegen, in Energy and Economic Myths. Institutional and Analytical Economic Essays, Ed. N. Georgescu-Roegen, (Pergamon Press, New York, 1976). M. Kalecki, Political Quarterly, 14, (1943). K.W. Kapp, in Economics in the Future: Towards a New Paradigm, Ed. K. Dopfer, (Macmillan, London, 1976).

Systemic Openness of the Economy and Normative Analysis

161

10. B.J. Loasby, Equilibrium and Evolution. An Exploration of Connecting Principles in Economics (Manchester University Press, Manchester, 1991).

11. B.J. Loasby, Knowledge, Institutions and Evolution in Economics (Routledge, London, 1999).

12. B.J. Loasby, Econ Journal Watch 2(1), 56-65, 2005 13. M.A. Louie, K.M. Carley, The Role of Dynamic-Network Multi-Agent Models of 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.

Socio-Political Systems in Policy, CASOS Technical Report, (2007), http://reports-archive.adm.cs.cmu.edu/anon/isri2007/CMU-ISRI-07-102.pdf . A.M. Law, D.W. Kelton Simulation Modelling and Analysis (McGraw-Hill, New York, 1999). S.G. Medema, W.J. Samuels in Economics, Governance and Law. Essays on Theory and Policy, Ed. W.J. Samuels, (Elgar, Cheltenham, 2002), pp. 151-169. B. Nooteboom, Cambridge Journal of Economics 23, 127-150 (1999). K. Polanyi, in Trade and market in the early empires: economies in history and theory, Ed. K. Polanyi et al., (The Free Press, New York, 1957), pp. 243-270. M. Polanyi, Personal Knowledge. Towards a Post-Critical Philosophy (Routledge, London, 1962). D. Rooney, et al., Public Policy in Knowledge-Based Economies. Foundations and Frameworks (Elgar, Cheltenham, 2003). W.J. Samuels, A.A. Schmid, in The Economy as a Process of Valuation, Ed. W.J. Samuels, S.G. Medema, A.A. Schmid, (Elgar, Cheltenham, 1997). A.A. Schmid, Property, Power, and Public Choice. An Inquiry into Law and Economics, 2nd edition, (Praeger, New York, 1987). W.R. Scott, Institutions and Organizations (Sage Publications, Thousand Oaks, 1995). A. Sen, in Choice, Welfare and Measurement (Basil Blackwell, Oxford, 1982). A. Sen, in Development, Democracy and The Art of Trespassing. Essays in Honor of Albert O. Hirschman, Ed. A. Foxley, M.S. McPherson, G. O'Donnel, (Indiana University of Notre Dame Press, Notre Dame, 1986), pp. 343-354. H.A. Simon, in Method and Appraisal in Economics, Ed. S.J. Latsis, (Cambridge University Press, Cambridge, 1976), pp. 129-148. H.A. Simon, in The Sciences of the Artificial (MIT Press, Cambridge, MA, 1981). H.A. Simon, A. Newell, Human Problem Solving (Prentice Hall, Englewood Cliffs, 1972). I. van Staveren, The Values of Economics. An Aristotelian Perspective (Routledge, London, 2001). L.G. Zucker, American Sociological Review 42, 726-743 (1991).

This page intentionally left blank

MOTIVATIONAL ANTECEDENTS OF INDIVIDUAL INNOVATION

PATRIZIA PICCI, ADALGISA BATTISTELLI Department of Psychology and Cultural Antropology University of Verona - Italy E-mail: [email protected]

The current work seeks to focus on the innovative work behavior and, in particular, on the stage of idea generation. An important factor that stimulates the individual to carry out the various emergent processes of change and innovation within the organization is known as intrinsic motivation, but under certain conditions, the presence of different forms of extrinsic motivation, as external regulation, introjection, identification and integration, positively influences innovative behavior at work, specifically the creative stage of the process. Starting from this evidence, the organizational environment could be capable of stimulating or indeed inhibiting potential creativity and innovation of individuals. About 100 individuals employees of a local government health department in Central Italy were given an explicit questionnaire. The results show that among external factors that effect the individual such as control, rewards and recognition for work well done, controlled motivation influences overall innovative behavior whereas autonomous motivation plays a significant role in the specific behavior of idea generation. At the same time, it must also be acknowledged that a clearly articulated task which allows an individual to identify with said task, seems to favor overall innovative behavior, whilst a task which allows a fair degree of autonomy influences the behavior of generating ideas. Keywords: innovation, antecedents of individual innovation, motivation, self determination theory, work characteristics.

1. Introduction One of the most common convictions nowadays is the understanding that in terms of innovation, it is not only the remit of the organization to be innovative but also of its individual parts. Whereas the organization can provide experts and specialists in Research and Development, it is also possible to develop and use individuals’ innovative potential, in order to respond successfully to the constant challenges of the market and social system. Betting on personnel becomes a deciding factor in competitive advantage, often reflected in a high quality service, based on the idea of continuous improvement. Currently, the focus of research on organizational citizenship behavior, employees’ creativity, positive individual initiative and on critical and reflective

163

164

P. Picci and A. Battistelli

behavior patterns which accompany people at work is the primary motivation of personnel to commit themselves to various proactive behaviors, identified as “extra-role”. It is the same general concept, according to which individuals in certain situations will do more than requested and comes under the definition of innovative work behavior (IWB), whereby individuals begin a certain course of action and intentionally introduce new behaviors in anticipation of the benefits of innovative changes (Janssen, Van De Vliert and West, 2004 [26]). For many modern organizations, public or private, competitive advantage depends on their ability to favor and program innovation, by activating and converting the ideas within the innovative process and transforming them into marketable products (Salaman and Stirey, 2002 [32]). The emerging character of innovation, when adopting the framework of Theories of Emergence (see, for instance, Crutchfield 1994 [12]; Holland 1998 [24]; Minati and Pessa 2006 [28]), appears as a sort of challenge for every theory of this kind. Namely, by its very nature, the innovation itself is unpredictable, a circumstance which seems to rule out any attempt to state the conditions granting for the occurrence of this special form of ‘intrinsic’ emergence (to use Crutchfield’s classification). Nevertheless, while a complete model of emergence of innovation is obviously unfeasible, we could still try to characterize the different psycho-social factors which seem to have some relevant influence on the development of innovation itself. We could then try to answer to questions such as: what particular factor favors individuals’ innovative behavior at work? What stimulates them to be creative or to adhere and contribute to processes of change and improvement in their specific work? By concentrating on the individual innovative process in general and on the phase of idea generation in particular, the current paper proposes to examine in what way specific motivational factors, linked to the individual and their work, can stimulate the frequency of the creation of ideas and the expression of new and better ways of doing things by individuals in the workplace. The psychological motivational construct of creativity and innovation at work as an antecedent factor has been defined on the basis of studies, carried out by Amabile (1983; 1988) [2,3], according to the distinction between intrinsic and extrinsic. Such a distinction built upon previous studies of Deci (1971) [14] and Deci and Ryan (1985) [15] and led to the Gagné and Deci (2005) [19] theory of self-determination which represents a recent attempt to put extrinsic motivation (autonomous vs controlled) into operation. Furthermore, by

Motivational Antecedents of Individual Innovation

165

continuing to follow the line taken by Amabile, various characteristics of the organizational/work environment which can stimulate or inhibit the expression of potential individual creativity, are described. Among the abovementioned characteristics, the important role of the various elements of the task (Hackman and Oldham, 1980 [21]), will be examined, highlighting in particular possible influences on the creative phase of the individual innovative process. 2. Motivational factors which influence the emergence of individual innovation and idea generation. Individual innovation or role, understood as “the intentional introduction within the role of new and useful ideas, processes, products and procedures” (Farr and Ford, 1990 [18, p. 63]), is that type of innovative behavior which the individual puts into practice to improve the quality of their work. A recent widespread current of thought considers individual innovation as a complex process made up of phases, often defined in different ways, which can be essentially divided into two distinct behaviors, namely, the generation and implementation of ideas (Rank, Pace and Frese, 2004 [29]). The precise moment of idea generation is the initial phase of innovative wok behavior. It is considered to be the most relevant factor in creativity, that is to say the most closely linked, to “the production of new and useful ideas” (Scott and Bruce, 1994 [31, p.581]), in which it is principally the individual, who acts according to their interpretation of internal environmental factors. The abovementioned phase, which is characterized by “subjectivity”, differentiates itself from other phases of innovative behavior at work which are more “social” (idea promotion and idea realization), that is to say, those which give space for necessary moments of interaction between individuals. Given its fundamental importance for the development of emerging innovation, idea generation is among the most discussed tasks of individual innovative behavior. These innovations may be born both spontaneously and intentionally from individuals and groups at work, with the aim of making the work at hand better, simpler and more efficient (among the many papers devoted to a theory of this subject we will quote West, 1990 [34]; McGahan, 2000 [27]; Alkemade et al., 2007 [1]; Cagan, 2007 [11]). One specific study, aimed at researching individual characteristics which stimulate or inhibit creativity, as expressed by employees towards their work, is that of Amabile and Gryskiewicz (1987) [8]. They analyzed individual performance within a problematical situation in the workplace. Among the qualities of the problem solver, which favor

166

P. Picci and A. Battistelli

creativity, not only do a number of typical personality traits, such as persistence, curiosity, energy and intellectual honesty, emotional involvement in the work per se and willingness to accept a challenge emerge but also the possession of fundamental cognitive abilities for certain sectors, and finally characteristics more closely linked to the particular situation. These abilities include being part of a team with dependable intellectual and social qualities and showing good social, political and relational skills. Amabile (1988) [3] proposes a componential model, in terms of psychosocial creativity, in which three necessary components/elements for good creative performance are described, namely, domain-relevant skills for the task, creativity-relevant skills and intrinsic task motivation. Intrinsic motivation involves people carrying out an activity in and of itself, given that they find the activity interesting and that it gives a spontaneous satisfaction. Extrinsic motivation, on the other hand, requires a combination of activity and certain consequences thereof, in such a way that satisfaction does not originate in the activity per se but rather in the consequences from which they derive (Porter and Lawler, 1968 [30]). A number of authors (Woodman, Sawyer and Griffin, 1993 [35]), maintain that it would be preferable to avoid the development of extrinsic motivation among workers, given that it would direct the focus/attention “beyond the heuristic aspects of the creative task and towards the technical and normative aspects of performance” [35, p. 300] even if certain conditions exist in which these aspects play a favorable role in the creative carrying out of the work performance and in which it may even be necessary and desirable that the positive effects actually increase. For example, with the imposed limitations of deadlines, expectations, controls and contractual rewards, work tends to be completed on time and well. Furthermore, not only do people need financial recompense for their work but they also need positive rewards of other types, such as feedback, recognition and behavioral guidelines (Amabile, 1988 [3]). Amabile in particular dedicates a major part of his work to the study of the role of motivation in task performance, positing the hypothesis that intrinsic motivation may influence the emergence of the creative process, whereas extrinsic motivation may actually be destructive, even if at times in simpler tasks it can act in synergy with intrinsic motivation and actually increase the expression of creativity, to such an extent that high levels of performance such as innovative, emerge clearly (Amabile, 1996, 2000 [5,6]; Amabile, Barsade, Mueller and Staw, 2005 [7]).

Motivational Antecedents of Individual Innovation

167

The inclusion of intrinsic motivation as determinant in innovative behavior directs our attention towards the motivational potential of work characteristics, such as the variety of task and of competences required, the degree of significance and of perceived identity in and of itself, feedback and autonomy (Hackman and Oldham, 1980 [21]). Farr (1990) [17] confirms that compared to simplified tasks, complex tasks are more challenging and potentially encourage innovation. Hatcher, Ross and Collins (1989) [23] highlight a positive correlation between task complexity (a comprehensive measurement of autonomy, variety and feedback) and the generation of ideas phase. In order to understand the degree of commitment and level of motivation that an individual has with regards their work, it is therefore also useful at this point to consider the nature of the task, which as previously noted, is strongly related to satisfaction with same (Hackman and Oldham, 1975 [22]). From what has already been stated, it seems patently obvious that in order to be creative in carrying out tasks at work, it is necessary to be intrinsically motivated and this only becomes possible if two fundamental conditions exist, namely that a person loves what they are doing and that their work takes place in a positively motivational context. Only when these conditions have been met, does the probability of being ready for innovation within organizations increase, thus guaranteeing creative contributions by employees, which in turn produce important benefits in the long term. If it is taken as a given that the process of innovation not only includes the development but also the implementation of creative ideas, the objective of the present paper is to consider how individual work motivation and work characteristics influence the emergence process of idea generation. It has therefore been decided to concentrate on the initial phase, which also represents the phase of maximum individual creativity in the process. 3. Intrinsic and extrinsic motivation in the idea generation process The relationship that is established between intrinsic and extrinsic motivation has caused great interest in the pertinent literature. The most prevalent psychological models proposed to date in the field, have tended to concentrate on a deep antagonism between these two forms of motivation, in that as one increases, the other decreases. Nonetheless, as the abovementioned implies, various evidence points to a more complex and clearly expressed reality. Firstly, we will look at the fact that even though the relationship between intrinsic and extrinsic motivation is always inversely proportional, we will underline how

168

P. Picci and A. Battistelli

under certain conditions, the synergic presence of these two motivational forms may actually determine positive effects on creative performance. It is useful to remember that in this regard an innovative project is made up of various phases and that whilst it may be helpful in the initial phases to propose as many ideas as possible, in the successive phases it is more important however to dwell upon those produced, examining and choosing the most appropriate (Gersick, 1988 [20]). It is generally maintained that the synergic action of extrinsic motivators is more useful in those phases, where a high level of new ideas is not required, such as the phase of collecting data or the implementation of the chosen solutions. Amabile (1994) highlights intrinsic and extrinsic motivation as relatively independent factors, rather than completely opposing poles of the same dimension. Nonetheless, certain empirical evidence shows how people simultaneously maintain a strong orientation towards intrinsic and extrinsic motivation. An interesting fact has emerged from comparing the link between motivation and creativity. Not only do reported test scores for creativity in professionally creative individuals correlate positively with the so-called “challenge” component of intrinsic motivation, but they also correlate positively with “external acknowledgement”, a component of extrinsic motivation (Amabile, 1994). The current research uses the Gagné and Deci (2005) theory of selfdetermination as a reference point. The theory works within a continuum that distinguishes between autonomous and controlled motivation. These two forms of intentional motivation, by their nature, can be differentiated from “amotivation”, which implies a complete absence of motivation on the part of the subject. Autonomy presupposes a degree of willingness and of choice in the actions to be performed, e.g. “I am doing this job because I like it”, whereas controlled motivation, being partly intentional, differentiates itself by the fact that the subject acts under pressure from external factors , e.g. “I am doing this work for the money”. Intrinsic motivation is a classic example of maximizing autonomous motivation. With regards extrinsic motivation, however, the theory identifies 4 types of motivation along a continuum, from the complete absence of autonomy to its absolute presence (auto-determination). Among the abovementioned types of motivation, two belong to controlled motivation (externally controlled motivation and introjection) and two belong to autonomous motivation (identification and integration).

Motivational Antecedents of Individual Innovation

169

In order to be undertaken successfully, the activities, which may be of little personal interest, require external motivational forms, such as financial rewards, positive acknowledgements and promotions. This is a form of externally controlled motivation and is the prototype of extrinsic or controlled motivation. Other types of extrinsic motivation are related to those behaviors, values and attitudes, which have been interiorized in people at differing levels. It is possible to distinguish between three fundamental processes of interiorizing: introjection, identification and integration, which are differentiated by the degree of autonomy characterizing them. Introjection is the process, by which a value or a behavior is adopted by an individual but is not fully accepted or lived by said individual as their own. Unlike the other two types of autonomous motivation, namely identification and integration, the above type of extrinsic motivation is controlled. Identification is characterized by the fact that a behavior or a value is accepted by the subject/individual because they have judged it to be personally important and coherent with their identity and objectives. For example, if a nurse truly has the well-being of their patients at heart, they will be prepared to operate independently, undertaking their own initiatives with tasks of little interest or even with highly unpleasant ones. Ultimately, integration is a form of extrinsic motivation, characterized by a greater degree of autonomy, in which certain values and behaviors are not only tacitly accepted by the individual but also incorporated and integrated into their value system and way of life. The abovementioned motivational form, even if it shares many aspects of intrinsic motivation, is still part of extrinsic motivation, due to the fact that the person who is acting, is not interested in the activity per se but considers the activity at hand to be instrumental in reaching personal objectives “similar to” but “different from” said activity. As is stressed in the theory, people who are autonomously motivated, even if the motivation is extrinsic in nature, are potentially more inclined to introduce changes in the way they work because they constantly wish to do their work in the most efficient manner possible. It is for this reason therefore that the work itself becomes even more intrinsically motivating, without excluding however the wish on the part of the individual to have other forms of external acknowledgement, regarding the quality of their work. This datum implies that even those individuals with a more controlled motivation may potentially develop creative ideas while carrying out their work, by simply introjecting a value which does not belong to them, in order to adapt to their organization’s wishes to ultimately obtain a certain form of positive acknowledgement or to avoid other forms of disapproval.

170

P. Picci and A. Battistelli

Thus, in line with the salient aspects of the self-determination motivational theory, as reexamined by Gagné and Deci (2005) [19], particularly with regard to an as yet unstudied possible relationship to innovative behavior, it appears pertinent at this juncture to hypothesize that not only autonomous motivation (due to its proximity to intrinsic motivation) but also controlled motivation (externally regulated at differing levels), may have a positive influence on the behavior of idea generation. Let us therefore look at the following hypothesis: H1: Both autonomous motivation (in the form of identification and integration) and controlled motivation (in the form of external regulation and introjection) positively influence idea generation behavior. 4. Work characteristics influencing the emergence of the innovation process Undoubtedly, the motivation that a subject shows towards their work depends not only on individual personality but various studies have also shown the crucial role that the work environment has in stimulating creative capacity. In other words, external background and other factors stimulate the individual per se and condition their creative and innovative capacity. For example, the resources includes all those aspects which the organization places at the individual’s disposition, until such time as it becomes practical for them to effect a creative performance. This allows sufficient time to produce an innovative piece of work, to work with competent and prepared personnel, having the availability of funds, materials, systems and adequate processes, along with the relevant information, and to have the possibility of learning and training (Sigael and Kaemmerer, 1978 [33]; Ekvall, 1996 [16]). It has been consistently observed over time that the structural organization of work has direct effects on creative performance. The more complex the task, the more motivated, satisfied and productive the individuals become (Cummings and Oldham, 1997 [13]). A greater complexity of task should stimulate a higher creative potential, to the extent that a higher degree of responsibility and autonomy, in the choices made by an individual, is clearly implied. In this case, we are dealing with tasks that call for the necessity of adopting various perspectives and observing the same problem from different points of view. The abovementioned perspectives are characterized by the fact that they require a high level of ability until the tasks are carried out. They enable individuals to follow through with the task from beginning to end, in such a manner that the

Motivational Antecedents of Individual Innovation

171

individual is fully aware of the meaning of their work. They provide important feedback during the execution of the task. Finally, these various perspectives have a strong impact on people’s lives, both within and outside the organization. By way of contrast, the most simple or routine tasks tend to inhibit enthusiasm and interest and consequently they do not stimulate the expression of creative potential (Scott and Bruce, 1994 [31]). Some jobs in contrast to others however, offer people a greater opportunity for innovative behavior. Hackman and Oldham (1980) [21] identified three conditions, in which people can feel motivated by their work. Firstly, they must recognize the results of their work. Secondly, they have to experience the sensation of taking responsibility for the results of their work. Finally, they must live their work as something significant and relevant to their value system. The research literature, in this regard, highlights five useful work characteristics for handling the demands of the task at hand (Hackman and Oldham, 1975, 1980 [22,21]), skill variety, task identity, task significance, autonomy and job feedback. When combined together, the five job characteristics decide the motivational potential of a working role/position. For this reason, if a job has a low motivational potential, the intrinsic motivation will be correspondingly low and the feelings of a person will no longer be positively influenced, even by a job done well. Farr (1990) confirmed that “complex” jobs, when compared to simpler ones, are more challenging and require more thought and that consequently they tend to promote innovation. Those studies that follow this hypothesis generally confirm the existence of a relationship between the job characteristics and further confirm the creative phase of innovation, known as idea suggestion (Axtell, Holman, Unsworth, Wall, and Waterson, 2000 [10]). Using job complexity as a measurement, based on Job Diagnostic Survey (Hackman and Oldham, 1980 [21]), Oldham and Cummings (1996) found a positively significant correlation in creative scores, attributed to employees by their superiors, highlighting the interaction between job complexity and personality characteristics, in predicting idea generation. Overall, studies in job characteristics suggest that when individuals are committed to various tasks with high levels of controls, they have a greater propensity to find new solutions, in improving their work (Axtell et al., 2000 [10]).

172

P. Picci and A. Battistelli

It is therefore by following the objective of said work that we are in a position to outline the hypothesis mentioned below: H2: Job characteristics (task identity, task significance; feedback, autonomy and skill variety) positively influence the behavior of idea generation. Finally, in the light of such evidence and considering the predictable relationship between job characteristics and intrinsic motivation, it is quite plausible to hypothesize a synergic role for these factors, not only in the generational phase of ideas but also within the process of individual innovation. It is therefore ultimately proposed to test for the following hypothesis: H3: Autonomous and controlled motivation and job characteristics positively influence the entire behavior of individual innovation. 5. The method The application within the work environment context of said study was the Health Service, because of the changeable nature of the organization which underwent a notable number of reforms, legislated for during the 1990’s that transformed Hospitals into “Health Service Firms”. The research was carried out according to a quantitative and transversal type of methodology, through a specifically geared questionnaire. A research questionnaire was presented in a generic manner to the relevant subjects, providing them with general indications, whilst clarifying total anonymity and the final objective of the project. This was done in complete agreement with the Head of the Psychology Unit and the Head of Quality Business Centre for Health Service Firms, operating in a Region of Central Italy. 5.1. The sample The sample is made of 100 subjects currently employed in the Health Service and in the administrative area of the relative management service, situated in a region of Central Italy. 53% of the total sample are Health Service Personnel and the remaining 47% are made up of Administrative Personnel within the Hospital Service. 48% of the sample are male and the remaining 52% are female. The average age of sample participants is 44.7. The information regarding qualifications obtained by those involved in the sample, revealed a relatively diverse reality, divided as follows: 40% are in possession of a High School Certificate, 10% are in possession of a Diploma

Motivational Antecedents of Individual Innovation

173

from Professional Schools/Technical Institutes, 3% are in possession of a 3 year University Diploma, 26% are in possession of a Graduate Degree and finally 21% are in possession of a Postgraduate Qualification. The average tenure within the Health Service of those sampled is 15 years. Regarding the function served by the various subjects within their respective sectors, 22 are Directors/Managers/Referees in situ or in organizational positions. 20 are part of Management and the major part of the sample (58 subjects) declared that they belong to Ward Personnel. Finally all those involved in the study have officially spent an average of 12 years in Service. 5.2. The measure The questionnaire was composed of two sections, the first comprising of a general enquiry into personal details and the second included three scales, the function of which was to analyze innovative behavior, motivation and perceived job characteristics. The construct of innovative work behavior (IWB) by Scott and Bruce (1994) [31], revisited by Janssen (2000) [25], was to use to measure innovative work behavior. In order to observe the innovative behavior of idea generation, three specific items were used for this dimension, taken from the scale of nine items of innovative work behavior, as published by Janssen (2000).These items, based on three states of innovation, conceived three items that refer to idea generation, three that refer to idea promotion and three that refer idea realization. The response format was based on the 5 point Likert Scale, where 1= never and 5= always, and upon which the subjects indicated the level of frequency, with which they undertook innovative work behavior, e.g. “With what frequency does it happen to you to have to come up with original solutions to problems?” The measurement scale for job characteristics was taken from “Job Diagnostic Surveys” (JDS), by Hackman and Oldham (1980) [21]. The scale was composed of ten items and had already been used in previous non-published Italian Research that had tested the validity of the structure. Five dimensions were considered: task variety, identification of the subject within the task, significant, autonomy and feedback. Each dimension looked into two items of the scale. For example, “My job requires me to use a number of complex capacities at a high level.” (task variety); “My job offers me the opportunity to finish that part of the work, which I had previously started.” (task identification); “My job is not very significant or important in my life.” (task

174

P. Picci and A. Battistelli

significance); “In doing my job, I am constantly provided with a considerable degree of independence and freedom.” (autonomy) and finally “My job provides me with little or no indication upon which I may judge, whether I am doing well or badly.”,(feedback from the work itself). The subjects were requested to indicate their response on the 5 point Likert scale, where 1= absolutely false and 5= absolutely true, based on their level of agreement/disagreement with potentially descriptive aspects of their job. To observe the forms of motivation that drive individuals to perform their job, a recently constructed scale of 20 items is in the publishing and evaluation stage in Italy, based on the self-determination theory of Gagné and Deci (2005) [19]. The motivational forms considered, refer to intrinsic motivation that is completely autonomous and to various types of controlled and autonomous motivation, which may be identified along the continuum of extrinsic motivation as follows: externally regulated motivation (e.g. “I am doing this job because it allows me to have a high salary.”), introjection (e.g. “I am doing this job because the esteem in which my colleagues hold me, depends on my work.”), identification (e.g. I am doing this job because it is important to me.”), integration (e.g. “I am doing this job because it allows me to reach my goals in life.”). The subjects were asked to indicate their response on a 7 point Lickert scale, where 1= absolutely false and 7= absolutely true, based on their level of agreement with the example of motivation described in the items. 6. The results Among the measures used, Table 1 summarizes the average, the deviation standard and the reliability test (Alpha Cronbach). When compared to the motivational scale of Gagné (2005) [19], it emerged from an explorative analysis of the results that the original four-dimensional structure, as hypothesized by the author (controlled external motivation, introjection, identification and integration) never appeared in the data obtained by the sample used in this research. In fact, the resulting three-dimensional structure consists of the following: external regulated motivation (M=2.67; DS=1.12), introjection (M=2.58; DS=1.32) and identification/integration (M=4.14; DS=1.31). This last dimension incorporates two motivational types which are nearest to intrinsic motivation or to those motivational types which according to the theory, appear in the autonomous category. In order to obtain the current structure, 6 items of the 20 in the original scale were eliminated, due to a lack of saturation among the sample.

175

Motivational Antecedents of Individual Innovation Table 1. Descriptive Analyses of variables. VARIABLES Autonomy Task Variety Feedback Identification Significance Job Characteristics Innovative work behavior (IWB) IWB Idea suggestion M_Integration/Identification M_Externally Controlled Motivation M_Introjection

N 100 98 100 100 100 100 100 100 97 99 99

Range 1-5 1-5 1-5 1-5 1-5 1-5 1-5 1-5 1-7 1-7 1-7

M 3.77 3.63 3.74 3.81 3.97

D. S .93 .98 .80 .92 .77

3.26 3.18 4.14 2.67 2.58

.69 .80 1.31 1.12 1.32

.70 .88 .85 .90 .70 .69

Note: it is reported that the reliability coefficient of the global Job Characteristic scale was calculated out of a total of 10 items

In line with the central hypothesis of this research, we then proceeded with an analysis of possible specific relationships between each of the variables, hypothesized as antecedents (motivation and job characteristics) and the single phase of idea generation. Table 2 shows the results of the regressions carried out, in order to study the relationship between idea generation and motivation in their autonomous and controlled forms (H1 hypothesis): Within the sample, the results of the regression show a significantly positive influence in the dimension that covers forms of integration/identification of idea generation behavior. It is therefore possible to confirm that motivation, only in the form of identification/integration, influences the emergence of the innovative behavior of idea generation. From the abovementioned data, it clearly emerges that the H1 hypothesis cannot be confirmed in its totality, showing once again that only autonomous motivation which is the nearest to an intrinsic form, appears to be implicated to a greater degree in the creative process. In fact, this process which is based on the emergence of individual innovation, does not reveal any significant result in relation to the dimensions of controlled motivation (external regulation and introjection). Table 3, on the other hand, shows the possible relationships of influence between job characteristics, according to the job characteristics model (Hackman and Oldham, 1980 [21]), and always shows the specific phase of idea generation. As can be seen from the above Table, the behavior of idea generation gives a result of being positively influenced by two specific job characteristics, namely task variety and autonomy.

