Multicriteria Decision Aiding Interventions: Applications for Analysts 303128464X, 9783031284649

Citation preview

Multiple Criteria Decision Making

Maria Franca Norese María A. De Vicente y Oliva Irène Abi-Zeid   Editors

Multicriteria Decision Aiding Interventions Applications for Analysts

Multiple Criteria Decision Making Series Editor Constantin Zopounidis, School of Production Engineering and Management, Technical University of Crete, Chania, Greece

This book series focuses on the publication of monographs and edited volumes of wide interest for researchers and practitioners interested in the theory of multicriteria analysis and its applications in management and engineering. The book series publishes novel works related to the foundations and the methodological aspects of multicriteria analysis, its applications in different areas in management and engineering, as well as its connections with other quantitative and analytic disciplines. In recent years, multicriteria analysis has been widely used for decision making purposes by institutions and enterprises. Research is also very active in the field, with numerous publications in a wide range of publication outlets and different domains such as operations management, environmental and energy planning, finance and economics, marketing, engineering, and healthcare. This series has been accepted by Scopus.

Maria Franca Norese • María A. De Vicente y Oliva • Irène Abi-Zeid Editors

Multicriteria Decision Aiding Interventions Applications for Analysts

Editors Maria Franca Norese Politecnico di Torino Turin, Italy

María A. De Vicente y Oliva Universidad Rey Juan Carlos Madrid, Spain

Irène Abi-Zeid Université Laval Québec, Canada

ISSN 2366-0023 ISSN 2366-0031 (electronic) Multiple Criteria Decision Making ISBN 978-3-031-28464-9 ISBN 978-3-031-28465-6 (eBook) https://doi.org/10.1007/978-3-031-28465-6 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Introduction

This book was designed to help Multiple Criteria Decision Aid (MCDA) analysts who are familiar with multi-criteria (MC) methods, but who may lack knowledge or experience in the practical creation and validation of an MC model. The need for procedural and methodological approaches that are derived from practice has long been known and pointed out in the scientific literature (see, for instance, Belton and Pictet 2002; Sodhi and Tang 2008; Howick and Ackermann 2011; Ormerod 2014a; Merad et al 2014). Different approaches and tools, ranging from systems thinking (Mingers 2015) to elements of connections between MCDA and e-government (Bollinger and Pictet 2003), participatory democracy (Matielson et al 2008) and MCDSS (Multi-Criteria Decision Support Systems) (see Belton and Hodgkin 1999), have been proposed and analysed in the literature in terms of benefits for the potential users and the decision aid processes. The aim of this book is to contribute to the literature based on real-life applications by proposing some practical MCDA applications that had been conducted by analysts, some of whom are academic practitioners (as defined in Ormerod 2014b), while some others are consultants or researchers in national or European research centres. A decision aid intervention in an organisation can be complicated for several reasons, and the competences of an MCDA analyst should include how to identify and deal with difficulties and uncertainties by reducing or eliminating such complicated conditions. Attention should be paid to the facilitation of decision aid processes. Good facilitation skills, which ensure full participation of the involved actors while accommodating multiple positions and points of view, are crucial for effective model building (Ormerod 2014b). MC models and methods are used for reflexion processes and can also be used as tools to deal with several difficulties, as well as to increase the participants’ understanding of the problem, the organisational processes, the points of view and values of both the involved people and future users of the intervention results. An MC model is not a “given” a priori but is rather the result of a complicated modelling process. An MC method and the result of its application can aid decision makers, but the essential contribution of interventions is in supporting cognitive learning and communication processes. It is therefore important to stress v

vi

Introduction

the roles that a model can play during a decision aid process (Norese 2016). Explicit but possibly incompletely formalised models are often useful to clarify the decision problem and to favour behaviour that increases the consistency between the evolution of a decision aid process and the actors’ value system (Roy 1996). A list or a classification of the main types of complexity of a decision problem or process that have an impact on decision aid interventions is difficult to enumerate, and the concept of complexity cannot be considered absolute. An element of complexity for one analyst in a certain MCDA intervention may be a routine element for another analyst. Therefore, some examples of complexity, in relation to the nature of the problem, the decision process and its actors, the knowledge acquisition and use in the decision aid process, as well as to the modelling activities are briefly presented hereafter and further described in the various chapters. The main aim of each chapter is not to describe how the MC methods were used or how they should be used, but rather to show how these models and methods, when they are used to identify and control specific uncertainties and difficulties, can be adopted to reduce complexity and to ensure a harmonious and a rigorous MCDA intervention.

Main Complexities and Difficulties A decision problem can be complex because it is new and unique, therefore knowledge and data are not available. It could also be a familiar recurrent problem that is different from what had been encountered in the past. A decision problem can be classified as programmable or non-programmable and, as a more general and synthetic distinction, structured, partially structured, ill-structured or unstructured. Psychological, logical, material, organisational and/or political obstacles can be present. A decision problem is frequently presented, by the client to the analyst, as at least a partially structured one. However, it is often the case that the problem is not as structured as initially thought or that the understanding of its nature evolves during a decision aid intervention. In this kind of situation, the main difficulty is the implementation of an action oriented towards facilitating the definition or re-definition of a problem, rather than the choice of the correct method or data acquisition. When the problem is not new, the long history of an old, pre-existing decision process and the presence of other intersecting processes imply the need to know these elements before any technical proposal is made or intervention is planned. The evolution of an actors’ network during the decision process and sometimes the unwillingness of involved actors to change their perspectives make the situation even more complex. In MCDA, a multiplicity of points of view is the core element, and qualitative elements are accepted in MC models together with quantitative components. However, the involvement of decision maker(s) and/or actors should not be taken for granted at the beginning of an MCDA intervention and can be quite an achievement. The involvement of some actors may only be apparent and, in certain cases, some actors might not be able to or even want to be involved. The

Introduction

vii

organisational culture might not be of a participative kind or a conflictual situation may be present. A decision process can include actors from different organisations, or even the whole territory. The different visions are not always explicit and have to be identified. When a decision process involves different levels or departments of an organisation, or different organisations, their a priori relationships (hierarchical and institutional or personal, in relation to previous specific experiences, with results that can positively or negatively impact the decision process) can have a notable impact on the work and increase the complexity of any intervention. MC modelling is influenced by specific difficulties that have an impact on the generation of possible actions, on the elaboration of a consistent family of criteria, and on their weighting and action evaluation. In real-world studies, defining the possible actions and consistent criteria represents the major part of the work of an analyst (Bouyssou 1990). These two "activities" should not be considered as separate and consecutive steps (Corner et al 2001). Very often, the search for a legible, operational and consistent family of criteria leads an analyst to reconsider the definition of some criteria, to introduce new ones into the family and to aggregate some of them, and so on. Thus, the choice of a family of criteria interacts with the construction of the various criteria (Bouyssou 1990). When the definition of a set of actions is progressively elaborated during the course of the decision aid process, intermediate results can appear during the process and modify the nature of the actions, or a decision problem can arise in a naturally changing environment (Vincke 1992, Siebert and Keeney 2015).

Content and Structure This volume, from the Springer series on “Multiple Criteria Decision Making”, was proposed some time ago with the idea that each chapter would describe at least one MCDA intervention, and the differences between the chapters would not mainly be in the used MC method or application sector, but rather in the specific complexity of the decision problem and the approach adopted to identify and deal with a complicated situation. The authors’ work environments and roles (whether more oriented to research or practice) have generated different visions and languages to transfer the competences acquired during MCDA interventions in organisations. A not easy interaction with decision makers and actors of the decision process is described in all the cases, together with the adopted approaches, communication efforts and results. Time is often indicated as a critical resource, and the complexity of some problem situations, which implied long and expensive interventions, is described together with the main activities and tools adopted to reduce or control specific elements of complexity. Some of the authors of the chapters proposed a case analysis and underlined how difficulties and obstacles had been recognised in the MCDA intervention and dealt with. Other authors offered an analysis of decision problem classes and described the