176

P. Picci and A. Battistelli Table 2. Job characteristics in the phase of idea generation. Dependent Predictors Variables R² adjusted = .110; F= 13.219; p N. 20; range 19-55 years old) and origin (11

Environment and Urban Tourism: An Emergent System in Rhetorical …

351

Table 1. Transcription notation. (.) (2.0) [overlap] ↑ ↓ Underline > faster < Rea::lly . [(laugth)] Fann(h)y = ° ?

Micro-pause Pause length in seconds Overlapping speech Rising intonation Lowering intonation Emphasis Encloses speeded up talk (Brackets) Enclose words the transcriber in unsure about (empty brackets enclosed talk that is not hearable) Elongation of the prior sound Stopping intonation Immediate latching of successive talk Comments from the transcriber Loughter within speech without any silence Equal signs indicate a “latched” relationship talk appearing within degree signs is lower in volume relative to surrounding talk Question marks signal “questioning” intonation, irrespective of grammar

natives of Cagliari; 10 natives of Sardinia; 10 Italians born outside Sardinia; 9 Europeans). The choice of non-professional interviewers was based on the intention of creating a communicative exchange in which both parties could co-construct a dialogue based on a series of exchanges very similar to a “natural” conversation. The framework was that indicated by Speer and Potter: “The interviewer, for example, structures the talk by making certain issues and identities relevant and not others. Likewise, in such contexts, the “respondent” or “interviewee” orientates to the research interview as relevant, by speaking as a generic person who is keen to offer a suitably qualified response […]. Nonetheless, we would like to emphasize that we are treating these as natural materials in the specific sense that we are not privileging the actions and orientations of the researcher, but are instead treating her as an active implicated part of what is going on […]. If the participants’ institutional status (as interviewer/interviewee) is relevant to the interaction, then, it will be oriented to” (Speer, Potter, 2000 [27, note 7, pp. 565-566]). Interviews were transcribed by 3 psychology graduates likewise unaware of the final aims of the research project - based on “Jefferson’s transcription” rules (Jefferson, 1989 [18], Table 1]). Following the Discursive Action Model (Edwards, Potter, 1993 [14]), a specific analysis was implemented to identify rhetorical devices. Each interview was initialized by specifying the order of discursive analysis (casual),

352

M. Mura

interviewer (I)/ respondent (R), interviewer/respondent gender (M/F = male/female), interviewer/respondent bracket (Y/A = young/adult: 30); interviewer/respondent extraction (C – native of Cagliari; S = native of Sardinia; I = Italian, not born in Sardinia; E = European); date of interview, date of transcription, total number of linguistic passages identified (each passage was identified by means of a progressive number referred to the whole interview). The analysis was facilitated, where relevant, by some indications about the tone of the interview based on specific comments by the interviewer and other non-verbal elements. The interpretations were supported by a series of extracts. However, to save space, only some aspects felt to be particularly significant for the purposes of this study are analyzed here. Other significant partial statements present in the interviews are reported between inverted commas within the text. 3. Results: rhetorical devices and interpretational repertoires Discourse analysis displayed a central element of the co-constructed representations of the interviewer and respondent. It was never disassociated from individual self-presentation or else self-identity (extract 1). Interviewers and respondents adopted rhetorical devices to control interaction and demonstrate competence. The most common device adopted was “footing” (Goffman, 1979 [16]; Levinson, 1988 [21]; Edwards, Potter, 1993 [14]): while reporting and constructing questions or explanations, speakers were accountable for their own actions in speaking, for the veracity of their accounts, and for the interactional consequences of those accounts. In turn, the respondent replied in such a way as to present an image of himself/herself which was distinctive and felt to be most appropriate to the relational context. The interpretational repertoire most frequently used by interviewers, and also spontaneously by respondents, was the simulation of the role of a “friend” who identifies places which “he/she likes” and frequents during leisure time. The “friend” was asked to identify a pathway or recommend an holiday, because one can get to know a city only by “wandering around it” or, better still, being “taken around” to “see something”. The “pathway” was presented in two main ways: the most expert guides (tourist professionals) started from an elevated position (Monte Urpinu), offering an overview of the city which subsequently became more defined as the tourist/friend went through the historic quarters of the city and other significant historical sites. Non-natives of the city and non-professionals linked their representation to an element considered “touristic” ‘par excellence’: the sea (in Cagliari the

Environment and Urban Tourism: An Emergent System in Rhetorical …

353

Extract 1. [XXI /F-F/Y-Y/S-C/2003/2003/490]. 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214

I: and if your friend was from another place maybe from Florence and came here to Cagliari what would you [show? ] him = R: [of Cagliari in particular? ] = I: yes [(laughs)] R: uhm good question, well the sea is too obvio::us (.3) but, come on… it’s surely the first thing (.) why?: (.) they’ve never seen anything like that ↑ (.) and not Poetto’s beach/sea and the reasons: of its being ruined ↓ ((I smile )) , but because the mo::st beautiful places are Costarei °Calapira° (.) I don’t know if you’re familiar with them (.) ((I nod)) ↓ uhm while I don’t like Costa Smeralda ↑ (.)

reference is to Poetto beach) is presented as different/more attractive in regard to other beach areas frequented by non-Sardinians or non-residents and is often compared with the other coasts of the island. This is another rhetoric device, “factual report”, utilized to describe directly perceived events or by means of graphic description and sequential narrative that imply perceptual clarity. In this case, the interpretational repertoire displayed diverse “tourist city” aspects. Another is the historic center, the Castello quarter, which they found attractive for its steep, narrow streets and alleys which offer a view of the city from above, drawing the eye down to the port and gulf. One young female respondent also identified an area by pinpointing a trendy dive frequented in the small hours by youngsters (Extract 2). Sometimes the historic quarter received a negative evaluation, if compared with better else, but it was always presented as a tourist reference. Other respondents tended to suggest a specific thing (a church, a corner and so on) in order to present a distinct self, thus tourist-recommended resources were strongly linked to individual identity. “One-of-a-kind” elements were also present in the representation of a “tourist city” and referred to the model of co-construction centering on comparison: one-of-a-kind indicates the presence of elements not to be found elsewhere, in other cities or places visited. Even a small single characteristic element is sufficient, provided it is not “available” elsewhere. An essential

354

M. Mura Extract 2. [XI/F-M/Y-Y/S-C/2003-2003/ 890]. 259 260 261 262 263 264 265 266 267 268 269 270 271

I: then what would I show them from the roundabouts/here? R: surely not the area around Castello, the centre is certainly the ugliest part of Cagliari and the fact is that here uhm: I mean that it’s not ancient it’s just old and that’s something I just hate (.) The historical centre is not beautiful according to me (.) As in the case of other cities (.) However, there’s the area of Ibarium , I don’t know if you know where [it is]= I: [I do]= TO: and it’s all long the edge: and: (.) it is like hanging from a cliff high up above the city it’s beautiful that I surely would show(.) then museums and other places there’s nothing(.) one thing is a monument… and…a a monument (.) a: church, a beautiful one that very few know is San Michele(.) and the church of San Michele (.) that I would surely show

element of factual reports were “Original small dives” which can draw holidaymakers, originally oriented towards a beach holiday, like a magnet to the city. The “proof” of a place’s attractiveness and uniqueness is based on how crowded and well photographed it is. Factual reports and accountability (footing) produced an interesting interpretative repertoire: the simulated amicable relationship induced a sort of merging of the relationship between personal, social and place identity: friendship ensured veracity and the identified “special” places represented a distinctive place identity. This was confirmed by the fact that descriptions were rhetorically organized to undermine alternative ways of being: in their special role as “friend,” speakers revealed things that “only a few know”, or else mentioned and rejected better known spots, such as the “Costa Smeralda” (Emerald Coast – Northern Sardinia), highlighting their own diversity vis à vis unidentified, but always rhetorically (albeit implicitly) present “others” (Extract 1). Moreover, if the (non-Sardinian) respondent’s statements were contradictory or negative with respect to the evaluation of the city as attractive to tourists, the interviewer “compelled” the respondent to make a precise choice or moved the discourse in the direction of a choice, towards certain specific conclusions or, again, requested indication of resources to be enhanced, making the respondent take on the role of public administrator (Extract 3). Thus the positive identity of locals (both interviewers and native respondents) appeared linked to the presence of attractive characteristics in Cagliari from the point of view of tourism. This fact was particularly evident

Environment and Urban Tourism: An Emergent System in Rhetorical …

355

Extract 3. [XXI/F-M/Y-Y/C-E/2003-2003/ 902]. 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650

I: of a person [from outside]= [ from outside or anyway of a person that lives in Cagliari but travels]= R: [ but who travels a lot]= uhm boh(.) I told you I can simply say that: (.) it can potentially be a lot and: (.) Respect to those cities that really don’t have, that which they, that is, they don’t have (.) It has the means but doesn’t use them (.) this is the only thing I can say to you (.) ↓ And: because let me repe::at in Germany you look at the banks of the lakes they’re all extremely well kept ↑ (.) Then you look at Poetto without mentioning the blunder that they have recently done with the story of widening the beach: (.) Blunder ↓ (.) mean the idea was good, but it didn’t come out right However (.) all said, in the end, they ruined something (.) and they’re continuing to do:: it because they’re building I don’t know if you’re aware of it it (.) Anyway in the area of Villasimius they’re building right down close to the sea (.) violating all possibile and imaginable laws ↑ (.)

when the person appeared strongly attached to the place, through his/her selfcategorization as a young intellectual with roots running deep in the land of origin, passionate about its culture and traditions, very often summarized in a self-declaration of “Sardinian citizenship” (Extract 4). Extreme case formulation (Pomerantz, 1986) was another rhetoric device used by the high identified: a “trip” is seen only as something which takes the traveler outside Sardinia (defined as “home”), and this “departure” implies a change of Self marked by the wish to exhibit distinctive symbols such as language (defined as a “banner”), and the desire to return home as soon as possible. Paralinguistic phenomena highlighted and rhetorically communicated the “passion” at the basis of such statements, while laughter consistently preceded, accompanied, and followed affirmations that the respondent felt to be too strong or exaggerated. This paralinguistic rhetoric device, however, made it possible to express one’s identity in the conversational field and prevented the counterpart from assuming a negative approach or giving judgments of non-normality. In this “version of own world”, travel had meaning only if directed at some form of “attractive and worthwhile” activity, opposed to the passivity of the trips of “others”, defined as “purely to see something” or “waste of time”.

356

M. Mura Extract 4. [VI/F-M/Y-A/S-C/2003-2003/ 490]. 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89

R: For me to go on a trip is when you cross the sea when you go beyond the sea And : I realise that I change the moment I cross the sea why ?: (.) because > every time I have happened to be outside Sardinia < the need of identity has been evermore acute with a growing need to speak Sardinian, , co : n (0.5) in other words it’s as if one wants (.) to bring forward some sort of: banner ((laughs)) a flag of this I think that this thing is a little pathological ((laughs and coughs )) I: Can you tell me about you travel experience? R: Sure > ah here’s another important thing I didn’t tell you < for me there are no trips (.) > this is certainly a huge distortion < ↑ for me there are no trips uhmm (0.5) without it being > doing something < ((moving of hands in the air )) that has to do with work. I have already started to travel I have been part of a folklore groups since I was : 11 years old

This behavioral alternative was represented by symbol-localities: Maldives are compared to Cagliari’s beach (Poetto) as symbols, the former representing the passive, useless trips of the “others”. It is better to stay home if one is not equipped with an approach of “discovery and comparison with a completely different reality”, better still if referred to the great European capitals and visits to museums. Hence we found enunciations of what is attractive in the city, which for all intents and purposes includes everything: food and drinking water, the impossibility of staying “away” for “long periods”, “the mistral wind” whose “voice” is a “great companion”, the sun (emphasized by the tone of voice and a pause) which – unlike in other places – peeks through even on cloudy days, the sea, etc. It became clear that the identity of a place and its representation as being of tourist interest were intrinsically entwined in the profound conviction of those questioned when respondents (non-natives of Cagliari unable to provide the information requested, or pinpoint its tourist resources, due to lack of knowledge of the city), faced with a native-born interviewer from Cagliari, enacted the “footing” rhetorical device: 1. respondents expressed regret and at times ascribed their inability to provide positive answers to shortcomings, recognizing implicitly the affective link which binds any individual to his/her place of origin;

Environment and Urban Tourism: An Emergent System in Rhetorical …

2.

357

respondents, after a fraction of a second’s silence and other paralinguistic signs, followed with the affirmation that the city cannot be defined as tourist on first impact, but becomes so once it is “known” better, and only if one has the patience to “discover things”.

There were those who, so as not to offend the sensibilities of the interviewer, shifted the emphasis from recommending a holiday in the city to recommending it as a place to live: the city is suitable for a class of inhabitants in search of peace and quiet, who hate traffic, because Cagliari is a “people-friendly” city, beautiful in the sense that the “quality of life is excellent” and it is not chaotic. Based on the conversational context, to give rhetorical force to affirmations (accountability), the respondent might also utilize different forms of identity selfcategorization, perhaps stating that he/she does not truly feel ‘a citizen’ of Cagliari, although born there, because at the time his/her parents were still living in their city of origin. Here the aim was to indicate belonging to a famous city (in this case Florence). Whereas in other sequences the respondent defined him/herself as ‘a citizen’ of Cagliari to support the validity of his/her opinion on the tourist resources of the city. Again, in some cases he/she states not always living in the city, defining him/herself as a “person who travels a lot”, to support the role of “expert” in the evaluation of places and persons, while at other times he/she affirms living there to support the statement that places and events are not sufficiently publicized if he/she has found out about them only recently. As to be expected, a constant element in the repertoires of Cagliari as a tourist city was the fact that it is only potentially so, by using factual reports and systematic vagueness rhetoric devices in order to designate it as distillation of a common wisdom. This was also expressed through affirmations that the city has many “means”, especially in comparison with other similar cities, but fails to “use” them (Extract 3). To make a city attractive to tourists, it is necessary to “publicize” and “well keep” it. Moreover, in this discourse we also found another dimension of the profound conviction that the place and the identity of its inhabitants are closely related: indeed, it was stated that the “Sardinian” has no concept of “marketing” because it invests in “product quality”, but not in “image quality”; the failure of tourism to “take wing” would appear linked to the “local mentality”. The same judgment was made about the inhabitants of Cagliari, defined as “generous” (with a low tone of voice) but not open towards tourists, “sociable”, but only if “you know how to take them”, “courteous” but tending to see tourists as “strange beings”. In this they demonstrate having the same mentality as other “Sardinians”.

358

M. Mura Extract 5. [XVI/F-M/Y-A/C-C//2003-2003/ 772]. 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246

Re: Well then can you give me a definition of tourist city? To: Mmm well a tourist city (0.4) must be (.) should be a city that is at the disposal of the tourist but without forgetting (.) the ..the.. the people that live there. Let’s take Venice it’s a great tourist city However I find that for the people living there it must be a nightmare Because they don’t own their own city = = owners at home mmm for this reason I think that a city a tourist city should have the: (0.8) should show and put its beauties in the showcase (.) its own characteristics So that on the behalf of > the administrators there should be greater care < of all its esthetical features, (.) the gardens, the facades, to the urban furniture etc.etc,… But at the same time (0.3) think about them > think about these things < for those living there and not for those coming to see them

However, this judgment was attenuated by recognizing the fact that the “Cagliaritano-Sardinians” are also the only true artificers of the city’s beauty, and moreover by highlighting the fact that their characteristics might well derive only from a personal impression. In these interactional situations the interviewer moved the attention of inhabitants to the theme of organization and repeated the question on the city’s potential. The respondent generally intervened by reaffirming his/her position but also accepting the standpoint of the interviewer: the city is defined as “beautiful”, thanks to its sea, historic center and small tourist train, but lacking in organization and “in style”, as shown respectively by the difficulty in accessing information and the “shuttered” shops. On the theme of the city’s lack of a true characteristic tourist vocation, we saw the emergence, among natives of the city, of the need to achieve compatibility between the transformation of the city as a tourist center and the well-being of its residents (Extract 5): tourism brings with it the risk of no longer being “lords and masters in one’s own home” (the negative example quoted here is “Venice”). There was an implicit contrast between the population of relatively stable residents and those who are only passing through; the “hosts” present their beauty spots/characteristics as though in a “show case”, through which they

Environment and Urban Tourism: An Emergent System in Rhetorical …

359

evidently also present themselves, and the “visitors”, indicated as “guests”, must observe and admire, but they must be neither too numerous nor too intrusive. In this case, the “footing” rhetoric device was used to support the veracity of their accounts and the interactional consequences of such. An important role in the set-up of this city “show case” was assigned to public administrators, in an attempt to ward off negative judgment on the inhabitants (because the interviewer was an inhabitant). One of the “weakest” elements of the city was in fact identified as a lack of care or nurture: public places and areas are sketchily cleaned, administrators are responsible at times for true damage to the city, witness the attempt to restore Poetto beach with the introduction of dark sand or the damage caused by private building activity in defiance of scenic and heritage protection legislation (extract 3). 4. Conclusion It would appear clear that discursive analysis is capable of bringing to the fore “in vivo”, in conversation, the co-construction of interpretative repertoires which are rhetorically expressed, that is talk in action. As Edwards and Potter (1993) [14] claim, social action attributions are discursive action to invitation refusals, blaming, and defenses. Reports and descriptions were rhetorically organized to undermine alternatives: factual reports, systematic vagueness, extreme case formulations, graphic description, “footing” are rhetoric devices for managing interest, to externalize an account by constructing certain actions as universal or normative. Also, interview discourse analysis displayed interpretative repertoires which confirmed our expectations of finding in “versions of the world” the most important people-place transition: representations. Interpretative repertoires are the result of the place and social representations, constructively formed about what is “true” and “accepted” for interlocutors: We know that social representations influence group behavior and thus the “city of Cagliari” system as a “tourist place” (Bechtel and Churchman, 2002 [6]). “Versions of the world” gave us representations of the city as a system in which tourist resources are at once objective (a beautiful beach, an important historical center, entertainment services, and so on) and subjective (what the guide likes), possibly unique and authentic (the guide should be a friend). Resources were a native cultural expression, providing a sense of pride and individuality, like the entire city, and the only difference between natives and tourists is that the latter’s attitude is driven by the “tourist gaze” (Urry, 2002 [31]). Nevertheless, city tourist resources were an important part of the selfesteem of natives’ social and place identity, whereas pleasure references

360

M. Mura

constituted the tourists’ limited social-place identity. In repertoires, the social construction of Cagliari as a tourist city is strongly linked to the identity presented. This blending means that only an artificial separation of selfpresentation and representation of the object is possible since the statements perform both functions: they strengthen the persuasive value of arguments through the presentation of a positive distinct identity, constructed rhetorically with respect to “others”, implicitly present. The representation of the city was indeed marked by the presence both of the elements unanimously agreed upon, obviously representing “stereotypes” of the general representation of an urban “tourist venue” (sea and historic center) but also by evaluations of very different type and “idiosyncratic” places, which blend the distinctiveness of those indicating them in relation to the model of “visitor”, implicit or explicit, to which reference is made. Our analysis moreover provided confirmed the attempt to establish a conceptualization of place identity as constructed in the same terms as social identity (Tajfel, 1981) and Breakwell’s process of identity theory (Bonaiuto, Twigger-Ross, Breakwell, 2004 [8]). When the “ambiguous” origin of the subject brought about variable regional self-categorizations in relation to the conversational and rhetorical activity in progress, people chose that which was more conducive to a positive self presentation and accountability. Indeed, it would seem that a profound conviction emerges of the co-essence of place and inhabitant identity in the version of the world in which the place, as an end-product of the culture (becoming technique and art) of its inhabitants, represents both successes and failings. The place cannot be spoken of in a derogatory manner in front of a native inhabitant because this is regarded as offensive (Twigger-Ross, Uzzel, 1996 [30]). The identification of “tourist” resources themselves (buildings of historical importance, lesser places of typical morphology, cultural events etc.), for visitors and residents alike would seem to provisionally support the hypothesis that in the city and for its residents the representation of tourist behavior does not differ from that of leisure time: the places and activities objectifying it were the same. In conclusion, the social construction of this tourist place in communicative relationships simultaneously represented social reality and individual identity as possessing value (self-esteem) and as distinct (distinctiveness). Discourse analysis of different “residents” (native and non-native) allows us to identify the “tourist-place system” as a molar effect of physical and architectural characteristics, and social elements which form the basis for natives and visitors in awarding the “beautiful” label.

Environment and Urban Tourism: An Emergent System in Rhetorical …

361

Moreover the interpretative repertoire, in referring to natives as “masters of the house,” provided a very significant cue for maintaining sustainable tourism: visitors want a native-friend to take them sightseeing (they look for authenticity), they want new-unique “things” to “gaze” upon and this is recognized by everyone as the goal of a typical (socio-physical) environment. A “tourist” place is an emergence of a preserved bio-social diversity place: the system of socio-physical environmental features and the native people’s sense of belongingness give “identity” to that place. References 1. I. Altman, B. Rogoff, in Handbook of environmental psychology, Vol. 1, Ed. D. Stokols, I. Altman, (Wiley, New York, 1987), pp. 1-40.

2. C. Antaki, Explaining and Arguing: The social organization of accounts (Sage, London, 1994).

3. M. Argyle, The Social Psychology of Leisure (Penguin Books, London, 1996). 4. G.J. Ashworth, in Classic Reviews in Tourism, Ed. C. Cooper, (Channel View Publications, Clevedon, 2003).

5. R. Atkinson, The life Story Interview (Sage Pubblications, London, 1998). 6. R.B. Bechtel, A. Churchman, Eds., Handbook of Environmental Psychology (Wiley, New York, 2002).

7. M. Billig, Arguing and Thinking (Cambridge University Press, Cambridge, 1996). 8. M. Bonaiuto, C. Twigger-Ross, G. Breakwell, in Psychological Theories For 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

Environmental Issues. Ed. M. Bonnes, T. Lee, M. Bonaiuto, (Ashgate, Adershort, 2003), pp. 203-233. Bonnes, M., Nenci A.M., “Ecological Psychology”, in UNESCO - Encyclopedia of Life Support System, (Unesco-Eolls, Oxford, 2002). D. Capozza, R. Brown, Social identity process (Sage, London, 2000). F. di Castri, V. Balaji, Tourism, Biodiversity and Information (Backhuys, Leiden, 2002). W. Doise, La forza delle idee. Rappresentazioni sociali e diritti umani (Il Mulino, Bologna, 2002). W. Doise, A. Clemence and F. Lorenzi Cioldi, Reprèsentations sociales et analyses des donnèes (Pug, Grenoble, 1992). D. Edwards, J. Potter, Psychological Review 1, 23-41 (1993). R.M. Farr, S. Moscovici, Eds., Social representations (Cambridge University Press, Cambridge, 1984). E. Goffman, Semiotica 25, 1-29 (1979). M. Harris, Culture, People, Nature. An Introduction to General Anthropology. (Harper & Row Publishers, New York, 1985). G. Jefferson, in Conversation. An interdisciplinary perspective, Ed. D. Roger, P. Bull, (Multilingual Matters, Clevedon, 1989), pp. 156-197. K.M. Korpela, T. Hartig, F.G. Kaiser and U. Fuhrer, Environment and Behaviour, 33(4), 572-589 (2001).

362

M. Mura

20. E. Laszlo, The Systems view of the world (George Braziller, New York, 1971). 21. S.C. Levinson, in Erving Goffman: Studies in the interactional order, Ed. P. Drew and A. Wootton, (England Polity, Cambridge, 1988), pp. 161-289.

22. R.C. Mannell, S. Isho-Ahola, Annals of Tourism Research 14, 314-331 (1987). 23. M. Mura, in Culture, Quality of life and Globalization, Ed. R.G. Mira, J. M. Sabucedo, J. Romary, (Book of Proceedings, 2002), pp. 870-871.

24. Pearce, P.L. Tourist Behaviour. Themes and Conceptual Scheme (Channel View Publications, Clevedond, 2005).

25. J. Potter, M. Wetherell, in The psychology of Social, Ed. U. Flick, (Cambridge University Press, Cambridge, 1998), pp. 138-155.

26. C. Puchta, J. Potter, British Journal of Social Psychology 41, 345-363 (2002). 27. S. Speer, J. Potter, Discourse & Society 11(4) 543-572 (2000). 28. H. Tajfel, J.C. Turner, in Psychology of intergroup Relations, II Edition, Ed. S. Worchell, W.G. Austin, (Nelson-Hall, Cicago, 1986).

29. H. Te Molder, J. Potter, Eds., Conversation and Cognition (Canbridge University Press, Cambridge, 2005).

30. C.L. Twigger-Ross, D.L. Uzzel, Journal of Environmental Psychology 16, 205-220 (1996).

31. J. Urry, The tourist gaze (Sage Publications, London, 2002). 32. L. Von Bertalanffy, General System Theory (Braziller, New York, 1968).

EMERGENCE IN ARTIFICIAL INTELLIGENCE

This page intentionally left blank

DIFFERENT APPROACHES TO SEMANTICS IN KNOWLEDGE REPRESENTATION S. DAVID, A. MONTESANTO, C. ROCCHI DEIT – Università Politecnica delle Marche {s.david|a.montesanto|c.rocchi}@univpm.it There are different approaches to model a computational system, each providing a different Semantics. We present a comparison between different approaches to Semantics with the aim of identifying which peculiarities are needed to provide a system with a uniquely interpretable Semantics. We discuss different approaches, namely Description Logics, Artificial Neural Networks, and Databases, and we identify classification (the process of building a taxonomy) as a common trait. However, in this paper we also argue that classification is not enough to provide a system with a Semantics, which emerges only when relations between classes are established and used among instances. Our contribution also analyzes additional features of the formalisms that distinguish the approaches: closed vs. open world assumption, dynamic vs. static nature, the management of knowledge, and the learning process. We particularly focus on the open/closed world assumption, providing real world modeling examples to highlight the differences, and the consequences of one choice vs. the other. We also consider an important difference: in symbolic systems the notion of Semantics is ‘declared’ by means of axioms, rules, or constraints, whereas in subsymbolic ones the notion of Semantics emerges according to the evolution of a modeling system. Keywords: description logics, artificial neural networks, databases, open world assumptions.

1. Introduction Recently there has been a growing interest in the notion of Semantics. Probably pushed forward by the development of the Semantic Web, many researchers in computer science have started investigating in the field of Semantics. Such a notion has been already widely investigated in many fields like linguistics, philosophy and logic. Following the analytic stream as conceived in (Russell, 1908 [19]) and (Frege, 1918 [7]), the concept of Semantics in philosophy evolved in formal terms until Montague provided a formalization (Montague, 1974 [15]), which is widely accepted in the field of logic-based linguistic studies. In philosophy, a less formal trend brought to the definition of Semantics in terms of ”correspondence to the world” (Wittgenstein, 1921 [23]), an approach 365

366

S. David et al.

influenced by the formal work of Tarsky about the notion of truth (Tarsky, 1944 [22]). Meanwhile, the work in cognitive psychology explored the human process of categorization and classification, which led to the development of models, inspired by formal logic, but more focused on representational issues (scripts, frames, etc.). In (Balkenius and Gärdenfors, 1991 [4]) the authors show that by developing a high-level description of the properties of neural networks it is possible to bridge the gap between the symbolic and the subsymbolic levels (Smolensky, 1993 [21]). We can see this relation by giving a different interpretation of the structure of a neural network. They highlight ‘scheme’ as the key concept for this construction. Schemata are neutral with respect to the different views of cognition and have been used in many fields (Balkenius, 1993 [5]). Moreover Balkenius uses the term scheme as a collective name for the structure used in the works of (Piaget, 1952 [17]), (Piaget and Inhelder, 1973 [16]), (Rumelhart and McClelland, 1986 [18]), and (Arbib and Hanson, 1987 [2]), including also concepts such as Frames (Minsky, 1986 [14]) and Scripts (Schank and Abelson, 1977 [20]). Nowadays Semantics is involved in many research fields: natural language processing, semantic web, knowledge representation and medical informatics. Our purpose is to analyze the concept of Semantics in different approaches adopted in the design and implementation of computational systems. We consider three approaches to domain modelling: Description Logics (DLs), Relational Databases (DBs), and Artificial Neural Networks (ANNs). DLs and DBs establish their foundation in the theory of propositional logic and its connectives: AND, OR, and NOT, with the addition of universal ("x) and existential ($x) quantification from predicate logic, while some particular type of ANNs (e.g., see McCulloch and Pitts, 1943 [13]) can express connectives of propositional logic. Research and discussion in the fields of DLs, DBs and ANNs often involves the notion of Semantics. In this paper we closely look at each approach, highlighting peculiarities and common traits. Each approach is examined in relation with a simple domain, including entities such as wines and courses, which we introduced to show commonalities and differences between the three approaches. We will first consider in Section 3.1 Description Logics, a widely known approach in the knowledge representation field, which exploits logic tools (subsumption, consistency and satisfiability) to implement reasoning over structured data. Such an approach finds its roots in logical formalisms and exploits a clear definition of Semantics, often expressed in terms of set theory.