viii

Introduction

approaches that had been (and are) frequently adopted in these cases, the weakness of these approaches and related procedures, as well as the reasons underlying an MCDA intervention and the implementation of specific methodologies. The usefulness of tools that facilitate understanding and communication in the different steps of an intervention is described in all the chapters. A topic that is associated with the quality and the different possible outcomes of an MCDA intervention is analysed in the first chapter (What are the Results of an MCDA Intervention? Some Reflections) by M.F. Norese. These results may be identified in the answers to questions posed by actors involved in a decision process, but also in outcomes internal to the intervention. A framework of this topic analysis is proposed, on the basis of some interventions in relation to different decision problems and organisations, and it is used to describe how some results can be acquired, in relation to different problem situations and their main difficulties.

Case Analyses The need to reorganise the urban postal network in Switzerland led to the activation of a decision-making process that involved a private consultancy firm for years. The problem was related to an aspect of the reorganisation of the overall Swiss postal network, which was itself included in the reorganisation of the Swiss Post. Therefore, all the technical actions and results, during the MCDA intervention, were implemented in a broader social, political and economic context and are analysed in the second chapter (Spatial decision support to reorganise the Swiss postal network) written by F. Joerin. A temporary working group was created at the Fuenlabrada Hospital in Madrid to aid the administrators of the hospital in managing a change in the procedure of the economic management of patients. Some MCDA analysts were involved in the working group, the MC tools were viewed with interest and open-mindedness and a decision support system was elaborated for the frequent technological decisions. An ex-post analysis of this intervention produced some reflections, which are synthesised in the third chapter (A decision aiding methodology to compare patient classification systems) by M. de Vicente y Oliva. In the fourth chapter (A decision-aiding tool for the choice of road pavements and surfacing), A. Fiordaliso, O. Pilate and M. Pirlot analyse some unusual elements of an MCDA intervention that started as a master thesis and became a research project to develop a decision-aid tool for a Ministry, in Belgium. Some elements of uncertainty and complexity, which were recognised and underlined during the intervention, are analysed in the chapter. In the fifth chapter (Multicriteria decision aiding for a shift towards best environmental practices in agriculture, with a focus on viticulture), F. Macary describes a project adopted to study the use of pesticides and their impacts on ecosystems in the Bordeaux region, in France. The involvement of three national Agencies and the close collaboration with the stakeholders, and above all with the professional

Introduction

ix

winegrowers over the five years of the project, were essential to face several critical difficulties and to legitimate the project and its results. New practices of more sustainable agriculture, with less impact on the environment and human health, were proposed to be used as a basis for decision making. Two MCDA interventions, conducted with the City of Quebec to support problem solving in strategic decisions contexts, are described in the next chapter (Multicriteria Decision Aiding—An overview of some challenges of interventions in real life applications) by I. Abi-Zeid, F. Marleau Donais and J. Cerutti. The sixth chapter illustrates an action research context that brings together players from business, government and the academic community to exchange and transfer knowledge and develop innovative solutions to problems inherent to the City of Quebec through the creation of an urban laboratory. A comparison of the two cases (in the fields of transportation and urban planning) points out the main difficulties that were encountered and some insights, based on the global experience of the Université Laval team in conducting socio-technical interventions.