Different Approaches to Semantics in Knowledge Representation

367

We will then analyze in Section 3.2 the relational database approach, which allows for fast retrieval of structured data via queries over tables. This approach is widely adopted in commercial applications, where data are organized in terms of tuples. Finally we will consider in Section 3.3 Artificial Neural Networks, which show a fairly different and peculiar approach to the modeling of a domain, comprising a necessary learning phase to classify input data. Our analysis has been done in terms of some main features, namely: static vs. dynamic nature, closed vs. open world assumption, management of implicit knowledge, need for a learning phase, and classification. We identify the last one as a common feature. Nevertheless we argue that classification is not enough to have a ’system with Semantics’. We believe Semantic arises when relations between classes are considered. We also highlight the different nature of Semantics in the three approaches: in DBs it is explicitly declared (via tables), in DLs it partially explicited (via axioms) and partially calculated through reasoning algorithms, and in ANNs Semantics emerges from the state transitions of a system. 2. Motivating example In order to ease the understanding of our purpose, we present an example. We describe in plain English a sample scenario, which we expect an output from (e.g., a result from a query): this scenario will later be represented in each of the three approaches, assuming the same domain and granularity. We consider how the formalization varies across different representation approaches and we finally discuss the different outcomes observing how the different features of the approaches behave and sum up to form the Semantics. As we want to keep our example simple, we neither describe nor represent in details our domain, but concentrate on the relevant aspects concerning the purpose of our work. We introduce a domain containing wines and courses, and a relation that states when a given wine is ideal with a course, because the flavor of the wine fits the taste of the course. We describe the general elements of the domain and their characteristics as follows: • Wine is a class. • Wine has an alcoholic degree. • Wine has a color. • Wine is a liquid.

368

• •

S. David et al.

Course is a class. ”ideal with” is a relation between Wine and Course.

With the term class, we refer to a set of elements that share the same characteristics. Assertions about particular elements (instances) of the domain are: • Marzemino is an instance of wine. • Marzemino alcoholic degree is 12. • Marzemino’s color is red Ruby. • Rabbit is a course. • Marzemino is ideal with rabbit. 3. Formalization of the Example In this section, for each of the approaches investigated, we introduce its formalism, then we formalize the example presented in Section 2, and finally we show how the different features of the formalisms are used. 3.1. The Example in Description Logics Description Logics are a family of languages for the representation of knowledge, which also allow reasoning over it. Before formalizing the example in Description Logics, we briefly introduce the basic concepts underlying DLs: components, syntax, semantics, and reasoning. We point the interested reader to the Description Logic Handbook (Baader et al., 2003 [3]) or to the bibliography for a deeper presentation. 3.1.1. Components of a DL system The three main components of a Description Logic System are depicted in Figure 1. A Knowledge Base (KB) is a set of assertions (also called statements or axioms) about a domain, defined by means of Classes and their Properties and Relationships; it can be described in Description Logics by means of a concept language. The axioms are organized in a TBox and in an ABox, and reasoning services provide the ability to deduce additional information from the knowledge stored in the KB. 3.1.2. TBox and ABox The TBox contains the intensional knowledge, which is general knowledge concerning the domain of discourse fixed in time and usually not subjected to

Different Approaches to Semantics in Knowledge Representation

369

Figure 1. The Components of a Description Logic System.

change. The information stored in a TBox represents Concept and Roles (i.e., properties of concepts and relationships among concepts), which are aggregates of elements in the domain and form the terminology, i.e., the vocabulary used to describe the whole domain and to assign names to complex concepts description. Concepts are defined by unary predicates and Roles by binary predicates; Concepts can be either Atomic or complex. As an example, consider the following axioms: 1. Wine Β Liquid hasColor hasAlcoholicDegree. “A wine is liquid, has a color, and has an alcoholic degree”. Liquid is an atomic concept (not defined in terms of others). Wine is a complex concept, hasColor and hasAlcoholicDegree are Roles. 2.

-

idealWith.Course, idealWith .Wine. This is subtler, but it basically says ”A wine is ideal with an animal”. It expresses the range (Course) and domain (Wine) of the role idealWith. The value after the ”.” (e.g., in idealWith.Course it would be Course) is called filler of the property. The superscript denotes the inverse property. The DL syntax indeed does only allow specifying the filler (i.e., the range) of a property and not its domain.

On the other hand, the ABox contains the extensional knowledge, which can change over time and represents assertions about individuals in the domain. Individuals are instances of concepts and roles defined in the TBox, e.g., Wine(marzemino), idealWith(marzemino, rabbit). The first assertion states that Marzemino is a specific instance of class Wine, whereas the second is an instance of the role idealWith, stating that Marzemino is ideal with rabbit.

370

S. David et al.

TBox and ABox form the KB, denoted by Σ = (T,A), where T is the TBox and A the ABox. A single statement (or axiom) denotes information contained in either the TBox or ABox and is represented by α, β. The example presented in Section 2 can be expressed in the DL formalism as follows:

Τ = {Wine Β Liquid -

hasColor

hasAlcoholicDegree, idealWith.Course,

idealWith .Wine}

Α = {hasAlcoholicDegree(marzemino, 12), hasColor(marzemino, redRuby), Course(fish), idealWith(marzemino, rabbit)} A KB also contains so-called implicit knowledge, information not explicitly stated, but that can be logically deduced by the existing statements. Implicit knowledge can be discovered with the help of reasoning services, for example in answer to a query posed by a user to the system. A reasoning engine (reasoner) also provides additional inference services: depending on the input and on the goal, the system carries out different processes. Note that it is not necessary to explicitly state neither Wine(marzemino) nor Course(rabbit), i.e., that marzemino is a wine and that rabbit is an animal, since these conclusions are automatically derived from the idealWith(marzemino, rabbit) relation. 3.1.3. Semantics of DL Languages We will denote with ∆ the domain of discourse, with × the Cartesian product of two generic sets and with the subsumption relation between Concepts or Roles. The semantics of the languages is given in terms of an interpretation, defined as a pair I = (∆ I , ·I ), where ∆I is a non-empty set called domain and the interpretation function. ·I is a mapping from every Concept to a subset of ·I, from every Role to a subset of ∆ I × ∆ I and from every Individual to an element of ∆ I. An interpretation I is a model for a Concept C if the set C I is nonempty. 3.1.4. Reasoning A Description Logic system provides many basic inference services. Here we present some of them, along with the sketch of how they are used to perform complex operations.

Different Approaches to Semantics in Knowledge Representation

1. 2. 3.

4.

371

Subsumption: decide whether a concept is more general than another one. Upon Subsumption, the process of Classification is built, which is the process that builds the hierarchy of the concepts in the TBox T. Consistency Check: decide if Σ is satisfiable, i.e., if it is coherent and admits a model. Instance Checking: the problem to decide if an instance C(a) is satisfied in every model of Σ. On Instance Checking in based the process of Retrieval or Query Answering: given a Knowledge Base Σ and a concept C, find the set of all instances C(a) in Σ. Concept Satisfiability: decide whether a concept C is satisfiable in a Knowledge Base Σ.

3.1.5. Characteristics of DL Semantics CWA/OWA. The Closed World Assumption and the Open World Assumption represent two different approaches to the evaluation of implicit knowledge in a KB. The difference in their behavior is usually clarified by a comparison between the structure of a Knowledge Base Σ = (T,A), and a Database, where the schema of the latter (i.e., tables and structure) roughly corresponds to T and its tuples correspond to A. On one side, we have a single Database instance, which represents the only possible interpretation of stored data, while on the other we have one out of all possible interpretations of Σ. Hence, while in a Data Base if an information is not explicitly stated in a tuple, it is interpreted as ”negative” or false knowledge, in a KB it is considered false only if it contradicts some other axioms in the domain or if it is explicitly stated (see Section 3.2). Static and dynamic systems. We already noted that a KB stores two types of knowledge: intensional and extensional. The former is also said timeless, since it is unlike that it be changed over time. Description Logics and KBs are indeed suitable for describing domains that can evolve over time, but only at the ABox level, i.e., with assertions on the individuals and not on the structure stored in the TBox. TBox is designed in a way that it hardly can be changed. Description Logics Systems are therefore static systems in that they cannot automatically update the TBox, as this implies the redefinition of an existing concept (see De Giacomo et al., 2006 [6]). Only the interaction of the KB designers can modify a concept. Note also, that the literature about updates w.r.t. the ABox is very limited (investigated in Liu et al., 2006 [12]), and also ontology evolution (i.e., ontology update w.r.t. the TBox) is poorly investigated (an exception is Haase and Stojanovic, 2005 [8]).

372

S. David et al.

Incomplete knowledge. Description Logics (like other Knowledge Representation formalisms) provide constructors that allow to specify what an object is not or what it can be, rather than describe what it is. For example, it is acceptable and correct to define statements like the following. • ¬Student(paul): we know that Paul is not a student, but we do not care whether he is a professor or perhaps an assistant. • hasParent(john, paul) ∨ hasParent(jane, paul). Either John or Jane is parent of Paul, but whom of them? Perhaps both? Again, we do not care who is, or who are, the parent of Paul, we only know that he has (at least) one. The reason for this behavior is in one of the purposes of the DL formalism: it needs to be able to manage information that is added while the domain is described in more details. This amounts to have the ability to deal with objects that are not fully specified. 3.2. The Example in Relational Databases To the extent of this work, we consider a Data Base (DB) as a structured pool of data. We do not go into details of DB theory, so we point the reader to (Abiteboul et al., 1995 [1]) for a comprehensive introduction to the topic. A DB relies on a scheme, which describes the ’objects’ that are held in the DB. Among different approaches to the structure of a scheme, the most commonly known and commercially adopted is the relational one, which allows defining information in terms of multi-related records (tables). Less known approaches are the hierarchical one (e.g., used in LDAP) and the network model, which allows multiple inheritance via lattice structuresa. The representational tool in relational DBs is the table, which allows representing instances and relation between instances. For example a wine can be represented as in Table 1. The first row of the Table is the definition of the wine concept. It is named ’Wine’ and has three attributes: name, color, and alcoholic degree. The second row represents a particular instance of wine, whose name is Marzemino, red ruby colored and with 12 degrees. More formally both a definition and an instance can be expressed as a tuple. For example the following is a tuple representing the wine definitionb: Wine: name, color, alcoholic degree a b

It is important to note that, for the scope of this paper, we focus on the representational approach and not on technical details or implementations like SQL-based systems. Note that attributes are written lowercase and fillers capitalized.

Different Approaches to Semantics in Knowledge Representation

373

Table 1. Wine definition. name Marzemino

Wine

color Red Ruby

alcoholic degree 12

Table 2. Definition and instantiation of IdealWith. idealWith

wine Marzemino

course Rabbit

whereas the Marzemino instance can be expressed as: Wine(Marzemino, Red Ruby, 12) In the definition of the concept it is possible to specify the type of an attribute (e.g., integer, string, etc.) and whether it is mandatory or not. Attributes’ fillers (namely column values) have to be of the same type. For example the color of a wine has to be a string, the degree has to be an integer or decimal number, etc. A table can also express a relationship. Consider Table 2: the first row defines the IdealWith relation, which can hold between a wine and a course. Assuming rabbit is an instance of course, the second row states the fact that Marzemino is an ideal wine with a rabbit course. It is important to notice that Table 2 states a relation between two tuples, the Marzemino instance and the rabbit instance. This simple DB allows making queries, for example, to retrieve the list of wines or courses, the wines with a degree greater than ten, etc. The purpose of a relational DB, especially in commercial applications, is fast querying, supported by views built during batch procedures. A view is a sort of virtual table containing tuples of some particular table column, with pointers to the row associated with the value, usually built to quickly locate tuples (rows) in a table. Views simply quicken the retrieval but the application works anyway, just slower. Views can be built on any combination of attributes, which usually are the most frequently queried. We can distinguish three main actions, which are performed on a DB: schema construction, population, and querying. The first action is the definition of the tables and of the relations between them. During this phase some semantics is involved, especially during the establishment and statement of relations. For example, the definition of the IdealWith table requires that the elements that populate it have to come from the wine and the course tables. More formally, in set theory terms, the pairs that populate the IdealWith table have to be a subset of the Cartesian product obtained by coupling tuples of the Wine

374

S. David et al.

table and tuples of the course table. The same ’semantic requirement’ will apply in the subsequent phases: population and querying. Population is the constructions of instances, tuples describing the objects that populate the database. During the population there will be a sort of type check, to control the type-correctness of the tuples populating IdealWith. The querying phase involves also the specification of properties, which introduce additional semantic operations. For example it is possible to retrieve all the wines with an alcoholic degree greater than ten. In other words, the retrieval allows specifying constraints on the values of the tuples to be searched in a specific table or in a set of tables. 3.2.1. Characteristics of DB Semantics OWA/CWA. The underlying assumption of relational DBs is that the world is closed. This means that, during the querying, if a tuple does not exist (or is void) the result is an empty set of tuples. Static/Dynamic system. Relational DBs are static systems. In this case static means that, unlike ANN approaches (see Section 3.3), there is no training phase required for the system to work. Once the schema has been declared and populated, the system is ready to be used. Incomplete knowledge. The management of incomplete knowledge is not handled by relational DBs. A DB always retrieves information that describe what an element is, so it must have certainty about the data stored. It is important to notice that the main purpose of relational DBs is fast querying of explicitly stated knowledge. A DB, by design, would allow some ‘reasoning’ on data. For example the table IdealWith, instead of being populated by hand, could be dynamically constructed by implementing rules which couple wines and courses according to their characteristics. This mechanism resembles the one adopted by rule-based systems, which dynamically apply rules to structured data (knowledge bases). Although this might look as a strong similarity between two different approaches we should focus on the foundations of them. By foundations we particularly mean the representational approach on which they rely. For example a clear difference is about constraints. In a DB, constraints are taken into account during the population phase. For example, if color is a mandatory attribute and an instance of colorless wine is inserted, the database simply rejects such statement (integrity constraints check). A DL knowledge base, in the same situation, would simply insert an instance of wine with an unknown color. This is

Different Approaches to Semantics in Knowledge Representation

375

Figure 2. A generic ANN unit.

due to the purpose of the system. In a DL-based approach incomplete data are considered anyway as a source of possible computation. The DB is instead based on the assumption that incomplete data, namely data that do not fit the schema, are to be rejected. 3.3. The Example within Artificial Neural Networks An Artificial Neural Network (ANN) could be seen as a system of programs and data structures that approximates the operation of the human brain. They involve a large number of processors operating in parallel each with its own small ”sphere of knowledge” and access to data in its local memory. Figure 2 shows how a generic neural network unit works. The action potential function P(t) defines how a single unit collects signals by summing all the excitatory and inhibitory influences acting on it. This potential affects the activation function that calculates the output value of the unit itself. In (Balkenius and Gärdenfors, 1991 [4]), an artificial neural network N is defined as a 4-tuple . S is the space of all possible states of the neural network. The dimensionality of S corresponds to the number of parameters used to describe a state of the system. F is a set of state transition functions or activation functions. C is the set of possible configurations (that is weight distributions) of the network. G is a set of learning functions that describe how the configurations develop as a result of various inputs to the network. We can identify two interacting subsystems in a neural network: that governs the fast changes in the network, i.e., the transient neural activity, and that controls the slower changes, corresponding to the whole learning in the system.

376

S. David et al.

ANNs have a distributed representation of knowledge (Kurfess, 1999 [11]): an item is not represented by a single symbol or a sequence of symbols, but by the combination of many small representational entities, often referred to as micro-features. The concept ’wine’, for example, would not be represented as a string of characters, but as an entity that has the properties ’color’, ’alcoholic degree’, and other characteristics of wine. Such representational schemata have some properties like similarity based access, fault tolerance, and quick response time. We could say that a static scheme is a stable pattern of activity in a neural network. A scheme α corresponds to a vector in the state space S. A scheme α is currently represented in a neural network with an activity vector x = . This means that xi ≥ αi for all 1 ≤ i ≤ n. Let α, β be two schemata. If α β, then β can be considered to be a more general scheme than α, and α can thus be seen as an instantiation of the scheme β. Semantics in ANN. According to (Healy and Caudell, 2006 [9]), concepts are symbolic descriptions of objects and events, observable or imagined, at any arbitrary level of generality or specificity. They are organized as a multithreaded hierarchy ordered from the most abstract to the most specific. In this context, the semantics of a neural network can be expressed as an evolving representation of a distributed system of concepts, many of them learned from data via weight’s adaptation. Usually we use definitions or constraints to build a class, which are conditions to be satisfied, or better they are features and attributes of the same classes. Members representing a specific domain compose classes. ANNs create sub-symbolic class relations strongly related to the particular domain described. These relations are embedded into the dynamical system structure. The architecture of this dynamical system is a model of the learned domain. OWA/CWA. A clear distinction between closed world assumption or open world assumption in the field of ANN is not so easy. Usually the standard models of neural networks are closed world systems. But we could evaluate the ”openness” of a neural network first considering its physical structure: for example if we need a variable number of nodes, we can apply a pruning approach that removes redundant units from a network (Wynne-Jones, 1991 [24]). On the other hand, we can use a fixed structure but change the amount of information in the training set. An example can be found in (Rumelhart and McClelland, 1986 [18]), about learning the past tenses of English verbs. It is a simple perceptron-based pattern associator interfaced with an input/output

Different Approaches to Semantics in Knowledge Representation

377

encoding/decoding network, which allows the model to associate verb stems with their past tenses using a special phoneme-representation format. Static/dynamic system (Learning and relational semantics). Learning modifies the structure of the weights in the neural network in order to maximize the number of constraints satisfied. In this way ANNs catch the constraints structure of the particular context modeled, so we can say that it has “learned” the relational semantics of that domain. This point of view shows that semantics is a kind of Gestalt that constrains data into a coherent structure. The understanding of meaning could consist of the emergence of a coherence starting from a chaotic initial state through a phase transition. Even (Balkenius and Gärdenfors, 1991 [4]) have shown that by introducing an appropriate schema concept and exploiting the higher-level features of the resonance function in a neural network it is possible to define a form of nonmonotonic inference relation. So, the ”truth” in ANNs consists of the dynamic state in which a node is active or not, that is, the truth is embedded into the knowledge state of the system. The particular dynamic system represented by a specific ANN structure is the model of the learned domain. Typically, a neural network is initially ”trained” with large amounts of data and rules about data relationship. One of the most important features of a neural network is its ability to adapt to new environments. Therefore, learning algorithms are critical for the study of neural networks. Learning is essential to most of these neural network architectures and hence the choice of the learning algorithm is a central issue in the development of an ANN. Learning implies that a processing unit can change its input/output behavior as a result of changes occurred in the environment. Since the activation functions are usually fixed when the network is constructed and the input/output vector cannot be changed, the weights corresponding to that input vector need to be adjusted in order to modify the input/output behavior. A method is thus needed, at least during a training stage, to modify weights in response to the input/output process. A number of such learning functions are available for ANN models. Learning can be either supervised, in which the network is provided with the correct answer for the output during training, or unsupervised, in which no external teacher is present. Multiple Layer Perceptron’s (MLPs) training algorithms are examples of supervised learning using the error backpropagation rule (EBP) (Rumelhart and McClelland, 1986 [18]). EBP is a gradient descent algorithm that uses input vectors and the corresponding output vectors to train a multiple layer network

378

S. David et al.

until it can approximate a given function. It was proved that MLPs, which are networks with biases, a sigmoid layer, and a linear output layer, can approximate any function with a finite number of discontinuities. Self-Organizing Maps (Kohonen, 2000 [10]) are based on non-supervised learning. The preservation of neighbourhood relations is a very useful property that has attracted a great interest, for which similar input patterns from the input space will be projected into neighbouring nodes of the output map, and conversely, nodes that are adjacent in the output map will decode similar input patterns. All self-organizing networks have been generally considered as preserving the topology of the input space, and this is the consequence of the competitive learning. However, by following recent definitions of topology preservation, not all self-organizing models have this peculiarity. Incomplete knowledge. In order to manage incomplete knowledge, logical full negation is necessary, but ANNs can only implement a logical atomic negation (e.g., see McCulloch and Pitts, 1943 [13]). For example, we can build an ANN that distinguishes between “Wine” and “not Wine”. However, in ANNs, there is no correspondent of DLs's full negation, and therefore we can not represent negation between relations among classes, for example the concept of “marzemino not idealWith fish or verdicchio not idealWith rabbit”, like in Section 3.1. So, incomplete knowledge can not be implemented in ANNs in the same way it is intended in knowledge representation. Example. If we want to built a suitable model of neural network that represent the concept IdealWith talking about wines and courses, we can choose among different neural networks architectures. We decided to use a Multilayer Perceptron, that models the relation between the different instances of wine (input data) and the different typology of course (rabbit and fish). As figure 3 shows, the input layer of the MLP consists of 3 nodes, the completely connected hidden layer consists of 2 nodes and the output layer consists of only 1 unit. The latter unit codes the different typologies of courses, by using 1 for rabbit and 0 for fish. We chose a sigmoidal activation function for the output unit in order to have a more refined classification, so the output will be not merely 0 or 1, but we can have a range of values between [0,1]. In this way we have more complete information about not only the course concept but also on the concept IdealWith. The learning algorithm of the MLP is represented by equation 1, taken from (Rumelhart and McClelland, 1986 [18]).

Different Approaches to Semantics in Knowledge Representation

379

Figure 3. Multilayer Perceptron that models the concept IdealWith.

∆ wij = −η

∂E + α ∆ wij (n − 1) ∂ wij

(1)

The minimum error (E) obtained during the training depends on a suitable learning rate (η) and momentum (α). Once completed the training, it is necessary to test its effective ability to recognize the concept IdealWith. The training set will be represented by all the instances of wine associated to a specific course. So, when we minimize the error on the association training set of wine and typology of course, we will have a model of this kind of concept embedded in the Perceptron structure. Moreover this model can associate, after the training phase, a new instance of wine (never seen before) to one of the class of course. In this way a neural network is able to fill missing information. 4. Résumé and Discussion Table 3 summarizes the results we obtained from our work. It shows the component that the formalisms share and those that differ. We argued that classification does not suffice to provide Semantics, but we claimed that it is a fundamental trait, and is the first building block to provide Semantics to a system. Relations represent the second necessary building block. They are defined among classes and then used between instances of classes. These two blocks are shared by all approaches. We want to point out the attention on the difference between a taxonomy and classification in logic-based formalisms. Taxonomy is a hierarchy of concepts, connected by means of a IS-A relation, that is, we can only express that a class is superclass of another one, or, in other terms, that the set of the

380

S. David et al. Table 3. Summary of the components that build Semantics in different approaches.

Classification Relations among classes OWA/CWA Static/Dynamic Incomplete knowledge Learning

Description Logics yes yes open static yes no

Databases yes yes closed static no no

Artificial Neural Networks yes yes both dynamic no yes

instances of a class is contained in the set of instances of the superclass. There are no possibilities to define additional relations among classes, for example if they belong to different hierarchies. Hence, we do not consider taxonomy expressive enough to provide the semantics to a system. A classification, on the other hand is the process that modifies an existing taxonomy, by adding new elements to it. In particular, we can distinguish classification at two different levels, according to (Baader et al., 2003 [3]). • Classification of concepts (at TBox level) determines subconcept/ superconcept relationships (called subsumption relationships in DL) between the concepts already present in a given taxonomy and a freshly defined concept, placing the latter in the correct place in the hierarchy, or notifying its inconsistency with the existing taxonomy. Hence, classification of concepts allows one to structure the taxonomy of a domain in the form of a subsumption hierarchy. • Classification of individuals (at ABox level) determines whether a given individual is always an instance of a certain concept (i.e., whether this instance relationship is implied by the description of the individual and the definition of the concept). It thus provides useful information on the properties of an individual. From this formal definition, it might not be straightforward to see that a DB can classify elements, since there is no relation among tables other than foreign keys or constraints. However, such relations can clearly be identified in the conceptual model of the DB itself. Moreover, as every tuple must belong to a table, we can see the process to add a tuple to a table as a classification of individuals. We have shown that, besides classification and relation among classes, there are other characteristics that are needed to build the Semantics of a system. Some of these are peculiar to a formalism, i.e., the learning phase of an artificial neural network, or are common to some of them but used differently, i.e., the

Different Approaches to Semantics in Knowledge Representation

381

assumption of a closed world or an open world, resulting in different effects when querying the data stored. 5. Conclusion and further work We have investigated and presented three popular formalisms used in Knowledge Representation that use different Semantics and we identified their commonalities (Classification and relations among classes), and several other peculiarities (closed vs. open world assumption, dynamic vs. static nature, the management of knowledge, and the learning process), used in different ways. The approaches have been chosen because they cover a wide range of combinations in the different use of their peculiarities. We showed that Semantics is based on classification and on relationships between classes, and is refined by additional peculiarities that are used differently, according to the purpose of the resulting system. At the moment, we were not able to find a definition of Semantics that can be used across different systems. Besides the commonalities shared be all formalisms, there are indeed peculiarities too different to be reconciled in a unique definition. We have also shown the difference between the notion of Semantics in symbolic systems (e.g., DLs and DBs) and in subsymbolic systems (e.g., ANNs). In the former, Semantics is described with the use of axioms, rules, or constraints, whereas in the latter, Semantics emerges according to the evolution of a modeling system. Nevertheless, we foresee two directions in which this work might be extended. On the one side, we can put additional efforts in the review of other formalisms, like datalog, lambda calculus, fuzzy logic, SQL and so on, to confirm that we identified the correct peculiarities, or if we missed some. On the other side, we have spent a lot of time in the discussion of what can sum up to define Semantics, but we have also left behind a lot of issues and ideas, that we decided to investigate in a future time. Hence, additional effort should concern a deeper investigation in the following peculiarities: • Algorithms: are they part of the semantics of a system, or are they the consequence of the semantics? • User interaction: can a user modify the Semantics of a system, or is she constrained to use it as-is? • Context: to state whether it is involved or not in Semantics we need further investigation.

382

S. David et al.

References 1. S. Abiteboul, R. Hull, and V. Vianu, Foundations of Databases (AddisonWesley, 1995). 2. M.A. Arbib and A.R. Hanson, Eds., Vision, brain, and cooperative computation (MIT Press, Cambridge, MA, USA, 1987). 3. F. Baader, D. Calvanese, D. McGuinness, D. Nardi, and P.F. PatelSchneider, Eds., Description Logic Handbook: Theory, Implementation and Applications (Cambridge University Press, 2003). 4. C. Balkenius and P. Gärdenfors, KR, 32-39 (1991). 5. C. Balkenius, in Selected readings of the Swedish conference on connectionism, Ed. M. Bodn and L. Niklasson, (1993). 6. G. De Giacomo, M. Lenzerini, A. Poggi, and R. Rosati, in AAAI, 2006. 7. G. Frege, Beiträge zur Philosophie des Deutschen Idealismus, 58-77 (1918). 8. P. Haase and L. Stojanovic, in ESWC, (2005), pp. 182-197. 9. M.J. Healy and T.P. Caudell, Axiomathes 16(1-2), 165-214 (2006). 10. T. Kohonen, Self-Organizing Maps (Springer, 2000). 11. F.J. Kurfess, Applied intelligence 11(1), 5-13 (1999). 12. H. Liu, C. Lutz, M. Milicic, and F. Wolter, KR, 46-56 (2006). 13. W.S. McCulloch and W.A. Pitts, Bull. Math. Biophys 5, 115 (1943). 14. M. Minsky, The society of mind (Simon & Schuster, New York, NY, 1986). 15. R. Montague, Formal Philosophy: Selected Papers of Richard Montague (Yale University Press, New Haven, Connecticut, 1974), (Edited, with an introduction, by R.H. Thomason). 16. J. Piaget and B. Inhelder, Memory and intelligence (Basic Books, 1973). 17. J. Piaget, The Origins of Intelligence in Children (Norton, New York, NY, 1952). 18. D.E. Rumelhart and J.L. McClelland, Parallel distributed processing (MIT Press, Cambridge, MA, USA, 1986). 19. B. Russell, American Journal of Mathematics 30, 222-262 (1908). 20. R.C. Schank and R.P. Abelson, Scripts, Plans, Goals and Understanding: an Inquiry into Human Knowledge Structures (Erlbaum, Hillsdale, NJ, 1977). 21. P. Smolensky, On the proper treatment of connectionism (1993), pp. 769799,. 22. A. Tarsky, in Philosophy and Phenomenological Research 4, 341-376 (1944). 23. Wittgenstein, Logisch-Philosophische Abhandlung, (1921). 24. M. Wynne-Jones, in 13th IMACS World Congress on Computation and Applied Mathematics, Volume 2, Ed. J.J. Vichnevetsky, R. Miller, (International Association for Mathematics and Computers in Simulation, 1991), pp. 747-750.

BIDIMENSIONAL TURING MACHINES AS GALILEAN MODELS OF HUMAN COMPUTATION†

MARCO GIUNTI Dipartimento di Scienze Pedagogiche e Filosofiche, Università di Cagliari via Is Mirrionis 1, 09123 Cagliari, Italy E-mail: [email protected]

Even though simulation models are the dominant paradigm in cognitive science, it has been argued that Galilean models might fare better on both the description and explanation of real cognitive phenomena. The main goal of this paper is to show that the actual construction of Galilean models is clearly feasible, and well suited, for a special class of cognitive phenomena, namely, those of human computation. I will argue in particular that Turing’s original formulation of the Church-Turing thesis can naturally be viewed as the core hypothesis of a new empirical theory of human computation. This theory relies on bidimensional Turing machines, a generalization of ordinary machines with one-dimensional tape to two-dimensional paper. Finally, I will suggest that this theory might become a first paradigm for a general approach to the study of cognition, an approach entirely based on Galilean models of cognitive phenomena. Keywords: Turing machines, galilean models, Church-Turing thesis.

1. Introduction Typically, a model of a cognitive phenomenon H is a dynamical system that (i) is implemented on a digital computer by means of appropriate software and (ii) allows us to produce correct simulations of the phenomenon H. Even though simulation models are the dominant paradigm in cognitive science, it has been argued (Giunti 1992, 1995, 1996, 1997, 1998a, 1998b, 2005) [7-13] that Galilean models might fare better on both the description and explanation of real cognitive phenomena. Galilean models are dynamical models of a different kind, in that they are dynamical systems with n (1 ≤ n) state components, where each component has a precise and definite empirical interpretation, as it corresponds to a measurable magnitude of the real phenomenon that the model describes. Ordinary Turing machines operate on a potentially infinite linear tape divided in adjacent squares. Bidimensional Turing machines (Dewdney 1989) † The main ideas of this paper were first presented at the seminar “Prendere Turing davvero sul serio”, Poligono delle Idee Seminar Series, Dip. di Scienze Pedagogiche e Filosofiche, Università di Cagliari, October 21, 2005.

383

384

M. Giunti

[5] work instead on a potentially infinite checkerboard, where they are capable of moving one square right or left (as ordinary Turing machines do) and, in addition, one square up or down. Bidimensional Turing machines are computationally equivalent to ordinary ones, and they are mostly known for the complex geometric patterns they can generate on the checkerboard. I will argue in this paper that Turing’s original formulation (1936, [20, sec. 9.] ) of the Church-Turing thesis can naturally be interpreted as implying (A) a detailed description of a type of cognitive phenomenon, namely, the one of human computation; ( B) the claim that, for any specific phenomenon H of human computation, there is an appropriate bidimensional Turing machine that turns out to be a Galilean model of H. I will then sketch how claim ( B) might function as the core hypothesis of a new empirical theory of human computation and, finally, I will suggest that this theory might become a first paradigm for a general approach to the study of cognition, an approach entirely based on Galilean models of cognitive phenomena.a 2. Phenomena simulation vs. Galilean models A simulation model of a real phenomenon H (Giunti 1995, 1997) [8,10] is a mathematical dynamical system that (i) is implemented on a digital computer by means of appropriate software and (ii) allows us to produce empirically correct simulations of H. A simulation is empirically correct if we are able to empirically establish that the simulating process is similar to H in some relevant respect. Which respects are to be considered relevant, and which empirical methods we may employ to establish the similarity, is usually clear in each specific case. Simulation models are the dominant paradigm in cognitive science. However, because of their design, they have severe limitations with respect to both data description and explanation. The descriptive limit concerns the correspondence between simulation data and real ones, which is not direct and intrinsic to the model, but at most indirect and extrinsic. For a simulation model does not incorporate measurable properties (magnitudes) of the real phenomenon among its basic components; in contrast, quantitative descriptions are typically obtained by matching such properties with superficial or global features of the model. The explanatory limit concerns the quality of the a

The spirit of the Galilean approach is somehow consonant with some of the ideas by Wells 1998, 2006.

Bidimensional Turing Machines as Galilean Models of Human Computation

385

explanations supported by the model. Typically, they are neither realistic nor comprehensive, as they are rather cast in a somewhat fictional and “in principle” style. This second limit, like the first one, is due to the fact that the basic components of a simulation model do not directly correspond to real aspects of the phenomenon itself. As a consequence, any explanation that is based on analysis of a model of this kind is bound to introduce a whole series of fictional characters, which do not have any real counterpart in the phenomenon. As a first approximation, we can think of a Galilean model as a dynamical system with n (1 ≤ n) state components, where each component has a precise and definite empirical interpretation, as it corresponds to a measurable magnitude of the real phenomenon that the model describes. A more precise characterization of a Galilean model presupposes a preliminary analysis of the very notion of a phenomenon. In general, a phenomenon H can be thought as a pair (F, BF) of two distinct elements. The first one, F, is a functional description of (i) an abstract type of real system ASF and (ii) a general spatio-temporal scheme CSF of its causal interactions; in particular, the functional description of the abstract system ASF specifies its structural elements (or functional parts) and their mutual relationships and organization, while the description of the causal scheme CSF specifies the initial conditions of ASF ’s evolution. The second element, BF , is the set of all concrete systems of type ASF that also satisfy the causal interaction scheme CSF ; BF is called the application domainb of the phenomenon H. For example, let He = (Fe , BFe ) be the phenomenon of the free fall of a medium size body in the vicinity of the earth (from now on, I will refer to He just as the phenomenon of free fall). In this case, the functional description Fe is as follows. The abstract type of real system ASFe has just one structural element, namely, a medium size body in the vicinity of the earth; the causal interaction scheme CSFe consists in releasing the body at an arbitrary instant, and with a vertical velocity (relative to the earth’s surface) and position whose respective values are within appropriate boundaries. BFe is then the set of all concrete medium size bodies in the vicinity of the earth that satisfy the given scheme of causal interactions.

b

Since the functional description F typically contains several idealizations, no concrete or real system RS exactly satisfies F, but it rather fits F up to a certain degree. Thus, from a formal point of view, the application domain BF of a phenomenon (F, BF) might be better described as a fuzzy set.

386

M. Giunti

Let DS = (X 1 × …× Xn, (g t )t ∈T ) be a dynamical systemc whose state space M = X 1 × …× Xn has n components Xi (1 ≤ i ≤ n, where i, n ∈ Z + = the non negative integers). An interpretation IH of DS on a phenomenon H consists in identifying each component Xi with the set of all possible values of a magnitude Mi of the phenomenon H, and the time set T with the set of all possible instants of the time T of H itself. An interpretation IH of DS on H is empirical if the time T and all the magnitudes Mi are measurable properties of the phenomenon H. A pair (DS, IH), where DS is a dynamical system with n components and IH is an interpretation of DS on H, is said to be a model of the phenomenon H. If the interpretation IH is empirical, then (DS, IH) is an empirical model of H. Such a model is said to be empirically correct if, for any i, all measurements of magnitude Mi are consistent with the corresponding values xi determined by DS. An empirically correct model of H is also called a Galilean model of H (Giunti 1995; Giunti 1997, ch. 3). A Galilean model is then any empirically correct model of some phenomenon. As an example, let us consider the following system of two ordinary differential equations dx(v)/dv = k, dy(v)/dv = x(v) , where k is a fixed real positive constant. The solutions of such equations uniquely determine the dynamical system DSe = (X×Y, (hv)v∈V), where X = Y = V = R (the real numbers) and, for any v, x, y ∈ R, hv(x, y) = (kv + x, kv2/2 + xv + y). On the other hand, let us consider again the phenomenon of free fall He, and let IHe be the following interpretation of DSe on He. The first component X of the state space of DSe is c

A dynamical system (Arnold 1977 [1]; Szlensk 1984 [19]; Giunti 1997, 2006 [10,14]) is a kind of mathematical model that formally expresses the notion of an arbitrary deterministic system, either reversible or irreversible, with discrete or continuous time or state space. Examples of discrete dynamical systems are Turing machines and cellular automata; examples of continuous dynamical systems are iterated mappings on R, and systems specified by ordinary differential equations. Let Z be the integers, Z + the non-negative integers, R the reals and R + the non-negative reals; below is the exact definition of a dynamical system. DS is a dynamical system iff there is M, T, (g t )t ∈T such that DS = (M, (g t )t ∈T ) and 1. M is a non-empty set; M represents all the possible states of the system, and it is called the state space; 2. T is either Z, Z +, R, or R +; T represents the time of the system, and it is called the time set ; 3. (g t )t ∈T is a family of functions from M to M; each function g t is called a state transition or a t-advance of the system; 4. for any t, v ∈ T, for any x ∈ M, g 0( x ) = x and g t + v (x) = gv (g t ( x ) ) .

Bidimensional Turing Machines as Galilean Models of Human Computation

387

the set of all possible values of the vertical velocity of an arbitrary free falling body, the second component Y is the set of all possible values of the vertical position of the falling body, and the time set V of DSe is the set all possible instants of physical time. Since all three of these magnitudes are measurable or detectable properties of the phenomenon of free fall He, IHe is an empirical interpretation of DSe on He, and (DSe, IHe ) is thus an empirical model of He. For an appropriate value of the constant k, such a model also turns out to be empirically correct.d Then, according to the previous definition, the pair (DSe, IHe ) is a Galilean model of He. It is quite clear that Galilean models can go well beyond the descriptive and explanatory limits of simulation models. Note first that data description in Galilean models is direct and intrinsic, for each component of a model of this kind determines the values of a specific magnitude of the corresponding phenomenon. Second, the explanations supported by a Galilean model are realistic and comprehensive, as each of its components corresponds to a specific magnitude of the intended phenomenon, and so any explanation based on an analysis of such a model cannot introduce any arbitrary or fictional character. For these reasons, anyone interested in improving the results of Cognitive Science, both on the descriptive and explanatory score, should seriously consider the prospect of constructing Galilean models of cognitive phenomena.e This, however, surely is not an easy task. The main problem we face is that of focusing on a particular class of cognitive phenomena for which the construction of Galilean models be clearly feasible and well suited. This, in turn, entails the availability of a sufficiently detailed description of this class, in such a way that (i) a few cognitive magnitudes relevant to the phenomena of this type be clearly identified, and (ii) a suitably detailed sketch of a specific kind of Galilean model, appropriate for this particular class of phenomena, be also given. My contention is that a quite natural interpretation of Turing thesis (1936, sec. 9.I ) indeed provides us with a detailed description of a particular class of cognitive phenomena, namely, those of human computation, and that such a description does satisfy both requirements (i) and (ii) above. I will explicitly

d

e

Quite obviously, if k = the value of the acceleration due to gravity, the model (DSe , IHe ) turns out to be empirically correct within limits of precision sufficient for many practical purposes. For a further defense of this tenet, see Giunti 1995, 1997, 2005 [8,10,13]. Eliano Pessa and Maria Pietronilla Penna pointed out to me (personal communication) that they pursued the idea of employing models of a Galilean kind in some of their work on iconic memory (Penna and Pessa 1992, 1998 [17,18]; Penna and Ciaralli 1996 [16]).

388

M. Giunti

address this issue in sec. 4. In the next section, however, we need to take a preliminary look to bidimensional Turing machines. 3. Bidimensional Turing machines An ordinary Turing machine can be thought as a device formed by a head, just one slot of internal memory, and a linear tape (external memory) infinite in both directions. The internal memory slot always contains exactly one symbol (internal state) taken from a finite alphabet Q = (q1, … , qm) with at least one element. The tape is divided in adjacent squares, where each square contains exactly one symbol taken from a second finite alphabet A = (a0, a1, … , an) with at least two elements. The first symbol a0 (the blank) is a special one, and is usually indicated by b. The blank is special in that a square containing it should in fact be thought as being empty. At any discrete instant of time (step), only a finite number of tape squares contains non-blank symbols or, in other words, the tape is always completely empty except for a finite portion. The head is always located on exactly one square of the tape (the scanned square), and it is capable of performing five basic operations: read the symbol on the scanned square, write a new symbol on such a square, move one square to the right (indicated by R), move one square to the left (L), do not move (H ). At each time step, the head performs a sequence of exactly three operations: the first is a read operation, the second a write operation, and the third a moving one. The result of the read operation is the symbol aj contained in the scanned square. However, exactly which writing and moving operations the head performs next is determined by such a symbol, by the current internal state qi and by the set of instructions of the machine. In fact, for each possible pair qi aj internal-state/ scanned-symbol, the machine has exactly one instruction (quintuple) of the form qi aj : ak Mqr , where ak indicates the symbol to be written, M (either R, L or H ) is the moving operation to be performed, and qr is the internal state the machine goes in at the next time step. It is thus clear that any ordinary Turing machine is a dynamical system whose state space has the same three components, as the future behavior of an arbitrary machine is completely determined by its set of quintuples and by the current values of the following three state variables: the head position (expressed by an integer coordinate that corresponds to the scanned square), the complete tape content (expressed by a function that assigns a symbol of the alphabet A to each tape coordinate), and the machine’s internal state. The simplest form of a bidimensional Turing machine is obtained by just replacing the linear tape with a checkerboard infinite both in the right/left

Bidimensional Turing Machines as Galilean Models of Human Computation

389

direction and in the upward/downward one. Accordingly, the head can now also move one square up (U ) and one square down (D). As before, the dynamic behavior of the machine is completely determined by its set of quintuples. The only difference is that now a generic quintuple has the form qi aj : ak Mqr , where M stands for either R, L, U, D or H . Bidimensional Turing machines of this simple kind are known for the complex geometric patterns they can generate on the checkerboard (Dewdney 1989). For our present purposes, we need bidimensional Turing machines of a somewhat more general kind. The main difference with respect to the simple ones concerns the more sophisticated syntax we allow for the machine quintuples. In fact, in the general case, an instruction is not just a simple quintuple, but a quintuple schema that typically represents a possibly infinite set of quintuples of a specific form. This is obtained by allowing four different kinds of basic symbols, and precisely: (i) constants a0, a1, … , an, … (ii) variables x0, x1, … , xn, … (iii) function terms f1, f2, …, fn, … (iv) simple internal states q1, q2, … , qn, … In addition, we also allow internal state schemata, which are obtained by concatenating a simple internal state with strings of other symbols. The exact definitions of both an internal state schema and a quintuple schema will be given below. Constants are the only symbols that can in fact be written on the checkerboard squares. As usual, the first constant a0 is the blank, and it is also indicated by b. Each variable may stand for a specified set of constants (metalinguistic variable), or for a set of numbers or other specified entities. Whenever a new variable is introduced, its variation range (i.e. the set of all its possible values) must be declared explicitly; it is permitted, but not mandatory, to also declare the initial value of the variable. The value of a constant may be explicitly declared as well, and in this case the constant is said to be bound (or interpreted). A constant that is not bound is called free. Besides constants and variables, we also allow the function terms f1, f2, …, fn, … as basic symbols of our bidimensional Turing machines. Function terms stand for functions, and whenever a function term is introduced, the corresponding functionf must be declared explicitly. Functions should be thought as auxiliary operations that can be performed as needed during the execution of a routine. Function terms, together with variables and constants, f

Any such function must be computable in the intuitive sense. A function f is computable in the intuitive sense just in case there is a mechanical procedure P that computes f, where P computes f iff, for any argument x, if P is started with initial data that correspond to x, it terminates in a finite number of steps, and its final data correspond to f (x). For an explicit characterization of a mechanical procedure, see sec. 4, par. 2.

390

M. Giunti

allow the formation of simple functional expressions. Complex functional expressions, obtained by (repeatedly) nesting function terms, are also allowed. For example, let us suppose that, for a given machine, the following basic symbols have been introduced: two numeric variables m and n and, as constants, the arabic numerals 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 with their usual meanings; in addition we also introduce two function symbols, + and ×, which respectively stand for the addition and multiplication operations on the non-negative integers. Then, the following are all simple functional expressions (m+n), (2+4), (n+7), (m×m), (m×2), (5×3), while the next two are complex ((m×m)+3), (((m×m)+3)+n). It is convenient to distinguish between functional expressions in which variables occur (open functional expressions) and those in which there is no occurrence of variables (closed functional expressions). Closed functional expressions can be thought as additional constants, while open ones as additional function terms. Functional expressions in which no free constants occur are called bound. A functional expression that is not bound is said to be free. A closed and bound functional expression always stands for some object, and it can thus be thought as an additional interpreted constant. An open and bound functional expression always stands for a function, and it can thus be thought as an additional function term together with the relative function declaration. Free functional expressions can instead be thought as either free constants or function terms for which the relative function declaration has not been given. Internal state schemata are obtained by concatenating a first string of variables, a simple internal state, and a second string of constants, variables, or bound functional expressions. Thus, the general form of an internal state schema is vqi s, where v is an arbitrary string of variables, s is an arbitrary string of either constants, variables or bound functional expressions, and qi is any simple internal state; either v or s may be empty. Note that internal state schemata include simple internal states as particular cases, for a simple internal state is an internal state schema where both strings v and s are empty. By definition, an internal state is any sequence of objects singled out by an internal state schema. The simple internal state qi occurring in internal state schema vqi s is better interpreted as a disposition of the machine to do a certain action that takes as objects the values of the constants, variables or functional expressions in the string s, and may also depend on the values of the variables in the string v. These latter variables are in fact parameters, that is, global variables of a certain

Bidimensional Turing Machines as Galilean Models of Human Computation

391

routine. More, precisely, a parameter is a variable whose value does not change during the whole routine execution, and which is available at each step of the execution itself. The instructions of a bidimensional Turing machine are quintuple schemata, where a quintuple schema is any sequence of five strings of symbols that satisfies the following conditions: (i) the first and the fifth string is an internal state schema; (ii) the second and the third strings are either constants, metalinguistic variables, closed and bound functional expressions that stand for constants, or open and bound functional expressions that stand for functions whose codomain is a set of constants; (iii) the fourth string is either H or any finite string of the four basic movement signs R, L, U, D. Such a string stands for the complex movement obtained by sequentially combining (from left to right) the basic movements expressed by each string component. A combination of n (0 ≤ n) movements of the same kind is respectively expressed by R n, Ln, D n, U n. If n = 0, then, by definition, R 0 = L0 = D 0 = U 0 = H and, if n = 1, then R1 = R, L1 = L, D1 = D, U 1 = U. More generally, we may also allow functional movement signs of the forms R e, Le, D e, U e, where e is either a closed and bound functional expression that stands for a non-negative integer, or an open and bound functional expression that stands for a function whose codomain is a set of non-negative integers; in the general case, the fourth string of a quintuple schema is then either H or any finite string of movement signs (basic or functional). Finally, by definition, a quintuple is any sequence of five objects singled out by some quintuple schema. The whole behavior of a bidimensional Turing machine is determined by its machine table or set of instructions. A machine table is any finite and consistent set of quintuple schemata, where a set of quintuple schemata is consistent iff the set of quintuples singled out by the quintuple schemata does not include any two different quintuples that begin with the same internal-state/scanned-symbol pair. A machine table is also intended to be complete, in the following sense. If the quintuple schemata of the machine table do not single out any quintuple that begins with some possible internal-state/scanned-symbol pair, then it is intended that the machine table also includes an identical quintuple that begins with any such pair. ( Where a quintuple is identical iff its initial and final internal states are identical, the scanned symbol is identical to the written one, and the movement is H .) Similar to ordinary Turing machines, bidimensional ones are dynamical systems with three state components. In this case the three instantaneous state variables are: the head position (expressed by a pair of integer coordinates that

392

M. Giunti

corresponds to the scanned square), the complete checkerboard content (expressed by a function that assigns a constant to each checkerboard coordinate pair), and the machine’s internal state. EXAMPLE 1 I give below the table of a machine that computes the function whose domain is the set of the arabic numerals {0, 1, 2, 3, 4, 5, 6, 7, 8}, whose codomain is the set of the arabic numerals {1, 2, 3, 4, 5, 6, 7, 8, 9}, and whose values are specified as follows: (0) = 1, (1) = 2, (2) = 3, (3) = 4, (4) = 5, (5) = 6, (6) = 7, (7) = 8, (8) = 9. From now on, I will refer to as the numeral successor function. The constants of this machine are the blank b and the ten arabic numerals 0, 1, 2, 3, 4, 5, 6, 7, 8, 9. In addition, this machine has just one metalinguistic variable, d, whose range is the set of numerals {1, 2, 3, 4, 5, 6, 7, 8, 9}. No function terms are allowed for this machine. Finally, its simple internal states are just two. The first one, S, can be thought as the disposition to produce the next numeral, while the second simple internal state, E, should be thought as the disposition to end the computation. The input numeral is written on a square of an otherwise blank checkerboard. Initially, the head scans the input numeral, and the internal state is S. The machine table is as follows. Table 1 (1)

S S S S S S S S S

0 1 2 3 4 5 6 7 8

: : : : : : : : :

1 2 3 4 5 6 7 8 9

H H H H H H H H H

E E E E E E E E E

Writes the successor of any Numeral between 0 and 8, remains on the same square, and calls the ending routine (2);

(2)

E

d

:

d

H

E

STOP.

EXAMPLE 2 As a second example, let us consider the table of a machine that computes the successor function s in decimal notation. This machine has the same eleven constants as the previous one. The machine has two metalinguistic variables, d, whose range is the set of numerals {1, 2, 3, 4, 5, 6, 7, 8, 9}, and c, whose range is the set of numerals {0, 1, 2, 3, 4, 5, 6, 7, 8}. This machine has just one function term, , which stands for the numeral successor function (see example 1). The simple internal states are two. The first one, S, can be thought

Bidimensional Turing Machines as Galilean Models of Human Computation

393

as the disposition to add one to an arbitrary number, while the second simple state, E, can be interpreted as the disposition to end the computation. The input number is written in decimal notation on a row of an otherwise empty checkerboard. Initially, the head scans the rightmost digit of the input number, and the machine is in internal state S. The machine table is as follows. Table 2 (1a)

S

c

:

(c)

H

E

TEST: finished; writes the successor of any numeral between 0 and 8, remains on the same square, and calls the ending routine (1ac1);

(1b)

S

9

:

2

L

S

TEST: not finished; writes 0 in place of 9, goes one square to the left and calls test (1);

(1c)

S

b

:

1

H

E

TEST: finished; writes the carried digit 1, stays on the same square, and calls the ending routine (1ac1);

(1ac1)

E

d

:

d

H

E

STOP.

EXAMPLE 3 Below is a machine that adds to an arbitrary non-negative integer n an integer d such that 1 ≤ d ≤ 9. This machine has the same eleven constants as the ones of examples 1 and 2. Its variables are three: n, d and c. The range of n is the whole set of the non-negative integers, and its initial value is the input number n; d is an integer variable such that 1 ≤ d ≤ 9, and its initial value is the other input number d. The variable c is metalinguistic. If is an arbitrary numeral, let ( ) be the number denoted by ; then, the range of c is the set of all arabic numerals such that 0 ≤ ( ) < d, and its initial value is the numeral 0. This machine has three function terms, s, and ; the first one stands for the successor function, the term stands for the numeral successor function, and the term stands for the function that, to any number d (1 ≤ d ≤ 9), associates the corresponding arabic numeral. There is just one simple internal state, S, whose intuitive meaning is the disposition to add one to an arbitrary number. The initial value of the counter c (i.e. the numeral 0) is written on a square of an otherwise blank checkerboard, and the computation starts with the head on that square. The machine table is below. Note that the output (n+d ) is given by

394

M. Giunti

the final value of the variable n, and that the variable d is in fact a parameter, for its value is constant during the whole computation, and available at each step. Table 3 (1a)

dSn

(1b)

dSn

c

(d)

:

(c)

H

dSs(n)

TEST: not finished; increases both the counter c and the output variable and then calls test (1);

:

(d)

H

dSn

TEST: finished; output is n, and thus STOPS.

It is surprising to realize that this simple machine has a quite intriguing psychological interpretation. Let us notice first that the machine only needs one external memory location, for it always stays on the same square of the checkerboard. Let us then consider the cognitive phenomenon Hf of a human being that computes the sum (n+d ) with the help of her hands. Initially, she memorizes both the number n, and the one digit number d , while both her hands are closed. She then opens one finger while she mentally adds one to the number n, and keeps repeating this pair of operations until the number of open fingers is equal to the number d. Now she stops, having in mind the result (n+d ). Let us then consider the following quite natural interpretation of the bidimensional Turing machine above on phenomenon Hf . In the first place, the machine’s discrete time steps correspond to the steps of the human being’s computation (that is to say, they correspond to the discrete time of the real Second, each internal state of the Turing machine phenomenon Hf ). corresponds to a possible content of the human being’s working memory during the computation. Third, the position of the head on the checkerboard corresponds to the spatial location attended to by the human being while computing (that is to say, the location of her hands). And finally, the checkerboard possible contents (i.e. the numerals 0, 1, 2, 3, 4, 5, 6, 7, 8) correspond to the possible configurations of her hands (closed, with one finger open, with two fingers, etc.). EXAMPLE 4 This machine writes from left to right in decimal notation an arbitrary non-negative integer n and then stops. The machine has the same eleven constants as the ones of the previous examples (see example 1). The variables are two: n and s. The range of n is the whole set of the non-negative integers, and its initial value is the number n to be written. The range of the

Bidimensional Turing Machines as Galilean Models of Human Computation

395

second variable s is the set of all strings of decimal digits, that is, all strings made up of the ten arabic numerals 0, 1, 2, 3, 4, 5, 6, 7, 8, 9. This machine has two function terms: h and t. The term h (short for head) stands for the function that takes as argument either an arbitrary non-negative integer n or a string of decimal digits c and, respectively, returns as value the leftmost digit of the decimal representation of n or the leftmost digit of c. The term t (short for tail) stands for the function that takes as argument either an arbitrary non-negative integer n or a string of decimal digits c and, respectively, returns as value the string obtained by deleting the leftmost digit of the decimal representation of n or the string obtained by deleting the leftmost digit of c; if either the decimal representation of n or c has just one digit, then, respectively, t(n) = 0 or t(c) = 0. There are two simple internal states, E and ??. E is the disposition to write the leftmost digit of the input number; while ?? is the disposition to check whether the writing of such number has been completed. The computation begins in simple internal state E, on an arbitrary square of an empty checkerboard. As said, the initial value of n is the number n to be written. The machine table is below. Table 4 (1)

En

b

:

h(n)

R

??t(n)

Writes the leftmost digit of number n, goes to the right, keeps the tail of n, and calls test (2);

(2a)

??s

b

:

h(s)

R

??t(s)

TEST2: not finished; writes the leftmost digit of the kept tail, goes to the right keeps the new tail, and calls test (2);

(2b)

??0

b

:

b

H

??0

TEST2: finished; STOP.