Analysis of Decision Problem Classes The constructivist approach that characterises an MCDA application consists in considering concepts, models, procedures and results to envisage them as suitable tools to establish a dialogue, develop opinions and visions, knowledge pieces and working hypotheses in order to evolve towards a recommendation (Roy 1993). In the seventh chapter (Contrasting applications in environmental planning and public procurement), J. Pictet and D. Bollinger describe these tools and their methodological uses in relation to two classes of cases, the former of which was associated with strategic decisions in environmental planning, while the latter was associated with a new legislative context in Switzerland, on the Public procurement. A methodological framework of public policy evaluation and conflict management is described in the next chapter (Social Multi-Criteria Evaluation of Policy Options) by G. Munda. A public policy process should consider civil society and future generations along with policy objectives and market conditions. Only an integration of different scientific approaches can be used to face these components of complexity and the closely related difficulties. Some lessons that emerged from interventions in different geographical contexts are proposed in the eighth chapter. An MCDA application in the regional planning context, where the legislative corpus continuously evolves and the planning process must include the present and future worldviews of actors and citizens, is a challenge that is described in the ninth and the last chapter (GIS based/MCDA platform for strategic planning and complexity management in conflictual socio-ecosystems). The experiences and practices of J.F. Guy and J.Ph. Waaub in the field generated a new prospective framework and the use of formalisation and adapted modelling tools to address complex decision problems. A case study approach was chosen to illustrate the methodology, which

x

Introduction

was intended for both experienced planners and new practitioners who need new tools to cope with the ever-growing complexity of the planning process. Maria Franca Norese María A. De Vicente Y Oliva Irène Abi-Zeid

References Belton V, Hodgkin J (1999) Facilitators, decision makers, D.I.Y. users: is intelligent multicriteria decision support for all feasible or desirable? Eur J Opl Res: 113(2): 247-260. https://doi.org/10.1016/S0377-2217%2898%2990214-4 Belton V, Pictet J (2002) Talking about the practice of MCDA. In: Bouyssou D, Jacquet-Lagrèze E, Perny P, Slowinski R, Vanderpooten D, Vincke P (eds) Aiding decisions with multiple criteria. Essays in Honor of Bernard Roy, Kluwer, Dordrecht: 71-88 Bouyssou D (1990) Building criteria: a prerequisite for MCDA. In: Bana e Costa CA (ed) Readings in multiple criteria decision aid. Springer-Verlag, Heidelberg: 58-80 Corner J, Buchanan J, Mordecai H (2001) Dynamic decision problem structuring. J Multi-Crit Decis Anal 10(3):129-141. https://doi.org/10.1002/mcda.295 Howick S, Ackermann F (2011) Mixing OR methods in practice: past, present and future directions. Eur J Opl Res 215(3): 503–511. https://doi.org/10.1016/j.ejor. 2011.03.013 Matielson D, Ekenberg L, Ekengren A, Hokby T, Lidén J (2008) Decision process support for participatory democracy. J Multi-Crit Decis Anal 15(1-2): 15–30. https://doi.org/10.1002/mcda.406 Merad M, Dechy N, Llory M, Marcel F, Tsoukiàs A (2014) Towards an analytics and an ethics of expertise: learning from decision-aiding experiences in public risk assessment and risk management. EURO J Decis Process 2 (1-2): 63-90. https://doi.org/10.1007/s40070-013-0022-5 Mingers J (2015) Helping business schools engage with real problems: The contribution of critical realism and systems thinking. Eur J Opl Res 242(1): 316–331. https://doi.org/10.1016/j.ejor.2014.10.058 Ormerod RJ (2014a) Critical rationalism in practice: Strategies to manage subjectivity in OR investigations. Eur J Opl Res, 235(3), 784–797. https://doi.org/10. 1016/j.ejor.2013.12.018 Ormerod RJ (2014b) OR competences: the demands of problem structuring methods. EURO J Decis Process 2(3-4): 313–340. https://doi.org/10.1007/ s40070-013-0021-6 Roy B (1993) Decision science or decision-aid science? Eur J Opl Res 66(2): 184-203. https://doi.org/10.1016/0377-2217(93)90312-B