EXAMPLE 5 I give below (see table 5) a bidimensional Turing machine that computes the sum of an arbitrary number of addends in decimal notation, by implementing a procedure very similar to the well known column based rule. Each addend is written in a different row of the checkerboard, with its rightmost digit in a specified column (that is, the addends are justified to the right). Immediately above the uppermost addend and immediately below the downmost one there are two horizontal lines, as long as the longest addend; the two

396

M. Giunti

horizontal lines are justified to the right as well. The result will be written below the lower line, justified to the right. This machine has twelve constants; they are the eleven constants of the previous machines (see example 1) and, in addition, the horizontal line sign –. The variables are three: n, d, and s. The range of the variable n is the whole set of the non-negative integers, and its initial value is zero; d is a metalinguistic variable whose range is the set A of the nine arabic numerals 1, 2, 3, 4, 5, 6, 7, 8, 9; s is a variable whose range is the set of all strings of decimal digits, that is to say, all strings made up of the ten arabic numerals 0, 1, 2, 3, 4, 5, 6, 7, 8, 9. This machine has seven function terms: , +, u, r, l, h, t. The term stands for the function that, to each numeral in the set A above, associates the number denoted by that numeral. The function term + stands for the addition of an arbitrary non-negative integer n and an integer d such 1 ≤ d ≤ 9 (see example 3). The term u stands for the function that takes as argument an arbitrary nonnegative integer n and returns as value the units' digit of the decimal representation of n. The term r stands for the function that takes as argument an arbitrary non-negative integer n, and returns as value the number obtained by deleting the units' digit of the decimal representation of n; if the decimal representation of n only has one digit, r(n) = 0. The term l stands for the function that takes as argument an arbitrary non-negative integer n, and returns as value the number of digits of its decimal representation. The terms h and t stand for the functions (respectively, head and tail) specified in example 4. There are six simple internal states: S, E, W, C, ?, ??. S can be thought as the disposition to sum all the numbers in a given column; E is the disposition to write the leftmost digit of the sum of the leftmost column; W is the disposition to write the units' digit of the sum of any other column; C is the disposition to perform a carrying operation; ? is the disposition to check whether or not the current column sum is the final one; ?? is the disposition to check whether the writing of the final result has been completed. Note that the machine of example 4 is in fact part of the final subroutine (2a1)-(2a2) of table 5. The computation begins in simple internal state S, with the head scanning the rightmost digit of the uppermost addend. As said, the initial value of n is zero.

Bidimensional Turing Machines as Galilean Models of Human Computation

397

Table 5 (1)

Sn Sn Sn Sn

d 0 b –

: : : :

d 0 b –

D D D L

S(n+ (d)) Sn Sn ?n

START: adds n and all the numbers in a column, at the end of the column, goes one square to the left keeps the sum result, and calls test (2);

(2a)

?n

b

:

b

R 2D Ll (n)

En

TEST2: finished; goes to the appropriate square to start writing the final result, and calls the ending routine (2a1);

(2a1)

En

b

:

h(n)

R

??t(n)

writes the leftmost digit of the result of (1), goes to the right, keeps the tail of the result, and calls test (2a2);

(2a2a)

??s

b

:

h(s)

R

??t(s)

TEST2a2: not finished; writes the leftmost digit of the kept tail, goes to the right, keeps the new tail, and calls test (2a2);

(2a2b)

??0

b

:

b

H

??0

TEST2a2: finished; STOP.

(2a2c)

??0

d

:

d

H

??0

TEST2a2: finished; STOP.

(2b)

?n

–

:

–

RD

Wn

TEST2: not finished; calls routine (2b1);

(2b1)

Wn

b

:

u(n)

LU 2

Cr(n)

writes the units' digit of the result of (1), goes to the bottom of the column to its left, keeps the carrying number, and calls routine (2b2);

(2b2)

Cn Cn Cn Cn

d 0 b –

: : : :

d 0 b –

U U U D

Cn Cn Cn Sn

carries the number kept by routine (2b1) up to the uppermost column square, and calls routine (1);

398

M. Giunti

4. Turing thesis as the core hypothesis of an empirical theory of human computation Section 9 of Turing’s well-known 1936 paper begins with the explicit question as to the kind of operations that a human being typically performs when engaged in computing: The real question at issue is “What are the possible processes which can be carried out in computing a number?” (Turing 1936, [20, p.249]; my emphasys) By the terminology of sec. 2, we can say that Turing is asking for a suitably detailed description of a specific type of cognitive phenomenon, namely, the one of human computation. In general, by a phenomenon of human computation we mean any activity of a human being that consists in executing a purely mechanical or effective procedure, where a mechanical procedure is a finite set of clear-cut formal instructions for symbol manipulation; given a finite series of data, a human being must be able to carry out such instructions in a definite sequence of steps, with the exclusive aid of paper and pencil (or equivalent external devices), and without resorting to any special insight or ingenuity.g A few lines below, in sec. 9.I , Turing provides us with a comprehensive answer to the above question. In the first place, he proposes to just focus on computations carried out on a tape divided into squares (one-dimensional paper), with the motivation that “… it will be agreed that the two-dimensional character of paper is no essential of computation.” After introducing this simplifying hypothesis, Turing points out that a computation can always be thought as involving a finite number of symbols that can be printed on the paper, and a finite number of different “states of mind” of the human being that carries out the computation (the computer). Furthermore: The behavior of the computer at any moment is determined by the symbols which he is observing and his “state of mind” at that moment. We may suppose that there is a bound B to the number of symbols or squares which

g

This, except for one point, is Copeland’s characterization of a mechanical procedure (2002). The only difference is that Copeland also requires that a mechanical procedure “… will, if carried out without error, produce the desired result in a finite number of steps”. This requirement is not essential for a correct general characterization of a mechanical procedure and, in fact, it makes it too restrictive. For this condition would immediately rule out any program that is not ending for some input. On this point, also see Corradini, Leonesi, Mancini and Toffalori (2005, sec. 1.3).

Bidimensional Turing Machines as Galilean Models of Human Computation

399

the computer can observe at one moment. If he wishes to observe more, he must use successive observations. (Turing 1936, [20, p. 250]) Turing then invites us to … imagine the operations performed by the computer to be split up into “simple operations” which are so elementary that it is not easy to imagine them further divided. Every such operation consists of some change of the physical system consisting of the computer and his tape. We know the state of the system if we know the sequence of symbols on the tape, which of these are observed by the computer (possibly with a special order), and the state of mind of the computer. (Turing 1936, [20, p. 250]; my emphasis) This is a crucial passage, for Turing is in fact suggesting that any phenomenon of human computation is completely described by exactly three different magnitudes (the state variables of the phenomenon), which are (i) the complete content of the (one-dimensional) paper, (ii) the exact location of the symbols observed by the human computer and (iii) his state of mind. Having thus singled out the state variables, Turing goes on to describe the kind of change that these variables may undergo and, on the basis of this analysis, he finally concludes that we can always construct a machine that does essentially the same operations as the ones carried out by the human being: We may now construct a machine to do the work of this computer. To each state of mind of the computer corresponds an “m-configuration” of the machine. The machine scans B squares corresponding to the B squares observed by the computer. In any move the machine can change a symbol on a scanned square or can change anyone of the scanned squares to another square distant not more than L squares from one of the other scanned squares. The move which is done, and the succeeding configuration, are determined by the scanned symbol and the m-configuration. (Turing, 1936, [20, p. 251-252]) This machine is very similar to an ordinary Turing machine, for it is formed by a tape, an internal memory, and a head that, at each time step, scans a fixed small number of adjacent squares. Turing thesis properly so-calledh can thus be formulated as follows: h

Church thesis (1936) [2] is the following assertion: [CT ] A numeric function is computable in the intuitive sense iff it is recursive (or, equivalently, lambda-

400

M. Giunti

[TT ] Apart from inessential simplifications or idealizations, ordinary Turing machines are adequate models of the phenomena of human computation. As we have just seen, [TT ] should be interpreted in the strong sense that, apart from inessential details, the mechanical procedure that is executed by the human being involved in an arbitrary phenomenon of human computation is identical to the one executed by an appropriately chosen Turing machine.i As stated, [TT ] is a philosophical hypothesis, which turns out to be exceedingly plausible in the light of Turing’s speculative analysis of human computing. However, the very same analysis also provides us with the necessary elements for transforming [TT ] in the core methodological hypothesis of an empirical theory of the phenomena of human computation. First, we should notice that most of the idealizations or over-simplifications introduced by Turing in his analysis of human computing can be eliminated by taking into account not ordinary Turing machines, but bidimensional ones. In particular, this kind of machine allows us to directly deal with (a) human computations carried out on two-dimensional paper, (b) more complex movements than one square at a time, and (c) more complex operations as well, for the reading, writing, moving, and change-internal-state operations can now be carried out with the help of auxiliary functions. Second, Turing’s analysis tells us how to interpret the three state variables of any bidimensional Turing machine BT on the three corresponding magnitudes of any phenomenon of human computation C. For, according to the crucial passage quoted above, we should obviously interpret (i) the head position (expressed by a pair of integer coordinates that corresponds to the scanned square) as the spatial location (typically, the region of the two-dimensional paper) attended to by the human being at each computation step, (ii) the complete checkerboard content (expressed by a function that assigns a constant definable). The non-trivial part (left-right implication) of [CT ] is usually called Church-Turing thesis (Copeland 2002 [3]; Gandy 1980, [6, p.124 ] ), for Turing thesis [TT ], in conjunction with the equivalence between the two formal notions of recursive and Turing computable function, logically entails [CT ]. It is widely agreed that this inference, sometimes called the analogic argument, is the strongest available argument in favor of [CT ]. For an extended discussion of the relationships between [TT ] and [CT ], see Giunti and Giuntini 2007, [15, sec. 5]. i

The mechanical procedure executed by an ordinary Turing machine can be identified with its set of quintuples. The set of quintuples of a Turing machine obviously is a mechanical procedure in the intuitive sense specified in the second paragraph of this section.

Bidimensional Turing Machines as Galilean Models of Human Computation

401

to each checkerboard coordinate pair), as the complete content of the paper, and (iii) the machine’s internal state as the content of the human being’s working memory during the computation. Furthermore, Turing’s analysis also makes clear that the machine’s discrete time steps should be interpreted as the steps of the human being’s computation (that is to say, as the discrete time of the real phenomenon C ). Since the bidimensional Turing machine BT is a dynamical system with three components (see sec. 3, par. 13-14), the interpretation SC proposed above is in fact an interpretation of BT on C in exactly the sense defined in sec. 2, par. 6, so that (BT, SC) is a model of C. In addition, if SC is empirical and (BT, SC) is empirically correct, then (BT, SC) is a Galilean model of C. From now on, I will refer to the proposed interpretation SC as the standard interpretation of a bidimensional Turing machine BT on a phenomenon C of human computation. The previous observations thus motivate the following methodological version of Turing thesis: [MTT ]

For any specific phenomenon C of human computation, there is an appropriate bidimensional Turing machine BT such that (BT, SC) turns out to be a Galilean model of C.

Unlike the original formulation [TT ], this version of Turing thesis has a definite empirical content. In the next section, I will sketch how [MTT ] might function as the core hypothesis of a new empirical theory of human computation. 5. Developing the theory The proposed methodological version [MTT ] of Turing thesis claims that, for any phenomenon of human computation, there is an appropriate bidimensional Turing machine that, when interpreted in the standard way, turns out to be a Galilean model of that phenomenon. Let us then consider the set of all such models, and let us call this set the [MTT ]-based theory of human computation.j If [MTT ] is true, then such a set is obviously non-empty, and thus the [MTT ]-based theory is consistent in a semantic sense; in addition, it is also complete, in the sense that it contains at least one Galilean model for each j

More precisely, the [M TT ]-based theory of human computation is the set of all pairs (BT, SC ) such that, for some phenomenon C of human computation , BT is a bidimensional Turing machine, SC is the standard interpretation of BT on C, and (BT, SC ) is a Galilean model of C.

402

M. Giunti

phenomenon in its intended domain (that is to say, for each phenomenon of human computation). Conversely, if [MTT ] is false, then the [MTT ]-based theory is incomplete and, possibly, inconsistent as well. Thus, investigating the truth/falsity of [MTT ] is tantamount to investigating the completeness/incompleteness of the [MTT ]-based theory of human computation. At the moment, we do not know whether the [MTT ]-based theory is complete, or even consistent. However, the very formulation of [MTT ], together with Turing’s general analysis of human computation, suggests a definite empirical method for systematically investigating the content of this theory, so that we will then be able to put forth our best informed guess as to both its consistency and completeness. An all too concise sketch of this method is below. EMPIRICAL METHOD FOR INVESTIGATING THE [MTT ]-BASED THEORY OF HUMAN COMPUTATION 1. Focus on a specific phenomenon C = (F, BF) of human computation, where each specific phenomenon is singled out by its functional description F (see sec. 2, par. 4), which is based on the particular mechanical procedure P executed by the human computer involved in C; 2. try and specify a bidimensional Turing machine BT that executes a mechanical procedure (i.e. a set of quintuple schemata) as similar as possible to the one executed by the human computer of the phenomenon C; 3. consider the standard interpretation SC of BT on C, and claim that: (BT, SC) is a Galilean model of C. 4. Then, try to confirm this claim; that is to say, specify observation methods for each of the three state-magnitudes of the standard interpretation SC , as well as for its time-magnitude; 5. on the basis of the specified observation methods, gather empirical timeseries for each state-magnitude; 6. compare the observed time-series with the corresponding theoretical ones determined by BT ; 7. if the fit between observed and theoretical time-series is sufficiently good, (a) take claim 3 to be confirmed; otherwise, (b) do not take claim 3 to be confirmed; 7a1. if (a), consider a new specific phenomenon of human computation and start again from 1; 7b1. if (b), carefully revise the previous steps in reverse order; more precisely, first revise 7, then 6, 5 and 4;

Bidimensional Turing Machines as Galilean Models of Human Computation

403

7b2. if none of the previous revisions is sufficient to get confirmation of claim 3, revise claim 3 itself, by revising either step 2 (modify BT ) or step 1 (refine the functional description F that singles out C ); 7b3. then go on to step 4 and repeat from there. The method above will allow us to put forth an informed guess as to the consistency of the [MTT ]-based theory in the following sense. As soon as, for some phenomenon C of human computation and some model (BT, SC), a confirmation of claim 3 is reached, our guess about the consistency of the theory will be positive. If, on the contrary, after a sufficiently long and careful implementation of the method, no confirmation of claim 3 is forthcoming for any model of any phenomenon, then we will have good grounds to conjecture that the theory is inconsistent. As for completeness, if the application of the method yields the prospect of an ever increasing series of confirmed models, with no especially hard open puzzle or anomaly, then our guess about the completeness of the theory will be positive. If, on the contrary, after a sufficiently long and careful application of the method to a particular phenomenon C of human computation, no confirmation of claim 3 is forthcoming for any model of C, then we will have good grounds to conjecture that the theory is incomplete. A special feature of the method above is worth mentioning. Suppose that, for some phenomenon of human computation C, and some model (BT, SC), claim 3 has been confirmed; also suppose that the bidimensional Turing machine BT includes some auxiliary function f. Then, since the function f is computable in the intuitive sense (see note f ) there is a mechanical procedure Pf that computes f. As Pf is a mechanical procedure, it singles out a phenomenon of human computation Cf (see step 1 of the method above). We can thus go on applying the method to Cf , and eventually repeat, until we find a phenomenon C* and a model (BT *, SC *) whose bidimensional Turing machine BT * does not include auxiliary functions. The phenomenon C* can thus be thought as a basic phenomenon of human computation, while all the ones we encountered along our way from C to C*, taken in reverse order, can be thought as phenomena of increasing complexity. It is then natural to advance the following hypothesis as to the relationship between a phenomenon Cn and its more complex successor Cn+1 in the previous chain: The mechanical procedure Pn (constitutive of the simpler phenomenon Cn) has been previously rehearsed, and then internalized, by the human computer involved in Cn+1, so that Pn can now be internally and automatically executed as needed during the external and conscious execution of Pn+1.

404

M. Giunti

6. Concluding remarks − toward a Galilean approach to cognition The main thrust of this paper has been to show that the actual construction of Galilean models is clearly feasible, and well suited, for a special class of cognitive phenomena, namely, those of human computation. Whether this Galilean approach to cognition can be extended to other types of cognitive phenomena, or even to all of them, is a question that, at the moment, is difficult to settle in the positive sense. At the very least, however, there seems not to be any special reason for denying this possibility. Also, it is sufficiently clear how this rather abstract possibility of today might become a concrete one in the future. In order to extend the Galilean approach to new types of cognitive phenomena K1, K2, … , Kn, … we should first of all give an explicit characterization D1 of the phenomena of type K1. In particular, the level of precision of D1 should be comparable to the one of the informal definition of human computation given in sec. 5, par. 2. Second, on the basis of D1, a further analysis A1 of K1 should clearly point out (i) whether the time magnitude T of an arbitrary phenomenon of type K1 is discrete or continuous, and (ii) a finite number of magnitudes M1, M2, …, Mk which determine the state evolution of any phenomenon of type K1. The level of depth of A1 should be comparable to Turing’s speculative analysis of human computing (1936, sec. 9.I ). Third, we should then give an explicit and detailed characterization E1 of a special class of ideal systems, and a corresponding formal definition L1 of a special class of dynamical systems, in such a way that there is an obvious one-to-one correspondence between the E1-ideal systems and the L1-dynamical systems. In addition, an arbitrary L1-dynamical system should have as many state components as the number of magnitudes specified by A1 and, on the basis of both E1 and A1, each state component should be naturally identified with the set of possible values of exactly one of these magnitudes; in other words, E1 and A1 should provide a standard interpretation SH1 of any L1-dynamical system on any phenomenon H1 of type K1. Finally, for any phenomenon H1 = (F1, BF1) of type K1, it should be possible to naturally identify the abstract type of real system ASF1 (see sec. 2, par. 4) with an appropriately chosen E1-ideal system.k Fourth, we should state the basic methodological hypothesis for a new empirical theory of the cognitive phenomena of type K1: k

Note that, in the case of the phenomena of human computation, this identification is provided by the original formulation [TT ] of Turing thesis.

Bidimensional Turing Machines as Galilean Models of Human Computation

[MT 1]

405

For any specific phenomenon H1 of type K1, there is an appropriate L1-dynamical system DL1 such that (DL1, SH1 ) turns out to be a Galilean model of H1.

Fifth, we should then consider the [MT 1]-based theory of K1,l and start its empirical investigation by applying an empirical method analogous to the one described in sec. 5. If we are able to effectively carry out this detailed research program for K1, K2, … , Kn , … , then the Galilean approach to cognition will be implemented. References 1. V.I. Arnold, Ordinary differential equations (The MIT Press, Cambridge, MA, 1977). 2. A. Church, American Journal of Mathematics 58,345-363 (1936). 3. B.J. Copeland, in The Stanford Encyclopedia of Philosophy, Ed. E.N. Zalta, (2002). URL = . 4. F. Corradini, S. Leonesi, S. Mancini and C. Toffalori, Teoria della computabilità e della complessità (McGraw-Hill, Milano, 2005). 5. A.K. Dewdney, Scientific American 261(9),180-183 (1989). 6. R. Gandy, in The Kleene symposium, Ed. J. Barwise, H.J. Keisler and K. Kunen, (North Holland Publishing Company, Amsterdam, 1980), pp. 123-148. 7. M. Giunti, Computers, dynamical systems, phenomena and the mind, Ph.D. dissertation, (Indiana University, Bloomington, IN, 1992), (Published by University Microfilms Inc., Ann Arbor MI. UMI order number: 9301444). 8. M. Giunti, , in Mind as motion: Explorations in the dynamics of cognition, Ed. R.F. Port and T.J. van Gelder, (The MIT Press, Cambridge, 1995), pp. 549-571. 9. M. Giunti, in Proceedings of the 18th annual conference of the Cognitive Science Society, Ed. G.W. Cottrel, Mahwah, (L. Erlbaum Associates, NJ, 1996), pp. 71-75. 10. M. Giunti, Computation, dynamics, and cognition (Oxford University Press, New York, 1997). 11. M. Giunti, in Prospettive della logica e della filosofia della scienza: Atti del convegno triennale della Società Italiana di Logica e Filosofia delle Scienze, Roma, 3-5 gennaio 1996, Ed. V.M. Abrusci, C. Cellucci, R. Cordeschi and V. Fano, (Edizioni ETS, Pisa, 1998), pp. 255-267. 12. M. Giunti, in Storia e Filosofia della scienza: Un possibile scenario italiano. Atti del convegno Storia e filosofia delle scienze: lo stato delle ricerche italiane di

l

In general , the [M T i ]-based theory of Ki is the set of all pairs (DLi , SHi ) such that, for some phenomenon Hi of type Ki , DLi is a Li-dynamical system, SHi is the standard interpretation of DLi on Hi , and (DLi , SHi ) is a Galilean model of Hi.

406

13.

14. 15. 16. 17. 18.

19. 20. 21. 22.

M. Giunti punta, Padova, 28-30 maggio 1997, Ed. E. Bellone and G. Boniolo, (Milano, 1998), pp. 89-98. M. Giunti, in Atti del XIX congresso nazionale dell'Associazione Italiana di Psicologia, sez. di psicologia sperimentale, Cagliari, 18-20 settembre 2005, (AIP, Cagliari, 2005). M. Giunti, in Systemics of emergence: Research and development, Ed. G. Minati, E. Pessa and M. Abram, (Springer, Berlin, 2006), pp. 683-694. M. Giunti, and R. Giuntini, in Title yet to be announced, Ed. S. Mancini, (Mimesis Edizioni, Milano, 2007). M.P. Penna, and P. Ciaralli, in Third systems science european congress, Ed. E. Pessa, M.P. Penna and A. Montesanto, (Kappa, Roma, 1996), pp. 533-536. M.P. Penna, and E. Pessa, Comunicazioni Scientifiche di Psicologia Generale 8,151-178 (1992). M.P. Penna, and E. Pessa, in Proceedings of the second European Conference on Cognitive Modelling (ECCM-98), Ed. F.E. Ritter and R.M. Young, (Nottingham University Press, Nottingham, UK, 1998), pp. 120-126. W. Szlensk, An Introduction to the theory of Smooth Dynamical Systems (John Wiley and Sons, Chichister, England, 1984). A.M. Turing, Proceedings of the London Mathematical Society, Series 2, 42, 230265 (1936). A.J. Wells, Cognitive Science 22(3), 269-294 (1998). A.J. Wells, Rethinking cognitive computation: Turing and the science of the mind (Palgrave Macmillan New York, 2006).

A NEURAL MODEL OF FACE RECOGNITION: A COMPREHENSIVE APPROACH VERA STARA(1), ANNA MONTESANTO(1), PAOLO PULITI(1), GUIDO TASCINI(1), CRISTINA SECHI(2) (1) Università Politecnica delle Marche, Facoltà di Ingegneria, DEIT [email protected] [email protected] [email protected] [email protected] (2) Università degli Studi di Cagliari, Facoltà di Scienze della Formazione [email protected] Visual recognition of faces is an essential behavior of humans: we have optimal performance in everyday life and just such a performance makes us able to establish the continuity of actors in our social life and to quickly identify and categorize people. This remarkable ability justifies the general interest in face recognition of researchers belonging to different fields and specially of designers of biometrical identification systems able to recognize the features of person's faces in a background. Due to interdisciplinary nature of this topic in this contribute we deal with face recognition through a comprehensive approach with the purpose to reproduce some features of human performance, as evidenced by studies in psychophysics and neuroscience, relevant to face recognition. This approach views face recognition as an emergent phenomenon resulting from the nonlinear interaction of a number of different features. For this reason our model of face recognition has been based on a computational system implemented through an artificial neural network. This synergy between neuroscience and engineering efforts allowed us to implement a model that had a biological plausibility, performed the same tasks as human subjects, and gave a possible account of human face perception and recognition. In this regard the paper reports on an experimental study of performance of a SOM-based neural network in a face recognition task, with reference both to the ability to learn to discriminate different faces, and to the ability to recognize a face already encountered in training phase, when presented in a pose or with an expression differing from the one present in the training context. Keywords: face recognition, biometrics, neural networks.

1. Introduction Visual recognition of faces is an essential behavior of humans: human activity relies on the classification of faces as distinct from other objects and the ability to recognize facial stimuli has significant social implications. The importance of this recognition process can be easily understood when we take into account the fact that since birth infants show a special interest for faces (Johnson, 1991 [15], 407

408

V. Stara et al.

Johnson and Morton, 1991 [16]) and are able to discriminate mother’s face from a stranger’s face. Experimental studies performed with neurologically normal subjects have suggested that faces are perceived as a special class of stimuli, distinct from other patterned objects. Face recognition differs from object recognition in that the former involves representing a face as a single, complex whole, whereas the latter typically involves decomposition into constituent elements (Farah et al., 1999 [8]). According to this point of view, recognition performance is worse in neuropsychologically normal adults when faces are presented upside down than when objects are presented upside down (Valentine, 1988 [28]; Farah et al., 1995 [9]). Research on patients and neuroimaging studies have reported increased activation in the fusiform gyrus in concomitance with the presentation of faces, although less activation is observed if the faces are presented upside down (Kanwisher et al., 1997 [17]). This same group has reported greater activation in this region to faces than to human or animal heads (Kanwisher et al., 1999 [18]). Also using fMRI, Gauthier et al. (1999) [12] reported that as subjects acquired expertise in recognizing artificial stimuli, the middle fusiform gyrus in the right hemisphere was recruited and showed a pattern of activation that was indistinguishable from that elicited by faces. Similarly, under passive viewing, activation in this area was greater in a single subject with expertise in viewing “greeble” faces versus individuals lacking such expertise (Gauthier et al., 1997, 1998, 2004 [10,13,11]). Overall, these results suggest that the fusiform ‘face area’ becomes specialized with experience. In terms of facial emotion, Whalen et al. (1998) [30] reported increased activation in the amygdala to fearful faces, but decreased activation to happy faces. Morris et al. (1996) [20] observed that neuronal activity in the left amygdala was significantly greater in response to fearful as opposed to happy faces. Collectively, from the neuroimaging studies it appears that regions in and around the fusiform gyrus seem to play a role in face recognition, whereas the amygdala plays a particularly important role in the recognition of facial expressions. Within the amygdala, some nuclei have been found to be responsive to individual faces, whereas others respond to individual expressions (Aggleton et al., 1980 [2]; Nahm et al., 1991 [21]). Although an unambiguous picture is emerging that faces may be accorded special status by the brain, it remains unclear upon what basis face specialization develops. From an evolutionary perspective, recognizing faces would be adaptive, and thus, selected for through evolution.

A Neural Model of Face Recognition: A Comprehensive Approach

409

Both human and non-human primates use in fact their faces to produce a range of social signals; more importantly, non-human primates may depend more on this medium for communication than do adult humans, given the absence of oral language. Thus, it is not surprising that monkeys are adroit in both face and emotion recognition (Boysen and Bernston, 1989 [4]; Phelps and Roberts, 1994 [25]; Parr et al., 1998 [23]; Pascalis and Bachevalier, 1998 [24]). It is now to be taken into account that the body of knowledge summarized above is endowed not only with a scientific value but also with a practical valence. Namely there is a large number of commercial, security, and forensic applications requiring the use of face recognition technologies: crowd surveillance, access control, “mugshot” identification, face reconstruction, design of human computer interfaces, multimedia communication and contentbased image database management. So far, the performance in face recognition tasks of human subjects largely outperforms the one evidenced by Automatic Face Recognition Systems (AFR). This entails the need for emulating the ability of human visual system by relying on the knowledge previously quoted. In this regard, we must underline that, when trying to emulate human ability, some AFR could be forced to rely on genuinely systemic approaches. This could derive not so much from the need to integrate biological, as well as psychological knowledge, with technological one. Rather this could be a consequence of the fact that human visual system operates per se in a systemic way, as its recognition output emerges from the interactions occurring within a complex network of specific visual subsystems, each devoted to a specific detection task (see, e.g., Zeki, 1993 [31]; Bartels and Zeki, 2005 [3]). Of course, the adoption of such a perspective entails that, among the AFR, the ones based on artificial neural networks should be at the same time the most biologically plausible and successful. Namely artificial neural networks are, among the available models of emergent processes, perhaps the easier ones, as regards the concrete implementation and the operation. Just for this reason in this paper we will introduce a particular kind of AFR based on a specific neural network model. As regards the domain of AFR, we remark that in most cases the worry for a systemic approach is totally absent. An overall picture of this field is very complicated but will be shortly summarized in the following. Before beginning our exposition it is to be recalled that, after about 30 years of research, the field of face recognition gave rise to a number of feasible technologies whose development required an interdisciplinary cooperation between experts in very different domains, such as image processing, pattern recognition, neural networks, computer vision, computer graphics and

410

V. Stara et al.

psychology. This development was driven, in most cases, by practical needs. In particular, we had a growth of the need for user friendly systems that can secure privacy without losing person’s identity. Not only, whereas some biometric personal identification rely on the cooperation of the participants (for example fingerprint analysis or retinal/iris scan), face recognition systems are often effective without people’s cooperation. • For this reason, typical applications of Face Recognition are in: • Entertainment area for specific applications in video games, virtual reality, training programs, human robot interaction and human computer interaction; • Smart card area for specific applications in drivers’ licenses, entitlement programs, immigration, national ID, passport, voter registration, welfare fraud; • Information security area for specific applications in TV parental control, personal device logon, desktop logon, application security, database security, file encryption, intranet security, internet access, medical records, secure trading terminals; • Law enforcement and surveillance for specific applications in advanced video surveillance, CCTV control, portal control, postevent analysis, shoplifting, suspect tracking and investigation. The purpose of these systems is to recognize the shapes of the features of a person's face in a background. To achieve the goal, a camera captures the image of a face, and then the software extracts pattern information. The latter is then compared with the one contained in user templates. In this regard, two techniques are used: one compares feature sizes and relationships (for example the nose length and the distance between eyes); the other method matches person most significant image data with a record stored in a face database. Due to interdisciplinary nature of face recognition and its possible systemic view, in this contribute we decided, owing mostly to the reasons discussed above, to deal with face recognition through a comprehensive approach with the purpose to reproduce some features of human performance, as evidenced by studies in psychophysics and neuroscience, relevant to face recognition, through a computational system based on a neural network. What are the advantages and drawbacks of this systemic view? An obvious advantage would be the one of assuring the emergence of recognition in a way mimicking the one evidenced by neurophysiological studied on visual cortex operation. However, this advantage exists only in principle, as the details of the operation of human visual cortex are still poorly known. For this reason

A Neural Model of Face Recognition: A Comprehensive Approach

411

comparisons between data coming from neuroscience or psychology and the ones coming from computer science could be very useful. For psychologists, comparing human performance to the performance of computational models of face recognition can potentially give insight into the processes used by the human visual system to encode and retrieve faces. For computational theorists, knowledge of the way human face processing works may yield insights into processing strategies that will allow for more flexible and robust face processing systems (O'Toole et al., 2000 [22]). The principal drawback is obviously related to the difficulty in emulating the ability of human visual system. This difficulty is due to some open issues that need to be solved: 1. The face has a 3D shape. For this reason the appearance of the face could change depending on: projective deformations, which leads to stretching and foreshortening of different parts of face; self occlusion and dis-occlusion of parts of the face. If in the past experience a face was seen only from one viewing angle, in general it is difficult to recognize it from different angles. 2. The pose variation is an inevitable problem owing to the illumination variation of the background within and between days and among indoor and outdoor environments. The direct effect is due to the 3D shape of the face, that can create strong shadows and shading. It accentuates or diminishes certain facial features: the inherent amount of light reflected off of the skin and the non-linear adjustment in internal camera control, can have bad effects on facial appearance. 3. The face is also a non-rigid object: facial appearance varies with facial expression of emotion and paralinguistic communication along with speech acts. This problem is crucial for geometry based algorithms: facial expression affects the apparent geometrical shape and position of the facial features. 4. Faces change over time in hair style, makeup, muscle tension and appearance of the skin, presence or absence of facial hair, and over longer periods owing to effects related to aging. 5. Algorithms may be more or less sensitive to gender and race. Males and females might be harder to recognize owing to day-to-day variation in makeup or in structural facial features, for different local features and shape. Men’s faces have thicker eyebrows and greater texture in the beard region, whereas in women’s faces the distance between the eyes and brows is greater, the protuberance of the nose smaller, and the chin narrower than in men. Despite these problems, the synergy between neuroscience and engineering efforts allowed us to implement a model that had a biological plausibility,

V. Stara et al.

412

performed the same tasks as human subjects, and gave a possible account of human face perception and recognition. The obtained results evidence a good performance of SOM network we used. The latter, however, is crucially dependent on a correct choice of the number of categorizing units, as well as of learning parameter values. Moreover, also the gender of faces seems to play an important role, as female faces entail a lower performance with respect to the male ones. In any case, the use of a SOM network in face recognition appears as very convenient, owing to the fact that it allows a satisfactory performance despite its simplicity and the possibility of finding a neurophysiological interpretation of its operation. 2. The Neural Network Model The neural network model used in this research was a standard two-layered network with feedforward connections. Both input and categorization layer were 2-dimensional. The input stimulations were consisting in 2-dimensional images, suitably preprocessed through methods described in the next section. The training phase has been based on usual shortcut Kohonen algorithm. The explicit form of laws ruling time evolution of learning parameter and bubble radius, as well as the parameter values adopted in the simulations, were the following:

η 0 = 0.10 ; β = 0.0001; R0 = 4 ; b0 = 0.0001 R(t ) = R0 exp(−b0 t ) radius of activity bubble

α (t ) = η 0 exp(− β 0 t ) learning parameter

.

The duration of each training phase was directly fixed in advance by the experimenter. Once performed the training process, we left open two possibilities for using the network, keeping fixed the weight values obtained at the end of training, that is: a) a test (or categorization) task, in which a stimulus image was fed in input to the network and associated to the category represented by the winning categorization unit (according to Kohonen algorithm prescription), b) a recognition task, in which two sets of images were used, the target set, and the probe set. For each image of both sets, the output activation ui of each categorizing unit was computed through the formula:

ui =

k

wik xk

where wik denotes the components of the weight vector associated to the i-th categorizing unit, and xk is the output of the k -th input unit (to shorten the notations, we neglected here the 2-dimensional character of both input and

A Neural Model of Face Recognition: A Comprehensive Approach

413

categorization layer). In this way, each input image was associated to a pattern of activation of categorizing units. Then, for each image of the probe set, the Euclidean distances between its activation pattern and the ones of the single images of target set were computed. The image of the probe set taken into consideration was associated to the image of the target set whose activation pattern was characterized by the minimum distance. In most cases the images of target set and of probe set were referring to the same faces, the only difference being facial expression or pose. In this regard, if an image of the probe set was associated to another image of target set, containing the same face, the former face image was considered as recognized, while it was considered as not recognized in the contrary case. The face recognition accuracy was defined as the percentage ratio between the number of probe images recognized and the total number of probe images. 3. Face Image Preprocessing In order to make easier the operation of the network and to lower computational costs, a suitable preprocessing of face images was introduced. To this aim, a mask was superimposed on each face image, making some predefined points to coincide with eyes, mouth and nose. The mask dimensions can be adjusted to exactly match the points with corresponding face parts. Due to this operation, the images obtained with this method may have different dimensions. They were therefore rescaled to a standard size, through an interpolation routine, to allow a comparison between different faces. Then, the image was subdivided in two areas by means of a line passing vertically through the center of the nose, and the left part was specularly reproduced on the right, obtaining a perfectly symmetric face. This step was introduced to avoid the interference of natural asymmetries with the categorization procedure of the network, since the latter could mask the peculiar characteristics of the face due to different face parts positioning. Images were in gray-scale uncompressed PNG format. A normalization process was applied after network acquisition, to compensate for luminance differences. In these experiments we used the FERET dataset, a collection of images acquired through a 35-mm camera and then digitized. Two frontal views for each subject were taken into consideration: 1) a neutral facial expression (the learning probe), 2) a different facial expression (the testing probe). Each image of a given individual was taken on the same day with the same lighting. Moreover, we created 3 special classes of stimuli different for gender and race: 1) a class including 40 different male white faces (CLASS 1); 2) a

V. Stara et al.

414

Table 1. The experimental design.

Experiment 1 Experiment 2 Experiment 3 Experiment 4 Experiment 5 Experiment 6

Output NODES 6×8 9 × 12 12 × 16 12 × 16 12 × 16 12 × 16

Class of STIMULI Class 1 Class 1 Class 1 Class 1 Class 2 Class 3

Number of LEARNING STEPS 20000 20000 20000 40000 40000 40000

class including 40 different female white faces (CLASS 2); 3) a class including 40 different faces, of which 20 were female faces (10 white and 10 black) and 20 male faces (10 white and 10 black) (CLASS 3). For each class there was a learning set and a testing set, so as to have a total of 80 face images for class. 4. The Experimental Design A study of performance of a SOM-based network in a face recognition task must be designed in order to answer a number of questions, the most important being the following ones: 1. what is the optimal number of nodes of categorization layer? 2. what is the optimal number of learning steps? 3. are gender or race differences important in explaining network performance? To this aim, we performed six successive experiments, in each one of which the training was carried out on 40 different face images of the class taken into consideration. The same images used in the training phase were fed in input in the categorization phase, while the recognition phase was dealing with the other 40 face images of the same class, not taken into consideration in the training phase. As previously remarked, each face image used in the recognition phase represented the same face of one of the face images used in the training phase, the only difference being the facial expression or the pose. The features characterizing the six experiments are summarized in the following Table 1. We underline that the first three experiments were done in order to identify the best dimension of the SOM based on the idea that the number of neurons influences the recognition accuracy. Once known the best dimension, in the last two experiments we studied the accuracy of the SOM as regards gender and race giving homogeneous stimuli in the 4th experiment and inhomogeneous stimuli in

A Neural Model of Face Recognition: A Comprehensive Approach

415

Table 2. The activated categorization nodes for each face in Experiment 1. 0 1 2 3 4 5 6 7 8

0 F10,F14

1

2 F18

3

F38

4

5

F24,F28

F1, F2

F9,F19

F40 F25,F29,F31

F23

F4,F21 F37

F3

F12

F6,F30

F16

F33

F26

F22

F13,F20

6

F27,F34

F15,F17

F5,F39

F8,F11,F32

Accuracy 1 Experiment

%

100 50

77,5

65

0 5.000

10.000

77,5

15.000

77,5

20.000

Epochs Figure 1. Recognition accuracy vs. number of training steps in Experiment 1, using 6×8 output nodes.

the 5th experiments. The accuracies evidenced for these classes were then compared with the accuracy found in class 1. 5. The Outcomes of the Experiments 5.1. Experiment 1 The table 2 shows, in categorization phase, the specific nodes that every face image used in learning phase appears to be associated to. As one can see, there are 11 nodes associated to one and only one face, while we have a partial overlapping of the other 29 faces on the same nodes. Then, relying on the other 40 images of faces of the same class belonging to the probe set, we computed the recognition accuracy as a function of duration of training phase. The results are shown in Figure 1, and evidence that the accuracy grows from 65% after 5.000 steps to 77,5% after 10.000 step, corresponding to a correct recognition of 31 faces. A prolongation of training phase does not

V. Stara et al.

416

Table 3. The activated categorization nodes for each face in Experiment 2. 0 1 2 3 4 5 6 7 8 9 10 11 12

0 F33

1

2

3

F34

4 F5

5

F37

F23

F26 F30

7

F39

F14

F21

F40

F35

F12

8 F6,F38

9

F4

F27

F7,F22

6 F10

F20

F36 F18

F15

F13 F17

F25 F31

F24

F29

F3 F16

F9

F28

F2

F11 F1

F19 F8

F32

Accuracy 2 Experim ent 100 80

77,5 72 ,5

60

77,5

75

40 20 0 5000

10000

15000

20000

Ep o chs

Figure 2. Recognition accuracy vs. number of training steps in Experiment 2, using 9×12 output nodes.

produce an improvement of recognition accuracy, which lasts unchanged even after 20.000 steps. 5.2. Experiment 2 The activated categorization nodes (their total number is 36) in this experiment are shown in table 3. A partial overlapping of 4 faces occurs, more precisely in node 8,0 for faces 6 and 38 and on node 0,4 for faces 7 and 22. The recognition accuracy grows from 72,5% after 5000 steps to 77,5% after 15000 steps (see Figure 2). Even in this case a further increase of the number of learning steps does not produce an improvement of accuracy. From these data we can deduce that the new choice of number of nodes does not seem to

A Neural Model of Face Recognition: A Comprehensive Approach

417

Table 4. The activated categorization nodes for each face in Experiment 3. 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1

F17

2

3

4

F2

5

6

7

8

9

F20

10

11

12

F12 F13

F15

F1

F26

F34

F3,F22 F7

F16

F37

F5 F33

F30

F28

F24

F31

F36 F29

F23

F25

F40 F6

F9

F38 F18

F39

F19

F4 F14

F10

F8

F21 F32

F27

F35

F11

Accuracy 3 Experim ent 100 80 60 40 20 0

72,5

5000

80

77,5

10000

15000

80

20000

Epoc hs

Figure 3. Recognition accuracy vs. number of training steps in Experiment 2, using 12×16 output nodes.

influence the recognition accuracy. However, this choice led to a strong decrease of overlaps of different faces within the same category. 5.3. Experiment 3 The association between faces and categorizing nodes is shown in Table 4. Two faces (F3 and F22) are categorized in the same node 0,3 The recognition accuracy grows from 72,5% after 5000 learning steps to 80% after 15000 learning steps (see figure 3). This maximum value does not change if we add further 5000 learning steps. These data evidence how the new choice of the number of categorizing nodes be able to produce an increase in the

V. Stara et al.

418

Accuracy x Output Nodes 100

65

72,5 72,5

77,5 75 77,5

77,5 77,5 80

77,5 77,5 80

10000

15000

20000

50 0 5000

Epoc h s

6x8

9x12

12x16

Figure 4. Accuracy of Experiments x output nodes Accuracy CLASS 1

80

80 77,5

80

80

60 %

40 20 0 1000 0

2000 0

Epoch s

3000 0

4000 0

Figure 5. Recognition accuracy vs. number of training steps in Experiment 4.

maximum value of recognition accuracy and, at the same time, a decrease of categorization overlaps. These first three experiments showed us that the output layer configuration based on 12×16 nodes was the best among the tested ones (see in figure 4 a graphical representation summarizing our findings). Thus, in the following experiments we always used it. 5.4. Experiment 4 Even in this experiment we used 40 white male faces (CLASS 1) but we trained the network for 40.000 epochs. 80 % of faces were recognized (see the plot of accuracy vs. number of learning steps in figure 5). What should happen if into the CLASS 1 we would add 20 faces? Should we repeat the training phase? To answer this question we built a new dataset of 20 male faces (20 faces in the learning probe and the corresponding 20 faces in the

A Neural Model of Face Recognition: A Comprehensive Approach

419

Accuracy CLASS 2 80 60

60

60

60

60

% 40 20 0 1000 0

2000 0

Epoch s

3000 0

4000 0

Figure 6. Recognition accuracy vs. number of training steps in Experiment 5. Accuracy CLASS 3

80 60 %

70

72,5

72,5

72,5

40 20 0 1000 0

2000 0

Epoch s

3000 0

4000 0

Figure 7. Recognition accuracy vs. number of training steps in Experiment 6.

testing probe), and, using the weights obtained at the end of the previous experiment, we checked the recognition accuracy for 20 male faces. 95 % of the latter was recognized. Thus it seems that the network does not need to repeat the training phase when adding new faces. However, this optimal accuracy decreases if the stimuli are not of the same gender of those used in the learning phase: if we add 20 female faces, they are recognized for 42,5%, whereas if we add 20 faces, partly of female gender and partly of male gender, partly with white and partly with black skin, they are recognized for 60%. 5.5. Experiment 5 In this experiment we used 40 white female faces (CLASS 2) and we trained the network for 40.000 epochs. Faces were recognize for 60% (see figure 6). These data evidence that there are different performances associated to the different classes. Faces of CLASS 2 seem to be less recognized than faces of CLASS 1.

V. Stara et al.

420

Accuracy between classes 100 80

80

77,5

72,5

70 60

60

1000 0

2000 0

80

72,5

80

72,5

60

60

3000 0

4000 0

60 % 40 20 0

CLASS 1

Epoch s CLASS 2

CLASS 3

Figure 8. Recognition accuracy in the different classes.

Also in this case we built a new dataset of 20 female faces to check the accuracy for the added stimuli while keeping unchanged the weights of this experiment. The new faces were recognized for 90% and, also in this case, this optimal accuracy decreased if the stimuli were not the same of the learning phase: if we add 20 male faces, the accuracy is 85%, whereas if we add 20 mixed faces the accuracy falls to 67,5%. 5.6. Experiment 6 In this experiment we used 40 faces (CLASS 3), 20 female faces (10 white and 10 black) and 20 male faces (10 white and 10 black) and we trained the network for 40.000 epochs. Faces were recognized for 72,5% confirming that different kinds of stimuli were associated to different levels of recognition accuracy (see figure 7). As in other experiments, we built a new dataset of 20 mixed faces to check the accuracy for the added stimuli with the weights obtained in this experiment. The new faces were recognized for 90%. The accuracy fell to 85% using 20 male faces and to 50% using 20 female faces. The findings obtained in experiments 4, 5 and 6 are graphically summarized in figure 8. 6. Conclusions Even if the use of neural networks in face recognition dates to 25 years ago (Stoham, 1984 [26]; Abu-Mostafa and Psaltis, 1987 [1]; Kohonen, 1989 [19]; Golomb and Sejnowski, 1991 [14]; Brunelli and Poggio, 1992, 1993 [5,6]), nevertheless techniques based on NN deserve more study (Chellappa et al., 1995 [6]), especially as regards gender classification and facial expression. We used a neural network architecture in order to account for some features of visual information implied in human face processing. It is to be remarked that

A Neural Model of Face Recognition: A Comprehensive Approach

421

our model does not reproduce neither the processes underlying the individuation and the extraction of the face representation from the visual field nor the processes leading from face representation to the semantic information relative to individuals which the face belongs to. We evidenced how a Kohonen’s neural network could recognize face stimuli in a way which is dependent upon its past experience. Therefore we could say that, in principle, a SOM-based network could attain a satisfactory performance as a face recognizer. However the experiments 1-6 evidenced that: I) network performance is crucially dependent on the number of categorizing units, II) network performance is strongly dependent on the nature of face images used in the training phase, more precisely on their gender. With regard to the latter point, by using 3 kinds of stimuli classes we investigated how the recognition accuracy could vary as a function of the nature of the class: we found the best accuracy in a class composed of only men (80%) followed by a “mixed” class (men and women together) (72,5%). The worse performance was found in a class of only women (60%). Data evidences that the system is able to produce stable representations from exemplars as well as humans do. Into these categories, a gender effect exists. We explain it by the Face Space’s Theory. Faces may be encoded by reference to a generalized prototype: a sort of schema that emerges as a result of a lifetime’s experience with faces (Valentine and Bruce, 1986). The prototype can be considered as the ideal format of a subject or object, stored in the long-term memory. This assumption is based on the “prototypicality hypothesis”: relevant stimuli are represented within a multidimensional feature space and, thus, considering each face as a point in this space, faces are organized in a face space with the most prototypical one in the center (Valentine, 1991). Faces dissimilar from this prototype are the more distant ones from the center, and are called “distinctive faces” The distinctiveness effect shows that some faces are much easily recognizable than others. In particular, some faces are more distinctive in the general populations and for this reason they are also easier to recognize. Distinctive faces are recognized better because they are further from other neighbouring faces in the face space and so are less susceptible to confusion between faces located near each other in the space. Face space’s theory seems to explain our data. Kohonen’s network produces categories based on statistical features of the incoming stimuli. In these categories there are distinctive faces such as male faces. Male and female faces differ in feature-based information such as the size of the nose and prominence

V. Stara et al.

422

of the brow. We could therefore interpret these categories as modelling the salience maps emerging from the interactions between different modules in visual cortex, each one of them being identified with a particular categorizing unit. These results seem to confirm the plausibility of our biological model. Anyway, in order to improve the performance, as a face recognizer, of a SOM-based network, we should better understand the reason for the different performances with faces of different gender. In turn, this requires a further extension of both experimental studies on human subjects and computational modelling activities within the domain of Face Recognition. References 1. Y.S. Abu-Mostafa, D. Psaltis, Scientific American 256, 88-95 (1987). 2. J.P. Aggleton, M.J. Burton, R.E. Passingham, Brain Research 190, 347-368 (1980). 3. A. Bartels, S. Zeki, Philosophical Transactions of the Royal Society B 360, 733-750 (2005).

4. S.T. Boysen, G.G. Bernston, Journal of Comparative Psychology 103, 215-220 (1989).

5. R. Brunelli and T. Poggio, in Proc. DARPA Image Understanding Workshop, (1992), pp. 311-314.

6. R. Brunelli and T. Poggio, IEEE Transactions on PAMI 15,1042-1052 (1993). 7. R. Chellappa, C.L. Wilson, and S. Sirohey, Proc. IEEE 83, 705-740 (1995). 8. M. Farah, G.W. Humphreys, H.R. Rodman, in Fundamen-tal Neuroscience, Ed. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

M.J. Zigmond, F.E. Bloom, S.C. Landis, J.L. Roberts, L.R. Squire, (Academic Press, San Diego, CA; 1999), pp. 1339-1361. M.J. Farah, J.W. Tanaka, H.M. Drain,. Journal of Experimental Psychology: Human Perception and Performance 21, 628-634 (1995). I. Gauthier and M.J. Tarr, Vision Research 37(12), 1673-1682 (1997). I. Gauthier, M. Behrmann and M.J. Tarr, Neuropsychologia 42(14), 1961-70 (2004). I. Gauthier, M.J. Tarr, A.W. Anderson, P. Skudlarski, J.C. Gore, Nature Neuroscience 2, 568-580 (1999). I. Gauthier, P. Williams, M.J. Tarr and J. Tanaka, Vision Research, 38, 2401-2428 (1998). B.A. Golomb, T.J. Sejnowski in Advances in Neural Information Processing System 3, Ed. D.S. Touretzky and R. Lipmann, (Morgan Kaufmann, San Mateo, CA, 1991), pp. 572-577. M.H. Johnson, S. Dziurawiec, H. Ellis, J. Morton, Cognition 40, 1-19 (1991). M.H. Johnson, J. Morton, Biology and Cognitive Development: The Case of Face Recognition (Blackwell Press, Cambridge, MA, 1991). N. Kanwisher, J. McDermott, M.M. Chun, Journal of Neuroscience 17, 4302-4311 (1997). N. Kanwisher, D. Stanley, A. Harris, NeuroReport 10, 183-187 (1999). T. Kohonen, Self-Organization and Associative Memory (Springer, Berlin, 1989).

A Neural Model of Face Recognition: A Comprehensive Approach

423

20. J.S. Morris, C.D. Frith, D.I. Perrett, D. Rowland, Nature 383, 812-815 (1996). 21. Nahm, F.K., Albright, T.D., Amaral, D.G. Society for Neuroscience Abstracts 17, 473 (1991).

22. A.J. O'Toole, Yi Cheng, P.J. Phillips, B. Ross, H.A. Wild, 2000. in Fourth IEEE 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.

International Conference on Automatic Face and Gesture Recognition, (28-30 March 2000), pp. 552-557. L.A. Parr, T. Dove, W.D. Hopkins, Journal of Cognitive Neuroscience 10, 615-622 (1998). O. Pascalis, J. Bachevalier, Behavioural Processes 43, 87-96 (1998). M.T. Phelps, W.A. Roberts, Journal of Comparative Psychology 108, 114-125 (1994). T.S. Stonham, In Aspects of Face Processing, Ed. H.D. Ellis, M.A. Jeeves, F. Newcombe and A. Young, (Nijhoff, Dordrecht, 1984), pp. 426-441. T. Valentine, Quarterly Journal of Experimental Psychology 43A, 161-204 (1991). T. Valentine, British Journal of Psychology 79, 471-491 (1988). T. Valentine, V. Bruce, Perception 15, 525-535 (1986). P.J. Whalen, S.L. Rauch, N.L. Etcoff, S.C. McInerney, M.B. Lee, M.A. Jenike, Journal of Neuroscience 18, 411-418 (1998). S. Zeki, A vision of the brain (Blackwell, Oxford, UK, 1993). W. Zhao, R.Chellappa, A. Rosenfeld, and P.J. Phillips, Face recognition: A literature survey, CVL Technical Report, (University of Maryland, 2000), ftp://ftp.cfar.umd.edu/TRs/CVL-Reports-2000/TR4167-zhao.ps.gz .

This page intentionally left blank

ANTICIPATORY COGNITIVE SYSTEMS: A THEORETICAL MODEL

GRAZIANO TERENZI Department of Psychology, University of Pavia, Italy E-mail: [email protected]

This paper deals with the problem of understanding anticipation in biological and cognitive systems. It is argued that a physical theory can be considered as biologically plausible only if it incorporates the ability to describe systems which exhibit anticipatory behaviors. The paper introduces a cognitive level description of anticipation and provides a simple theoretical characterization of anticipatory systems on this level. Specifically, a simple model of a formal anticipatory neuron and a model (i.e. the -mirror architecture) of an anticipatory neural network which is based on the former are introduced and discussed. The basic feature of this architecture is that a part of the network learns to represent the behavior of the other part over time, thus constructing an implicit model of its own functioning. As a consequence, the network is capable of self-representation; anticipation, on a macroscopic level, is nothing but a consequence of anticipation on a microscopic level. Some learning algorithms are also discussed together with related experimental tasks and possible integrations. The outcome of the paper is a formal characterization of anticipation in cognitive systems which aims at being incorporated in a comprehensive and more general physical theory. Keywords: anticipation in cognitive systems, anticipatory neural networks, learning algorithms.

1. Introduction The ability to anticipate specific kinds of events is one of the most amazing features of biological organization. Mainly, it is related to the functioning of complex biological systems which perform cognitive functions. An anticipatory system is a system which decides its behavior by taking into account a model of itself and/or of its environment; i.e. it determines its current state as a function of the prediction of the state made by an internal model for a future instant of time (Rosen, 1985) [17]. A system endowed with anticipatory qualities is also named a proactive system. Proactivity and anticipation, thus, are two important features of biological organization. As a matter of fact, anticipatory properties play an important role in the mechanisms which rule learning and control of complex motor behaviors. Indeed, as it stands, motion is the only means biological organism have to interact both with their environment and with other organisms (Wolpert, 425

426

G. Terenzi

Gharamani and Flanagan, 2001 [24]). Recent empirical studies have demonstrated the involvement of the motor system in processes ranging from the observation and anticipation of action, to imitation and social interaction (Gallese, Fadiga, Fogassi and Rizzolatti, 1996 [4]; Rizzolatti and Arbib, 1998 [14]; Liberman and Whalen, 2000 [10]; Wolpert, Doya and Kawato, 2003 [25]). In this context the problem of anticipation is essentially the problem of understanding how cognitive systems of complex organisms can take into account the expected future evolution of events which take place in the interaction with their environment in order to take decisions and determine their behavior (Sutton, 1990 [20]; Stolzmann, 1998 [19]; Baldassarre, 2002, 2003 [1,2]). The study of anticipation in the context of sensory-motor learning and coordination has been carried out by developing an interesting ensemble of computational models which aim to take into account the anticipatory properties of biological neural networks. Among these models we must mention the Forward Models by Jordan and Rumelhart (1992), the architectures based on Feedback-Error Learning procedure (Kawato, 1990) [9] and on the Adaptive Mixture of Local Experts (Jacobs, Jordan, Nowlan and Hinton, 1991 [6]; Jacobs, 1999 [7]), such as the MOSAIC model (Wolpert and Kawato, 1998 [23]; Haruno, Wolpert and Kawato, 2001 [5]). Essentially, on the basis of computational studies it has been hypothesized that the central nervous system is able to simulate internally many aspects of the sensory-motor loop; specifically, it has been suggested that a dedicated neural circuitry is responsible of such processing and that it implements suitable “internal models” which represent specific aspects of the sensory-motor loop. Internal models predict the sensory consequences of motor commands and, for this reason, are called forward models, as they model forward causal relations (i.e. to a future instant of time) between actions and their sensory consequences. A forward model is employed to predict how the state of the motor system will change in response to a given motor command (motor command state of the motor system). A forward model is therefore a predictor or a simulator of the consequences of an action. The inverse transformation from the state to the motor command (state motor command), which is needed to determine the motor command required to reach a given goal, is performed by what is called an inverse model. Since they take into account only the global patterns of behavior of the system, a basic feature of these models is that they are essentially bounded to a coarse grained treatment of anticipation, thus giving rise to a macroscopic (i.e. global) notion of anticipation. Incidentally, they do not consider the possibility for this global

Anticipatory Cognitive Systems: A Theoretical Model

427

notion of anticipation to be a consequence of a more general theory of anticipation working on other levels as well; to step closer towards the aforementioned goal is one of the main focuses of this paper. On the other hand, it is also clear that biological systems are first of all physical systems. For this reason understanding anticipation is a problem for any physical theory which aims at explaining the emergence of biological organization. For this same reason, the quest for such an explanation requires both the clarification of the levels of description of the systems under study, and their integration in a unitary and more comprehensive theoretical framework. The main goal of this paper, then, is to identify a possible characterization of anticipation in biological systems which perform cognitive processing, in such a way as to bring to light some properties that any biologically plausible physical theory must incorporate. 2. Anticipatory Neurons Essentially, a formal neuron is function F which maps an input space I to a corresponding output space A. A non-anticipatory formal neuron is a function which has the form

at = F (it )

,

(1)

where it and at are respectively input and activation of the neuron at time t. An anticipatory formal neuron instead can have the following form

at = F (it , it +τ ) it +τ = G (it )

,

,

(2) (3)

where it+τ is computed by a suitable function G, named a predictor. If the activation at depends both on the input at time t and on activation of the neuron predicted at time t+τ the neuron has the following form

at = F (it , at +τ ) ,

(4)

at +τ = G (at ) .