Introduction

xi

Siebert J, Keeney RL (2015) Creating more and better alternatives for decisions using objectives. Oper Res 63(5): 1144-1158. https://doi.org/10.1287/opre.2015. 1411 Sodhi MS, Tang CS (2008) The OR/MS Ecosystem: Strengths, Weaknesses, Opportunities, and Threats. Oper Res 56(2): 267–277. https://doi.org/https://doi.org/10. 1287/opre.1080.0519 Vincke P (1992) Multicriteria Decision-Aid. Wiley, Chichester

Contents

What Are the Results of an MCDA Intervention? Some Reflections . . . . Maria Franca Norese 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 An Analysis Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Strategic Decision Problems and DA Interventions . . . . . . . . . . . . 2.2 Technical Problems and DA Interventions . . . . . . . . . . . . . . . . . . 2.3 Aims and Results of a DA Intervention . . . . . . . . . . . . . . . . . . . . 2.4 An Application of the Analysis Framework . . . . . . . . . . . . . . . . . 3 The Main Factors that Influence Decision Aid Interventions . . . . . . . . . 3.1 Time Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 A Collective Design of an Integrated Land Monitoring System § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 How to Improve Financing Actions in the Public Sector § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 The Possible Roles of a DA Analyst . . . . . . . . . . . . . . . . . . . . . . 3.2.1 A Project Selection Case Pertaining to the Public Administration § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Project Selection in a Process of IT Innovation § . . . . . . . . 3.2.3 A DA Analyst Involved in a Roundtable of Experts as an Expert in Methods § . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Decision System and Its Dynamics . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 A New Service to Improve Airport Revenues from off-Flight Services § . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 An MCDA Process that Transforms Data into Information § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 MCDA Interventions: Some of the Main Difficulties and Results . . . . . . 4.1 Technical Decision Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 A Simple Tool to Reflect on and Structure an MC Model § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Expected and Unexpected Results in a Process of Project Selection § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 3 4 4 5 6 7 7 8 9 11 11 12 14 15 16 18 19 19 21 22 xiii

xiv

Contents

4.2

Strategic Decision Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 The Urban Postal Network Reorganisation in Switzerland § . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 MCDA Analysts and the City of Quebec § . . . . . . . . . . . . 4.2.3 Methodological Approaches to Public Contexts of Decision-Making § . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Decisions for Complex Policy Problems . . . . . . . . . . . . . . 4.2.5 Research Projects and Strategic Decision Problems . . . . . . 5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23

Spatial Decision Support to Reorganize the Swiss Postal Network . . . . . Florent Joerin 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 The Decision Support Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Sources of Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 The Main Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Definition of Objectives and Criteria . . . . . . . . . . . . . . . . . . . . . . 2.4 Evaluation of Actions’ Performances . . . . . . . . . . . . . . . . . . . . . . 2.5 Synthesis by Multicriteria Analysis . . . . . . . . . . . . . . . . . . . . . . . 2.6 Design and Evaluation of Network Alternatives . . . . . . . . . . . . . . 3 A Decision Support Approach in a Decision-Making Process . . . . . . . . . 3.1 Who Decides What? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 What Role for Information in the Decision-Making Process? . . . . . 3.2.1 The Construction of a Complex Problem . . . . . . . . . . . . . . 3.2.2 Interaction with Information . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Credibility and Scientific Legitimacy of Decision Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Annexe 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35

A Decision Aiding Methodology to Compare Patient Classification Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . María A. de Vicente y Oliva 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 The Problem and the Organizational Context . . . . . . . . . . . . . . . . . . . . 2.1 Relationships and Communications . . . . . . . . . . . . . . . . . . . . . . . 3 Problem Formulation and Modelling Process . . . . . . . . . . . . . . . . . . . . 4 The Implementation of the ELECTRE III Method: Analysis and Validation of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Main Difficulties Encountered During the DA Intervention . . . . . . 5 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24 24 25 26 27 30 32