(5)

This amounts to saying that the activation of a neuron at time t does not depend only on its input at time t, but also depends on its input (or in case on its output) at time t+τ, as it is computed by a suitable function G. Here, τ represents a “local time” parameter strictly related to predictor G. In the context of this model, we have to distinguish indeed between the global time t and the local time t’= t+τ. Whereas global time t describes the dynamics of function F,

428

G. Terenzi it

at

it

it

at

F

F at

F

G (a)

it+τ

at+τ

G (c)

(b)

Figure 1. Three different models of a neuron: (a) a standard non-anticipatory formal neuron; (b) an anticipatory formal neuron which anticipates over input; (c) an anticipatory formal neuron which anticipates over its own output.

local time t’ describes both the activation dynamics and the learning dynamics of the predictor function G. Parameter τ identifies a temporal window over which the dynamics of the anticipatory neuron is defined. It sets up the limit of an internal counter which is used, step by step, to fix values of function G during the learning phase (approximation). In this sense τ represents a measure for the quantity of information that a neuron can store while executing its computation. This fact also means that the neuron is endowed with a kind of internal memory. This memory makes the neuron take into account more information than the information which is already available instantaneously. 3.

An Example of Dynamical Evolution of a Single Anticipatory Neuron

The following example illustrates the dynamics of a single anticipatory neuron trained to associate input and output by means of the Widrow-Hoff rule (Widrow and Hoff, 1960) [22]. The predictor also is trained by means of the Widrow-Hoff rule. Let us consider a neuron with the generic form like the one illustrated in Figure 2. Let us suppose that F is a sigmoidal function, that is it has the form

at =

1 1 + e − Pi

(6)

where P is the activation potential of the neuron and it is computed as,

Pi =

wij i j

(7)

j

and where wij are connection coefficients of the input links to the i-th unit, and ij are its input values. Let us suppose that G has the same form as F except

Anticipatory Cognitive Systems: A Theoretical Model

it

w11

w21

429

at

F it+τ

G

w12

Figure 2. An anticipatory formal neuron with suitable connections weights.

for computing the quantity, i 't +τ , devoted to approximate its input at time t+τ instead of computing output at that is:

i 't +τ =

1 1 + e − Pi

(7)

where P, i.e. the potential of G, is computed as the potential of F. As it can be seen, the anticipatory neuron is a system which includes two different subsystems, the anticipator F and the predictor G. Whereas F computes its activation on the global time scale t, G computes its activation on the local time scale t’; and the activation is computed by storing in memory the input to the anticipator F both at time t and at time t+τ . Function G, on the other hand, is approximated by resorting to these values. Table A illustrates synthetically a generic dynamics for the previously introduced anticipatory formal neuron. Essentially, in the course of the dynamical evolution of the neuron, for each couple of patterns presented to the neuron, a propagation cycle and a learning cycle are first executed for the predictor G, and then a propagation cycle and a learning cycle are executed for the anticipator F. It must be underlined that the choice of Widrow-Hoff rule for the training cycle of the neuron has been taken only for simplicity and for illustrative purposes; other algorithms could suit well. 4. Networks of Anticipatory Neurons By taking inspiration from the simple model of an anticipatory neuron introduced in the previous section, networks of anticipatory neurons can be designed that carry out complex tasks which involve the employment of suitable “internal models” (be they implicit or explicit) of their external or internal environment in order to take decisions.

430

G. Terenzi Table A. Example of the computation of the system at subsequent time steps

-t=0 -τ = 1 - t’ = (t + τ ) = 1 - it = i0

-t=1 -τ = 1 - t’ = (t + τ ) =2 - it = i1

-t=2 -τ = 1 - t’ = (t + τ ) = 3 - it = i2

- Initialize variables - Compute G

- Approximate “G” 1 - Compute Error

- Approximate “G” 1- Compute Error

i1' = G (i0 ) - Propagate Activation

a0 =

F (i0 , i1' )

- Approximate “F”

EG =

1 ' (i1 − i1 ) 2

2 - Weights Update

∆w21 = −η

1 E F = (a0 − a0T ) 2

1 ' (i2 − i2 ) 2

2- Weights Update

∆w21 = −η

∂EG = ∂w21

= −η (i1' − i1 ) ⋅ G ' ( PG ) ⋅ i0

= −η (i2' − i2 ) ⋅ G' ( PG ) ⋅ i1

where PG = ( w21i0 ) − sG

where PG = ( w21i1 ) − sG

- Compute Activation “G”

- Compute Activation “G”

i2' 1- Compute Error

∂EG = ∂w21

EG =

= G (i1 )

- Propagate Activation

a1 =

F (i1 , i2' )

i3' = G (i2 ) - Propagate Activation

a2 = F (i2 , i3' )

2- Weights Update

- Approximate “F”

- Approximate “F”

∂E ∆w12 = −η F ∂w12

1- Compute Error

1- Compute Error

1 E F = (a1 − a1T ) 2

EF =

2- Weights Update

2- Weights Update

∂E ∆w12 = −η F ∂w12

∆w12 = −η

∂E F ∂w12

∂E F ∂w11

∆w11 = −η

∂E F ∂w11

∂E ∆w11 = −η F ∂w11

∆w11 = −η

1 (a2 − a 2T ) 2

The scheme in Figure 3 describes a simple example of a neural network constructed by resorting to anticipatory neurons as its building blocks, which calculate their activation as a function of input values predicted at time t+τ. Input units within the scheme are represented by an array i of input values. Hidden units are represented by a composition of sigmoidal functions, i.e. g and

Anticipatory Cognitive Systems: A Theoretical Model

431

Figure 3. Simple neural network constructed by resorting to anticipatory neurons as its building blocks.

f respectively. Whereas the functions g within the hidden layer give rise to specific activation values, xg, represented here by an array xg, functions f give rise to another set of activation values. Similarly, for the output layer, the activation of output units is computed by means of composition of two suitable functions g and f, and their activation values are collected respectively in the arrays ug and uf. It must be stressed that each function g is endowed with a stack A, i.e. a memory which stores input vectors spread across a suitable temporal window t + τ ; it can be represented by a τ × N dimensional matrix M, where N is the number of storable patterns, and τ is the time step when they are given as input to g. For each layer of the network there are two kinds of connections which play different roles: connections to functions f (fconnections) and connections to functions g (g-connections). Indeed, fconnections between input and hidden layers are characterized by the connection weights w (which in the figure are represented without indices); g-connections, on the other hand, are represented by connection weights w’. Similarly, on the one hand, f-connections between hidden and output layers are characterized by connection weights c; on the other hand, g-connections are characterized by connection weights c’.

432

G. Terenzi

The case reported in figure shows a network in which each anticipatory neuron is endowed with a neuron-specific stack A which lets the neuron store and compare patterns given as input to the neuron in its temporal window (i.e. Input or output patterns, depending on the nature of the device). With regards to the latter, the parameter τ can be approximated by suitable learning algorithms, thus giving rise to networks which are able to perform tasks which require indefinite long-range time correlations for their execution. Training in such a network can be made in many ways. A basic option is to employ a back-propagation algorithm (Rumelhart, Hinton and Williams 1986) [18] in order to learn weights w and c of the functions f within the layers of the network, and to use Widrow-Hoff rule for the weights w’ and c’ of the corresponding functions g. In the case of an “on-line” learning procedure, a learning step for the whole network can be implemented, for example, through the following succession of phases: • Initialize Variables • Give a pattern as Input and Propagate Pattern • For each layer, Modify weights of functions g according to their temporal window (Widrow-Hoff) Compute g Compute f Modify weights of functions f according to the corresponding target pattern (Backpropagation). Table B lists of the basic variables of the model, i.e. a set of constructs that can be used to implement and simulate the model under study.

5. The τ -Mirror Architecture: Networks In the previous sections we have introduced both single anticipatory neurons and a simple network architecture. In the example of the last section each function g was connected to one and only one function f within its layer, and, moreover, it was characterized by a neuron-specific time window. This fact entails the employment of a specific stack for each function g. According to this description, the stack can be represented as a variable-structure quadrimensional matrix which comprises 1) the parameter τ of the temporal window, 2) the number N of storable patterns, 3) a dimension which represents neurons, and 4) a dimension which represents the value of τ for those neurons.

Anticipatory Cognitive Systems: A Theoretical Model

433

Table B. The basic variables of the model. Constructs i

Description Input vector

xf

Vector of the activation of functions f within the hidden layer

xg

Vector of the activation of functions g within the hidden layer

uf

Vector of the activation of functions f within the output layer

ug

Vector of the activation of functions g within the output layer

W

Weight matrix of the f-connections between input and hidden layers Weight matrix of the g-connections between input and hidden layers Weight matrix of the f-connections between hidden and output layers Weight matrix of the g-connections between hidden and output layers τ × N matrix which stores N input to functions g of inputhidden layer within temporal window τ τ × N matrix which stores N input to functions g of hiddenoutput layer within temporal window τ

W’ C C’ Aw Ac

Another possibility is to consider network architectures where functions g are connected in output not only to a single corresponding function f but to all of the other functions f belonging to the same layer. Moreover we can consider a generalized window t + τ for every g of the same layer or of the whole network. The set of all functions g of the network constitutes a real “temporal mirror” for the corresponding functions f. In a sense, we can say that a part of the network learns to represent the behavior of the other part from the point of view of the specific task at hand, by constructing an implicit model of its own functioning. Briefly, the network is capable of self-representation. For this reason the units that approximate functions g of the network within the temporal window t + τ (i.e. the ones we have previously dubbed “predictors”) are called “τ-mirror units” of the network. The schemes presented in Figure 4 and 5 represent this idea synthetically. Figure 4 illustrates an anticipatory network which predicts over its input spaces. As anticipated, if each function g is connected to every function f within the same layer, and not only to the corresponding ones, then they can be grouped together in a suitable layer, called a “mirror layer”. Moreover, if the same temporal window t+τ is generalized to all of the mirror units within the same layer, then the complexity of both the representation and the implementation of

434

G. Terenzi

Figure 4. An anticipatory neural network which anticipates over its input space.

the model can be further reduced. Figure 5 represents an anticipatory neural network which predict over its output spaces.

6. The τ -Mirror Architecture: Learning Algorithms We have previously introduced a learning scheme for the training of the neural network model under study. And this scheme is a supervised learning scheme which is based on 1) the employment of the Backpropagation rule for the adjustment of the weights of the functions f (W and C) and 2) the employment of the Widrow-Hoff rule for the adjustment of the weights of the functions g. By the way, there are other possibilities that can be explored.

6.1. Reinforcement Learning A straightforward solution is given by the employment of a reinforcement learning algorithm for approximating the weights of the functions f. Typically, reinforcement learning methods are used to approximate sequential dynamics in the context of the online exploration of the effects of actions performed by an agent within its environment (Sutton and Barto, 1998 [21]; Baldassarre, 2003 [2]). At each discrete time step t the agent perceives the state of the world st and it selects an action At according to the latter perception. As a consequence of each action the world produces a reward rt+1 and a new state st+1. An “action selection policy” is defined as a mapping from states to action selection probabilities : S × A [0, 1]. For each state s in S, a “state-evaluation

Anticipatory Cognitive Systems: A Theoretical Model

435

Figure 5. Anticipatory neural network which anticipates over its output space.

function” is also defined, V [s], which depends on the policy computed as the expected future discounted reward based on s:

V π [ s ] = E[rt +1 + γ rt + 2 + γ 2 rt + 3 + ...] =

[ p ssa ' ( rt +1 + γ V π [ s ' ])]], (8)

[π [ a, s ] a∈ A

and is

s '∈S

selects an action a given the where [a,s] is the probability that the policy state s, E is the mean operator and γ is a discount coefficient between 0 and 1. The goal of the agent is to find an optimal policy to maximize V [s] for each state s in S. In a reinforcement learning setting, the estimation of expected values of states s, V(s), is computed and updated typically by resorting to Temporal Differences Methods (TD methods); this amounts to saying that the estimation of V(s) is modified by a quantity that is proportional to the TD-error, defined as follows

δ = r + γ V (s' ) − V ( s) .

(9)

An effective way to build a neural netowrk which is based on reinforcement learning is to employ a so called Actor-Critic Architecture (Figure 6). Actor-Critic Architectures are based on the employment of distinct data structures for action-selection policy management (“actor”) and for the evaluation function (“critic”). The “actor” selects actions according to its input. In order to select an action, the activation of each output unit of the actor is given as input to a stochastic action selector which implements a winner-take-all selection strategy. The probability for an action ag to become the selected action aw is given by

436

G. Terenzi

Figure 6. An Actor-Critic architecture.

P[a g = aw ] =

mg [ yt ] k

mk [ yt ]

,

(10)

where mg[yt] is the activation of output units as a function of input yt . On the one hand, the “critic” is constituted by both an Evaluator and a TDcritic. The Evaluator is a network which estimates the mean future rewards which can be obtained in a given perceived state of the world yt = st. The Evaluator can be implemented as a feed-forward two-layer network with a single output unit which estimates V’[s] of V [s] , where the latter is defined as above. On the other hand, the TD-critic is a neural implementation of the function that computes the TD-error et as a difference between the estimations of V [s] at time t+1 and time t:

et = ((V π [ yt ]))t +1 − ((V π [ yt ]))t = (rt +1 + γ V π [ yt +1 ]) − V π [ yt ] ,

(11)

where rt+1 is the observed reward at time t+1, V’ [yt+1] is the estimated value of the expected future reward starting from state yt+1 and γ is, as usual, a suitable discount parameter. According to such a TD-error, the “critic” is trained by means of Widrow-Hoff algorithm; specifically its weights wj are updated by means of the following rule

∆ w j = η et y j = η ((rt +1 + γ V π [ yt +1 ]) − V π [ yt ] ) y j .

(12)

Anticipatory Cognitive Systems: A Theoretical Model

437

Also the “actor” is trained by resorting to the Widrow-Hoff algorithm and only the weights which correspond the winning unit are updated by the rule

∆ wwj = ζ et (4m j (1 − m j )) y j = ζ ((rt +1 + γ V 'π [ yt +1 ]) − V 'π [ yt ] )(4m j (1 − m j )) y j

(13)

where ζ is a suitable learning parameter between 0 e 1 and (4mj (1− mj)) is the derivative of the sigmoid function corresponding to the j-th output unit which wins the competition, multiplied by 4 in order to homogenize the learning rates of the two networks.

6.2. Reinforcement Learning for Anticipatory Networks Extending a reinforcement learning algorithm to the anticipatory neural networks described above is quite straightforward. Let us consider an anticipatory network like the one described in section 4., but with two-layers only. For this reason it will have a single layer of τ-mirror units. Then, let us consider an “Actor” constructed by resorting to the latter network; in this case, input units code for the state of the world at time t, whereas output units code for possible actions that can be executed at time t. The “Critic” can be considered as analogous to the one just discussed in the previous section. In this context, the approximation of weights for the functions f within the output layer of the network goes exactly as described in the previous section, that is by modifying the weights of the f-connections according to the TD-error computed by the Critic. But how can the weight over the g-connections be updated? Also in this case, the extension is straightforward if we consider the fact that the adjustment of the weights of these units is based on the WidrowHoff rule. The idea is to employ the TD-errors corresponding to the reference temporal window rather than the quadratic error corresponding to input (or output) patterns. This essentially means that the stack of the g units has to store, beside the input patterns to g within the temporal window τ, also the corresponding TD-errors. Therefore, the error signal is computed not by considering the quadratic error with respect to suitable targets, but by resorting to the TD-error corresponding to the reference time step.

6.3. Other Learning Algorithms Beside the ones discussed in the preceding sections, there is a wider set of learning methods that could be employed for the training of this kind of networks. However, their application must be suitably assessed and analyzed

438

G. Terenzi

firstly by discussing their theoretical implications. Even if it is not the goal of this paper to discuss this topic, we recognize that it is a very interesting possibility to extend the application of non-supervised learning algorithms (such as Kohonen-like-methods) to this training setting. Another possibility is to employ genetic algorithms.

7. Experimental Tasks for Simulation Assessment How can the viability of such an approach to the study and design of neural architectures be assessed? If, on the one hand, the assessment of the viability of a theoretical model essentially amounts to test its predictive and/or explicative power with respect to natural phenomena, on the other hand, the assessment of its utility from a practical point of view requires it to satisfy suitable constraints and goals within a design setting; in this sense it must solve specific scientific and design problems. A simpler simulation task, which is similar to the one just described, is to follow and anticipate a trajectory of an object in a more abstract manner, that is by employing the whole body of the simulated agent and not the simulated arm. Other experimental settings can be conceived that involve the execution of both planning and navigation tasks in simulated environments. More abstract simulation tasks can involve learning and anticipation of things such as numeric series, grammars and automata, which do not possess necessarily a material counterpart.

8. Conclusions and Future Work The construction of neural network architectures endowed with selfrepresentation abilities as well as with the proactive ability to show behaviors which do not depend only on their state at time t but also on their predicted state at a future time step is a strategic lever for understanding phenomena ranging from sensory-motor control to thought and consciousness. In this explorative paper we have introduced a possible form of both anticipatory neurons and networks which have been named “τ-mirror networks”. We have proposed some learning algorithms for this kind of networks and have put forward the possibility to extend other algorithms suitable for training. In a sense, it can be acknowledged that anticipation at the system level (i.e. what we have called the “global” notion of anticipation) can be considered as an emergent property of the interaction among the anticipatory components of the system. For this reason, the model presented in this paper provides a more

Anticipatory Cognitive Systems: A Theoretical Model

439

general treatment of anticipation on a fine grained basis, thus relating anticipation to the micro-behaviors of the components of the system. Among the future developments, a possibility (which has to be evaluated in the detail) is to change the form the anticipatory neurons and networks in order to make them more biologically and physically plausible. For example, Dubois (1998) has introduced an anticipatory form of the neuron of McCulloch and Pitts which is based on the notion of hyperincursion and that could be employed in the construction of another kind of anticipatory network. The hyperincursive computing systems are demonstrated to show stability properties much relevant if compared to the non-anticipatory counterparts. Therefore, a future development could be the integration of hyperincursivity within the theoretical framework presented in this paper. Notwithstanding the aforementioned theoretical developments, the most important and critical step is to develop a more comprehensive physical theory which will be able to describe anticipatory systems including those of the kind introduced in this paper. We strongly believe that such an integration is necessary to understand this very important aspect of biological organization. Without such a step, indeed, it would be impossible to account for a wide range of phenomena occurring in both the biological and cognitive realms.

References 1. Baldassarre, Planning with Neural Networks and Reinforcement Learning, PhD Thesis, (Computer Science Department, University of Essex, Colchester-UK, 2002). 2. Baldassarre, in Adaptive Behaviour in Anticipatory Learning Systems, Ed. Butz, Sigaud and Gèrard, (Springer, Berlin, 2003), pp. 179-200. 3. Dubois, in AIP Conference Proceedings 437, (1998), pp. 3-29. 4. Gallese, Fadiga, Fogassi and Rizzolatti, Brain 119, 593-609 (1996). 5. Haruno, Wolpert and Kawato, Neural Computation 13, 2201-2220 (2001). 6. Jacobs, Jordan, Nowlan and Hinton, Neural Computation 3, 79-87 (1991). 7. Jacobs, Trends in Cognitive Sciences 3(1), 31-38 (1999). 8. Jordan and Rumelhart, Cognitive Science 16, 307-354 (1992). 9. Kawato, in Neural Networks for Control, Ed. Miller III, Sutton and Werbos, (MIT Press, Cambridge, MA, 1990). 10. Liberman and Whalen, Trends in Cognitive Science 4, 187-196 (2000). 11. Luppino and Rizzolatti, News in Physiological Science 15, 219-224 (2000). 12. McClelland and Rumelhart, (Eds), Parallel Distributed Processing. Exploration in the Microstructure of Cognition (MIT Press, Cambridge, MA,1986). 13. Miller III, Sutton and Werbos, (Eds), Neural Networks for Control (MIT Press, Cambridge, MA, 1990).

440 14. 15. 16. 17. 18.

19. 20. 21. 22. 23. 24. 25.

G. Terenzi Rizzolatti and Arbib, Trends in Neurosciences 21, 188-194 (1998). Rizzolatti, Fogassi and Gallese, Nature Reviews Neuroscience 2, 661-670(2001) Rizzolatti and Craighero, Annual Review of Neuroscience 27, 169-192 (2004). Rosen, Anticipatory Systems (Pergamon Press, 1985). Rumelhart, Hinton and Williams (1986) in Parallel Distributed Processing. Exploration in the Microstructure of Cognition, Ed. McClelland and Rumelhart, (MIT Press, Cambridge, MA,1986). Stolzmann, in Genetic Programming 1998: Proceedings of theThird Annual Conference, (1998), pp. 658-664. Sutton, in Proceedings of the Seventh International Conference on Machine Learning, (Morgan Kaufman, San Mateo, CA, 1990), pp. 216-224. Sutton and Barto, Reinforcement Learning: An Introduction (MIT Press, Cambridge, MA, 1998). Widrow and Hoff, IRE WESCON Convent. Rec., 4, 96-104 (1960) Wolpert and Kawato, Neural Networks 11, 1317-1329 (1998). Wolpert, Gharamani and Flanagan, Trends in Cognitive Science, 5(11), (2001). Wolpert, Doya and Kawato, Philosophical Transactions of the Royal Society 358, 593-602 (2003).

DECISION MAKING MODELS WITHIN INCOMPLETE INFORMATION GAMES

NATALE BONFIGLIO, SIMONE PERCIVALLE, ELIANO PESSA Dipartimento di Psicologia, Università di Pavia Piazza Botta 6, 27100 Pavia, Italy E-mail: [email protected]

According to Evolutionary Game Theory decision making in games with incomplete information should be viewed as an emergent phenomenon. However, the adoption of this framework tells us nothing about the concrete modeling of the emergence of decisions within specific games. In this paper we took into consideration the case of Iterated Prisoner Dilemma Game (IPDG). In this regard we compared the outcomes of computer simulations of three different decision making models, two of which implemented through particular neural network architectures, with experimental data coming from observations about the behavior in IPDG of human players. The comparison was based on the use of a Genetic Algorithm, which let us know the best parameter values, for each kind of model, granting for the best reproduction of the observed pattern of experimental data. We found that the best fit was obtained by a model directly taking into account the inner expectancies of each player. This result suggests that the emergence of decision cannot be described by resorting to the simplest models of selforganization. More complex models are needed, including a detailed account of the operation of player’s cognitive system. Keywords: Iterated Prisoner Dilemma Game, decision making, expectancy, neural networks, genetic algorithms.

1. Introduction Decision making has been initially studied through the so-called normative approach, describing the way in which a person should make decisions if he/she would behave in a rational way. This approach is therefore prescriptive, in that it prescribes the principles an individual should refer to when making a rational choice. Later, because of evident failures of predictions made by the normative approach, a second perspective, called descriptive approach, arose, especially influencing psychologists studying thought and decision making [1]. This approach aims at building models that can describe and thus predict the decisional process implied in the choices individuals make, and determining the factors that influence them. The theoretical support for descriptive approach has been given by Evolutionary Game Theory [2,3,4,5,6,7], according to which a decision is nothing but the outcome of a process of emergence characterized by

441

442

N. Bonfiglio et al.

the complex interaction of different factors, as well as of a number of contextual features, in turn connected to the story of individual interactions occurred within a specific game. While this perspective underlines the capital role of a theory of emergence [8,9,10] in describing decision making, it, however, leaves unspecified the criteria to be adopted when choosing particular models of emergence in order to describe decision making processes in given game situations. In this regard, the only strategy so far available seems to be the one of choosing particular decision making models, on the basis only of reasons of mathematical convenience, and to compare their predictions with the observed behavior of human subjects engaged in a real game. This comparison should allow the individuation of the best model, among the ones taken into consideration. However, while this strategy could, at a first sight, appear as natural and easy, we will show that, besides a number of problems of implementation, it could be dangerous, leading to recognize as the “best ones” some models which, in practice, are unsuited to account for the observational data. In this regard, we will consider the decision making processes occurring within the context of incomplete information games, focusing our attention on Iterated Prisoner Dilemma Game. The latter, as it is well known, belongs to the category of incomplete information games. This choice was made for two main reasons: first of all, there is a large amount of literature on it [4,11,12,13,14], and, in second place, the iterated version of Prisoner Dilemma Game (PDG) allows of a number of different equilibrium situations (while one-shot PDG is characterized by only one equilibrium situation). Moreover, the relatively simple structure of IPDG makes simpler both the building of models of players’ behavior and the implementation of experiments on IPDG played by real human subjects. As regards the kind of models used to describe players’ behavior, our choice fell on artificial neural networks, or on computational models equivalent to them, in one way or another. This choice was suggested by a number of considerations, such as: 1. neural network models allow to describe in a simple way emergence phenomena; 2. these models are, in a sense, “universal”; namely a number of studies proved that most models, even of different nature, can be reduced to some kind of neural network; 3. in principle, we could make use of these models to find relationships between their behaviors and phenomena described by more refined models of behavior of biological neurons.

Decision Making Models within Incomplete Information Games

443

Table 1. Payoff matrix for IPDG with human subjects. Player 2 cooperation Player 1

cooperation defection

(5.000 ; 5.000) (30.000 ; -25.000)

defection (-25.000 ; 30.000) (0 ; 0)

In particular, in this paper we took into consideration three different models of this kind, two based on neural networks and one on a computational mechanism which could be expressed through a neural-network-like language. As a consequence of the foregoing discussion, the goals of the present study can be stated as follows: • to compare the performances of three different decision making models related to IPDG, two of which implemented through neural networks and one through a neural-like computational model; • to find which model best reproduces behavioral trends exhibited by human players. 2. Subjects and Experimental Procedure Before testing the models, we performed an experiment to investigate about the actual interactions between human subjects playing IPDG. The experiment was carried out on 30 subject pairs, each one composed by University students. The members of each pair confronted each other on an “economic” version of IPDG, the duration of which was fixed in advance as given by 100 moves. The latter information was not revealed to the subjects, in that, knowing the length of the interaction, the latter could transform the game into a series of single PDGs, each of which has a single equilibrium, defined by reciprocal defection [15]. The task the subjects were to face was that of pretending to be an entrepreneur who must choose to increase or reduce the price of an object. He could earn, loose or have a neutral outcome according to the choice made by the other player. The payoff matrix of this “economic” version is represented in Table 1. 3. The Models As specified before, the models tested through computer simulations are three. The underlying theories, bases, operation, neural networks and structures are described as follows.

444

N. Bonfiglio et al.

3.1. Pessa, Montesanto and Longobardi’s Model (1996) The difference between this model [16] and classical theories lies in the impossibility to describe the individual’s behavior without considering the context it refers to. Each player’s strategic choice changes with time, according to the experience accumulated during the game, and mainly through the interaction with the adversary, mediated also by the players’ cognitive systems. In turn, the mediation depends on three components: • a set of schemes which guides the decision as to what move to make in presence of particular interaction states; • a set of motivational values; • a motivational predisposition which influences the use of schemes in the game situation. In this regard previous studies [17,18] have identified three kinds of predispositions: to cooperation, competition and individualism. Pessa, Montesanto and Longobardi’s Model implies that the choice of the next move is determined on the basis of subjective maximum utility U . The latter is computed through a sum of three products: the previous move ( x is a numerical variable coding the kind of move) times the weight attributed to it ( α ), the value of the previous move of the adversary ( y ) times the weight attributed to it ( β ) and the previous outcome ( k ) times the weight attributed ( γ ). Formally:

U =α x+ β y+γ k Structurally, the neural network associated to the model in question is composed of two layers with feedforward connections (see Fig. 1). The first layer, of input, is composed of three units which code the values of previously introduced variables. The second layer, of output, is composed of two units that respectively code for cooperation and defection. The activation of the output unit is computed according to the “Winner-takes-all” mechanism, so that only one unit can produce an output signal. This unit is that associated with the maximum value of the weighed sum of its entries. The activation has the value 1 , if the winning unit is that of cooperation, while its value is −1 if the defection unit wins. Given that in the model in question the utility function is not fixed in time but varies as a function of the learning process activated by previous game history, a rule of connection weight modification is introduced. This modification takes place through a process based on the production of an association between stimulus and response given by temporal proximity. This mechanism was implemented in the model through the increase in value of the connection weights in the winning units, by means of a fixed quantity δ .

Decision Making Models within Incomplete Information Games

C

Player’s Previous Move

D

Adversary Previous Move

445

Move

Previous Outcome

Fig. 1. Artificial Neural Network of Pessa, Montesanto, Longobardi.

Formally:

wij′ = wij + δ Furthermore, a contemporary decrease of weights of the connections relative to the other output unit takes place, by means of the following law:

wij′ = ( wij − δ 1 ) 2 where: δ1 denotes a further parameter. 3.2. Bonfiglio’s Model (1999) The decisional process model put forward by Bonfiglio [19] can be considered as a development of the previous model by Pessa, Montesanto and Longobardi. Bonfiglio started from a series of hypotheses according to which every player who begins a game has expectations concerning the adversary’s strategy; this expectation often concerns a game model that is highly similar to one’s own, being it the only one known. Consequently, two situations may take place during the game:

446

1.

2.

N. Bonfiglio et al.

Both players start by experiencing the effects of different moves to get a knowledge of the adversary until one of the two, or both, adopts a strategy that is adaptable to the other; in this case, it is hypothesized that a cooperative subject should be more inclined to change his/her strategy to adapt to that of the adversary; a competitive player would instead pursue his/her own goals. Both reach an equilibrium already from the first moves and have no intention of trying alternative ones; they therefore keep unchanged the initial strategies.

The basic hypothesis can thus be formulated in terms of three points: • Expectation; each individual who is in a situation of uncertainty regarding a decisional choice about what behavior to adopt will have expectations about the adversary. • The behavioral model; each individual endowed with expectations regarding the behavior of his/her adversary begins to build a model of behavior, therefore supposing that the other person’s model is similar to his/her own. • Individual differences; each subject can be more or less reluctant to adapt his/her behavior to that of another. This depends on individual differences, on being more or less competitive, on personal motivations regarding a specific situation, on the goals an individual pursues and on his/her investment in the game. Structurally, the neural network in Bonfiglio’s model is more complex than the previous one. Each player is represented by a neural network, the scheme of which is represented in Fig. 2. The expectation at time t is computed on the basis of 5 previous moves played by the player and by the adversary (hence the ten units in the input layer), through a neural network with three layers with feedforward connections. Successively, a comparison is carried through with Expectation at time t − 1 , previously memorized. The comparison between the two expectations, the players’ moves and the payoff obtained by the player, all at t − 1 , concur to the decision making as to the actual move. 3.3. Busemeyer and Stout’s Model This model is generally known as the Theory of Valence Expectation Learning [20,21,22]. It is an adaptation to game situations of the Decision Field Theory (DFT) put forward by Busemeyer and Townsend [23] in 1993. The DFT is a

Decision Making Models within Incomplete Information Games

447

MOVE

Comparison of Expectations Player’s Move (t-1)

Expectation (t )

Expectation (t-1)

Adversary Move (t-1)

Player’s Payoff (t-1)

MEMORY

Figure 2. Structure of Bonfiglio’s artificial neural network (1999).

theory rooted in economics which aims at explaining the decisional process in conditions of uncertainty. The DFT is particularly interesting for its description of the dynamical processes that occur between the moment of the choice and the final decision. This dynamical formulation allows to explain the relationship between choice probability and time; also, it helps to explain the role of the effect of time pressure on the choice probability. Within it the temporal effects in the decisional process are accounted for by resorting to the influence of two components: • a memory of the decision made in the immediately previous time step; • the value attributed to the present situation, considered as stochastically fluctuating. However, DFT cannot be directly applied to IPDG in its original form, and this led to the Busemeyer and Stout model. According to the latter, the subject integrates the gains and the losses experienced in each test into a single affective

448

N. Bonfiglio et al.

reaction called valence. During the tests the player creates expectations relative to the valences of the possible alternatives, through a process of adaptive learning. These expectations then determine (in a probabilistic way) the choices that the subject makes at each test. The determination of the valence associated to the choice of a particular alternative A in the t -th test can be obtained through the following rule:

v A (t ) = w × PA (t ) + (1 − w) × G A (t ) where PA (t ) is the loss following choice A in the t -th test (of course expressed by a negative number), G A (t ) is the gain following choice A in the t -th test, w is a parameter called attention, whose value lies in the interval between 0 and 1, which determines how much sensitive the subject is to losses. The actual determination of the expectation relative to the value obtainable from choice A in the t -th test can be obtained through the following rule:

Asp[v A (t )] = a × v A (t ) + (1 − a ) × Asp[v A (t − 1)] In it, a stands for a parameter that fixes the speed of learning. This too lies between 0 and 1. If the value of a is too close to 1, there are strong and rapid changes, but also fast forgetting; opposite effects are obtained if a has a value too close to 0. Finally, the probability of choosing the alternative A in the (t + 1) -th test can be obtained through the following rule:

PrA (t + 1) =

e{s(t )× Asp[v A (t )]} k

e{s(t )× Asp[ vk (t )]}

Here, s (t ) denotes a sensitivity parameter. If the value of s (t ) is very low, then the choice becomes practically random, while if the value is high, the alternative with the highest expectation will be chosen in a deterministic way. It is further supposed that sensitivity increases with experience, according to the following law:

s (t ) =

t 10

c

where c is a further parameter. The DFT has been applied to IPDG by using: • a different model of decision for each player, with different values for w , a and c ; • different initial expectations for each player;

Decision Making Models within Incomplete Information Games

• • •

449

a probabilistic choice procedure; two possible types of choice: to speak or not to speak (cooperation or defection); the values attributed to each choice are modified on the basis of the payoffs obtained in each move.

Busemeyer and Stout’s model has not been implemented through a neural network, but simply through computational rules. However, it can be shown that their effect could be reproduced by the operation of a suitable equivalent neural network [24]. For this reason this model can be considered as “neural-like”. 4. Simulations As already said, to find the best model of IPDG we compared the sequences of outcomes of games played by neural network pairs – or Busemeyer and Stout pairs – with the ones occurring in games played by human subjects. This comparison required a suitable coding of game outcomes. In this regard each game was described by two vectors, each one with 100 components, coding the moves made by each single player in the different time game steps. The components of each vector were of binary nature, their value being 0 when coding a defection, and 1 when coding a cooperation. The comparison between the games played according to a given model and those played by human subjects was based on a single index C , in turn computed starting from two Bravais-Pearson correlation indices. The formula chosen for computing C was:

C=

(1 + ρ1 ) (1 + ρ 2 ) 4

where ρ1 denotes the Bravais-Pearson correlation coefficient between the series of moves played, within each players’ pair, by the first human player and the one of those played by the first neural network or Busemeyer-Stout player, while ρ 2 denotes the analogous correlation coefficient between the moves played by the second human player and the second neural network, or Busemeyer-Stout player. In this regard we remark that the behavioral models taken into consideration in this paper are all rather complex and contain many parameters. The main problem to solve is that of finding the correct values of the parameters granting for a good reproduction of specific games played by human beings. If this problem could allow of a solution, at least for a specific model, then we could

450

N. Bonfiglio et al.

reasonably assume that the latter would account for human decision making processes in IPDG. On the contrary, if we should recognize that such a solution does not exist, this would mean that none of the previous models can be used to explain the behavior of human players. However, it is to be taken into account that, owing to the complexity of models themselves, it is very difficult to find the correct values of models parameters, if any, by resorting to traditional methods. Therefore we tried to solve the problem through the use of Genetic Algorithms (GA). As it is well known, GA [25,26] are adaptive methods which can be used to solve research and optimization problems. They are based on computational rules inspired to genetic processes occurring in biological organisms. By imitating these processes, genetic algorithms can find solutions for real-life problems, if adequately coded. They work with a population of individuals, each of which represents a possible solution to the problem. Each individual is associated to an adaptation score, the fitness score, according to whether it is or not an adequate solution to the problem. The best individuals can reproduce, cross-breeding with other individuals of the population. This produces new descendants who share some characteristics of each parent. The least fit individuals are less likely to reproduce, and therefore extinguish. A whole new population of possible solutions is therefore produced by the selection of the best individuals of the current generation, who, reproducing among themselves, produce a new group of individuals. This new generation contains a higher proportion of the fit individuals’ characteristics of the previous generation. In such a way, after several generations, the good characteristics are spread to the whole population, in that they are mixed and exchanged with other good characteristics. By favouring reproduction among the fittest individuals, the most promising areas of the research space can be explored. If the GA has been well built, the population will converge towards an optimal solution to the problem. Each individual is also called chromosome, and the parameters are called genes. In this paper, each parameter is coded through a real number. As has been previously described, the GA needs a fitness function, which returns the fitness score to the individual considered. In our case, the function is described through the maximum value, on all games played by human subjects, of the correlation index C previously introduced. The operations carried out on chromosomes concern cross-over and mutation, the first through a random cut, only on individuals who have a fitness

Decision Making Models within Incomplete Information Games

451

Table 2. Average and maximum fitness values of each model. MODEL Pessa, Montesanto and Longobardi (1996) Bonfiglio (1999) Busemeyer and Stout (2002)

AVERAGE FITNESS 0.3901961

MAXIMUM FITNESS 0.4591469

0.4116290 0.3884554

0.4901118 0.4618777

that is superior to the threshold. The second consists of an increase or random reduction of a suitable percentage of single gene values. The fitness threshold under which individuals were eliminated was equal to 90% of the maximum fitness value registered in the whole population at time t . 5. Simulation Outcomes The values coming from comparisons between the games played by human subject pairs and artificial player pairs, as obtained from the best individuals found after applying GA to the models previously described are shown in the Table 2. As it is possible to see, none of the models was able to perfectly reproduce at least one of the games played by human subjects. 6. Conclusions The fact that none of the models taken into consideration can perfectly reproduce at least one of the games played by human subjects shows that a modeling of emergence of decisions, as postulated by Evolutionary Game Theory, is not feasible through simple modeling means. Cheap emergence is not possible! However, the models’ different abilities to reproduce the games must be highlighted. In particular, as Table 2 shows, the best performances occur in neural network models (Bonfiglio; Pessa, Montesanto and Longobardi) rather than in Busemeyer-Stout model. The data seem to suggest (even if in a very weak way) that, going towards a higher complexity of networks, related to the need for taking into account subjects’ expectations, it would be possible to reproduce in a better way the observed subjects’ behavior. Such a strategy, however, implies that, in order to account in a realistic way for the emergence of decisions in incomplete information games, we should resort to a detailed description of the operation of the whole cognitive system of each player, absent within the models taken into consideration. This, in turn, implies again that the hope of describing cognitive emergence by resorting only to simple general principles, expressed by simple computational rules, is vain. Such a

N. Bonfiglio et al.

452

circumstance, on the other hand, is nothing but a consequence of the failure experienced by actual Cognitive Science when trying to describe cognitive processing as an emergent phenomenon. New conceptual and technical tools are needed to make progress in this domain. References 1. H.A. Simon, The Quarterly Journal of Economics 69, 99-118 (1955). 2. R.M. May, Stability and Complexity in Model Ecosystems (Princeton University Press, Princeton, NJ, 1974).

3. J. Maynard Smith, Evolution and the Theory of Games (Cambridge University Press, 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Cambridge, UK, 1982). R. Axelrod, The Evolution of Cooperation (Basic Books, New York, 1984). E. Akiyama, K. Kaneko, Physica D 147, 221-258 (2000). H. Gintis, Game Theory Evolving (Princeton University Press, Princeton, NJ, 2000). E. Akiyama, K. Kaneko, Physica 167, 36-71 (2002). J.P. Cruchtfield, Physica D 75, 11-54 (1994). J.H. Holland, Emergence: From Chaos to Order (Perseus Books, Cambridge, MA, 1998). G. Minati, E. Pessa, Collective Beings (Springer, Berlin, 2006). R. Axelrod, in L. Davis, Ed., Genetic Algorithms and Simulated Annealing (Morgan Kauffman, Los Altos, CA, 1987), pp. 32-41. W. Poundstone, Prisoner’s Dilemma (Doubleday, New York, 1992). J. Andreoni, J.H. Miller, The Economic Journal 103, 570-585 (1993). R. Axelrod, The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration (Princeton University Press, Princeton, NJ, 1997). R.D. Luce, H. Raiffa, Games and Decision (Wiley, New York, 1957). E. Pessa, A. Montesanto, M. Longobardi, in E. Pessa, M.P. Penna, A. Montesanto, Eds., 3rd Systems Science European Congress (Kappa, Roma, 1996), pp.1017-1021. D.M. Kuhlmann, A.F.J. Marshello, Journal of Personality and of Social Psychology, 32, 922-931 (1975). G.P. Knight, S. Kagan, Developmental Psychology 17, 783-790 (1981). N.S. Bonfiglio, General Psychology, 2 (1999). J.R. Busemeyer, I.J. Myung, Journal of Experimental Psychology: General 121, 177-194 (1992). I. Erev, A.E. Roth, American Economic Review 88, 848-881 (1998). J.R. Busemeyer, J.C. Stout, Psychological Assessment 14, 253-262 (2002). J.R. Busemeyer, J.T. Townsend, Psychological Review 100, 432-459 (1993). R.M. Roe, J.R. Busemeyer, J.T. Townsend, Psychological Review 108, 370-392 (2001). D. Goldberg, Genetic Algorithms in Search, Optimization, and Machine Learning (Addison-Wesley, Reading, MA, 1989). S.N. Sivanandam, S.N. Deepa, Introduction to Genetic Algorithms (Springer, Berlin, 2007).

EMERGENCE IN MEDICINE

This page intentionally left blank

BURNOUT AND JOB ENGAGEMENT IN EMERGENCY AND INTENSIVE CARE NURSES

PIERGIORGIO ARGENTERO, BIANCA DELL’OLIVO Department of Psychology, University of Pavia Piazza Botta 6, 27100 Pavia, Italy. E-mail: [email protected]

Burnout phenomenon emerges from a constellation of factors which cannot be described in terms of cause-effect relationships. This study investigated levels of burnout in nurses working in Critical Care Units with a systemic approach, giving evidence of relation between nurses staff burnout and psychosocial workplace factors. The purpose of this study was to examine the relationship between job burnout in emergency and intensive care nurse with specific areas of work life in their organizations, using Maslach and Leiter work life model [23]. A cross-sectional survey was designed using the Italian version of the “Organizational Checkup System” in a sample of 180 Italian nurses. Results showed that high burnout levels were strongly related to high demands, low control, low fairness, lack of social support, and individual disagreement on values in the workplace. High professional efficacy levels were instead correlated to professional reward and leadership involvement. The article concludes by suggesting the possible areas for intervention in order to prevent job burnout and building job engagement. Keywords: critical care nursing staff, job burnout, emergence, systems.

1. Introduction Emergence is a fundamental feature of complex systems and can be thought of as a new characteristic or behavior which appears due to non-linear interactions within systems. Chaotic behavior in systems researches was observed within many fields of science, including biology, physics and social sciences. These newer developments of complexity theories made use of systems theories as well as theories of chaos and complex adaptive systems. Interrelated agents, self-organization, evolution toward the chaos and constant evolution are further characteristics of complex adaptive systems. Complex systems can exhibit multilevel feedback behavior. The complexity theory’s goal is the examination of a whole system and its environment. For example, within social systems, health care organizations are complex adaptive systems and as such, they share common characteristics with

455

456

P. Argentero and B. Dell’Olivo

these systems [25]. Emergent characteristics are dynamic in nature [12]: this means that emergent properties of a social system require continuous attention [17]. Thus, to provide an explanation for the complexity of the problems in health organization, there are a number of different investigative routes: from an individual’s behavior through the behavior of a whole group, workplaces, societies to that of the whole ecosystem. Furthermore, studies on complexity use chaos theory to suggest that the organizations should strive continuously to be at the edge of chaos, because living in a state of continuous chaos is costly. Working in emergency (to not be confused with ‘emergence’ quoted above) and intensive healthcare departments is often chaotic, and for this reason the staff is not always able to deal with the difficulties they may face during their shifts. This often leads to feelings of loss of control. Intensive and emergency care units are stressful workplaces, both for physicians and nurses [11]. Many emergency events, such as a patient with a severe trauma and patients with cardiac arrest, are difficult to manage and stressful for the staff. A further source of stress is the need to make important management decisions in a very short time. Stress associated with chaotic working conditions in hospital environments has been identified as a major cause of burnout, among health care staff. Recent work on burnout has developed new theoretical frameworks that more explicitly integrate both individual and situational factors. This approach of job-person fit, would seem to be an appropriate framework for understanding burnout. In particular Maslach and Leiter [23] formulated a model that focuses on the degree of match, or mismatch, between the person and his job environment. The greater the gap, or mismatch, between the person and the job, the greater the likelihood of burnout; conversely, the greater the likelihood of engagement with work. One new aspect of this approach is that the mismatch focus is on the enduring working relationship people have with their job. Leiter and Maslach [22] describe burnout as a psychological syndrome of Exhaustion, Cynicism and Inefficacy which is experienced in response to chronic job stressors. The first dimension: Exhaustion, measures fatigue without referring to other people as the source of one’s tiredness. The second dimension: Cynicism, reflects indifference or a distant attitude towards work in general, not necessarily to other people. Finally, Professional Efficacy encompasses both aspects of occupational and individual accomplishments. High scores on exhaustion and cynicism, and low scores on professional efficacy, are indicative of burnout.

Burnout and Job Engagement in Emergency and Intensive Care Nurses

457

Job engagement is assumed to be the positive antipode of burnout. As Maslach and Leiter [23] affirmed, Energy, Involvement and Efficacy, are the direct opposites of the three dimensions of burnout. In their view, burnout is an erosion of Engagement, whereby Energy turns into Exhaustion, Involvement turns into Cynicism, and Efficacy turns into Ineffectiveness [23]. According to Maslach and Leiter [23], job engagement is assessed by the opposite pattern of scores on the three burnout dimensions: that is, low scores on Exhaustion and Cynicism, and high scores on Efficacy, are indicative of job engagement. Maslach and Leiter model has brought order to the wide variety of situational correlates by proposing six areas of work-life that encompass the central relationships with burnout: Workload, Control, Reward, Community, Fairness, and Values. Workload refers to the relationship of work demands with time and resources. Increasing workload is consistently related to higher levels of emotional exhaustion [23]. Control incorporates role clarity within the organization that provides a clear understanding of expectations and responsibilities, witch leads to efficiency, decreased stress and autonomy in the work setting [20]. Reward refers to the actual time invested in the work setting and the recognition associated with it. Community includes the quality of social relationship within the organization. Fairness in the workplace involves trust, openness and respect and some extent a quality of managerial support [19]. Finally, Values refers to the congruence between organization’s expectations and those of its employees [23]. Maslach and Leiter [23] argued that low levels on each of these areas of work life reflect a job-person mismatch that, if it persists, may result in low levels of engagement [22]. According to this model, a mismatch in one or more of these areas of work life can result in burnout [22]. Some previous researches found that lower levels of burnout were associated with staff nurses’ perceived autonomy and control over practice [18], while high levels of emotional exhaustion among nurses, were related to time constraints and work overload [15,16,2]. Leiter and Maslach [22] examined the effect of the six areas of work life on burnout and they found that mismatched areas of work life most strongly related to emotional exhaustion were workload, fairness, and control. Cynicism was

458

P. Argentero and B. Dell’Olivo

most strongly related to a mismatch between value and fairness. Personal efficacy was most strongly related to control and values. Moreover Maslach and Leiter model has introduced other four dimensions, than the authors assume to be correlated to burnout, that is: Change and three management processes: Leadership, Skill Development and Group Cohesion. Change dimension is composed by 10 items, that explore changes perception inside organization, for example, regarding quality of leadership and sanitary team members cooperation. The three management processes are composed by 13 items. Leadership refers to the opinion about the management and the quality of communication inside the organization. Skill Development refers to the opportunities of professional development that the organization offers. Finally, Group Cohesion regards the group identity and the work group cohesion. This model, beyond to offer a theoretical framework for the study and understanding burnout, provides direction for improving work environments. According to Maslach and Leiter [23] as mismatches are interrelated, action on any one area will tend to improve at least some of the other ones. Addressing areas of mismatch fosters work engagement, thereby decreasing or preventing burnout. 2. Objective and Hypothesis The purpose of this study is to investigate the relationship between work-life perceptions and burnout levels in nurses working in Emergency and Intensive Health Care Units. It is clear that at this stage it is not possible to propose a complete systemic model of emergence of burnout phenomena. The first goal is therefore the one of individuating the factors which can influence the behavior of subjects and whose synergy makes emergent just burnout itself. This preliminary individuation is here based on the use of typically linear tools. In any case, this preliminary stage will be followed by a further stage in which it will be possible to check the nonlinearity of interactions produced by the factors quoted above. In particular, the objectives of the study are to examine: 1. the differences between nurses operating within four different Healt Intensive Care Departments in relation to job burnout and some work life areas; 2. the relationships between the three burnout dimensions and several areas of work life, organizational changes and process management. Leiter and Harvie [19], also suggested that it is realistic to expect that when nurses feel that they have reasonable workloads, control over their work,

Burnout and Job Engagement in Emergency and Intensive Care Nurses

459

adequate rewarding, positive working relationships and that their values are congruent with organizational values, they are less likely to experience burnout. Thus, in agreement to previous researches [23,20,15,16,2,22] we can formulate the following hypotheses: 1. Emotional exhaustion is related to workload, working control and fairness; 2. Cynicism is related to value, fairness and group cohesion; 3. Professional efficacy is related to values, rewarding and leadership involvement. 3. Methods 3.1. Design A cross-sectional correlational study was carried out in four Emergency and Intensive Health Care Departments in an Italian Hospital. Approval for the use of human subjects was obtained from the Hospital Technical-Scientific Committee of Bioethics. The nurses of each health care unit were informed on the aim of this study, stressing the opportunity to identify the critical work-life areas that needed enhancements so as to improve workplace quality and prevent burnout. In the introductory letter, the confidential and voluntary nature of our research participation was asserted. All the research participants were asked to fill in the informed consent attached to the questionnaire. They were also reassured on the anonymity and on the exclusive treatment of their data for the present research. 3.2. Participants Nurses (N = 180) working full-time or part-time in four intensive health care units in a hospital in Northern Italy: Emergency Care, General Intensive Care, Post-Surgical Intensive Care and Coronary Intensive Care Units, were involved in this research. 3.3. Instrument Job burnout was assessed using the Italian version [7] of the Organizational Check up Survey (OCS) [23], finalized to measure job burnout/engagement and workers evaluation on some aspects of their organizational context. This questionnaire is constituted by 4 Scales and 68 items, in order to measure subjects relation with their workplace and the areas of working life. The first

460

P. Argentero and B. Dell’Olivo

Scale is the Italian version of the Maslach Burnout Inventory - General Survey (MBI-GS) [26]. The MBI-GS includes 16 items distributed among three subscales: Exhaustion (five items; e.g., ‘I feel used up at the end of a work day’); Cynicism (five items: e.g., ‘I doubt the significance of my work’); Professional Efficacy (six items; e.g., ‘I can effectively solve the problems that arise in my work’). All items were scored on a seven-point frequency rating scale ranging from 0 (‘never’) to 6 (‘always’). High scores on Exhaustion and Cynicism and low scores on Professional Efficacy are indicative of burnout. The second Scale, finalized to investigate the Work Life Areas, is the Italian version of the Areas of Work Life Survey [23]. This scale is composed by 29 items, distributed between 6 sub-scales, that measures the six Areas of working life. The six areas (Workload, Control, Reward, Community, Fairness and Values) are distributed on a continuum (lowhigh). The third Scale is composed by 10 items that measure the perception of occurred organizational Changes. At last, the fourth Scale is composed by 13 items that investigate Management Processes, that is: Leadership, that refers to the opinion about one’s own bosses; Skills development, that is the possibilities of professional development that the Organization offers; Work group cohesion. For all affirmations, except the Change items, others Scales were expressed by 1 to 5 score: 1 = I disagree very much; 2 = I disagree; 3 = difficult to decide, 4 = I agree; 5 = I agree very much. As for the Change, the participants were asked to estimate the quality of changes that occurred inside the organization using a 5-point Likert scale ranging from 1 to 5: 1 = very negative change, 2 = negative change, 3 = no change at all, 4 = positive change, 5 = very positive change. After obtaining a written consent from the nurses, and before providing them with the Organizational Check up Survey, nurses were asked to fill in a questionnaire about their sociodemographic data (gender, age, marital status, children) and working conditions (permanent employment, seniority in hospital and weekly working hours). Nurses rated each item and reported their feelings about their job using a 7-point Likert scale ranging from 0 to 6 for the first OCS Scale, while a 5-point Likert scale ranging from 1 to 5 was used for the other three Scales. Each nurses out of the four health care units was rated according to the three burnout dimensions of Emotional Exhaustion, Cynicism and Professional Efficacy.

Burnout and Job Engagement in Emergency and Intensive Care Nurses

461

3.4. Data analyses Data analysis was carried out by means of the “Statistical Package for the Social Sciences” (SPSS) version 13.01. The following statistical analyses were performed on data collected: 1. descriptive analysis, to calculate the frequency distribution, percentages of category variables, the mean values and standard deviations for continuous variables; 2. analysis of variance, using the socio-demographic characteristics of workers and Health Care Units as independent variables, while questionnaire items regarding job burnout dimensions and work life areas, as dependent variable. Tukey’s multiple comparison post hoc analysis was used to compare each nurse group to the others; 3. correlational analysis (Pearson’s r) in order to determine correlations between the four OCS Scales. 4. rank order correlations analysis (Spearman’s Rho) between the three means burnout' measures and work environment variables for each Health Care Unit. 4. Results In total 140 nurses (78%) answered the questionnaires. As regards sociodemographic characteristics, with reference to gender, the majority of the nurses who answered the questionnaire were female (64%). Nurse averaged 35 years of age (SD = 8.28). Almost 54% were either married or cohabiting, and the majority (71%) had not children. The mean length of time spent working in heath care was 20 years (SD = 9.75), of which 10 years (SD = 9.25) were at the current workplace. With regard to the belongings Units Care, nurses were distributed in the following four Departments: Emergency Care (37%), General Intensive Care (27%), PostSurgical Intensive Care (19%) and Coronary Intensive Care (17%). The comparison of the mean values for all nurses in the 13 dimensions of the questionnaire, showed that the most problematic aspects regarded nurses workload (M = 20.40; SD = 3.58) and reward (M = 10.86; SD = 3.54) while the highest balance area was community or social integration (M = 16.89; SD = 3.56), that is the supportive climate between the work group members, that would represent a resource for nurses. The analysis of variance did not find significant differences between mean score on the main variables of the questionnaire in relation to sociodemographic (gender, age, marital status, having children) and working

462

P. Argentero and B. Dell’Olivo Table 1. Means (M) and Standard Deviations (SD) of the Organizational Check up Survey dimensions. PostSurgical Intensive Care M SD Emotional Exhaustion 18.90 4.48 Cynicism 22.70 4.77 Professional Efficacy 20.6 4.51 Workload 20.25 5.48 Control 9.26 2.38 Reward 9.17 2.93 Community 15.71 4.63 Fairness 15.23 5.82 Values 10.93 4.33 Change 28.39 7.85 Leadership 15.18 5.16 Skill Development 8.15 3.38 Group Cohesion 8.80 2.20 OCS dimensions

1 2 3 4 5 6 7 8 9 10 11 12 13

OCS dimensions 1 2 3 4 5 6 7 8 9 10 11 12 13

F

Emotional Exhaustion 6.56 Cynicism 6.45 Professional Efficacy 10.34 Workload 2.27 Control 3.34 Reward 6.62 Community 2.48 Fairness 2.46 Values 5.92 Change 4.97 Leadership 14.57 Skill Development 13.54 Group Cohesion 5.14

p