35 37 37 38 39 40 44 46 48 49 52 52 54 57 58 60 62 65 65 67 68 69 73 75 76 77

Contents

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Antonio Fiordaliso, Olivier Pilate, and Marc Pirlot 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Context of the Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 The Working Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Goal of the Management of the MET . . . . . . . . . . . . . . . . . . . . . . 2 Highlights of the Modeling Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 The Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Categorization of Road Works . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Choice of a Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Evaluation Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Qualitative Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Quantitative Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Tradeoffs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Revision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Tuning Based on RW Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Computing the Value Associated to an RPS for a Given RW . . . . . 4.2 RW-Dependent Tradeoffs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Logical Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Validation by Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 The Decision-Aiding Software EVAL-MET . . . . . . . . . . . . . . . . . . . . . 6.1 Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Using the Decision-Aiding Tool EVAL-MET . . . . . . . . . . . . . . . . 6.3 The Acceptance Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Looking Back on the Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Strengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Difficulties and Process Weaknesses . . . . . . . . . . . . . . . . . . . . . . 7.3 Success or Failure? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Multicriteria Decision Aiding for a Shift Towards Best Environmental Practices in Agriculture, with a Focus on Viticulture . . . . . . . . . . . . . . . Francis Macary 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Project Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 The Research Institutes and Different Stakeholders Mobilised for the Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 A Multicriteria Analysis of Decision Aiding . . . . . . . . . . . . . . . . . . . . . 3.1 Step 1. Assessing of the Risk of Pesticides Reaching a Viticultural Watershed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Step 2. Evaluation of the Environmental and Socioeconomic Performances of Viticultural Systems . . . . . . . . . . . . . . . . . . . . . .

xv

79 79 80 80 81 82 82 83 84 85 86 88 94 96 98 99 99 100 100 102 104 104 105 107 108 108 109 110 110 116 119 119 121 126 128 129 131

xvi

Contents

3.3

Step 3: Development of Realistic Scenarios of Production Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Multicriteria Models, Methods, and Results . . . . . . . . . . . . . . . . . . . . . 4.1 Step 1: Model and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Steps 2 and 3: Model, Methodology, and Results . . . . . . . . . . . . . 4.2.1 Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Parameters of the Model . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Applications of the Methods and Results of Step 2 . . . . . . . 4.2.4 Assessment of the New Realistic Scenarios of Production Systems and Results of Step 3 . . . . . . . . . . . . . . . . . . . . . 5 Main Difficulties and Some Recommendations for this Kind of Complex Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 The Subject of Pesticides as a Very Sensitive Area . . . . . . . . . . . . 5.2 Presentation of Results and Modelling Assumptions . . . . . . . . . . . 6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Principle of Multicriteria Methods by Outranking . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions . . . Irène Abi-Zeid, Francis Marleau Donais, and Jérôme Cerutti 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Intervention Process Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 First Intervention: Ranking Streets According to their Potential to Become Complete Streets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Decision Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Organizational Context and the Participants . . . . . . . . . . . . . . . . . 3.3 Main Aim of the Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Post-Project Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Second Intervention: Rethinking the Role and Jurisdiction of a Municipal Planning and Conservation Commission . . . . . . . . . . . . 4.1 Decision Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Organizational Context and the Participants . . . . . . . . . . . . . . . . . 4.3 Main Aim of the Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Post-Project Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contrasting Applications in Environmental Planning and Public Procurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jacques Pictet and Dominique Bollinger 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Applications in Environmental Planning . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

133 135 137 142 143 146 147 148 149 149 151 152 153 153 155 161 161 162 165 165 166 166 168 171 172 172 174 174 176 179 180 183 187 187 188 188

Contents

xvii

2.2

Content Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.4 Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.5 Aggregation Method and Results Presentation . . . . . . . . . . 2.3 Process Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Public Procurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Content Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Process Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Spiral of Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Aggregation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Overall Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 New Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 MCDA and Democracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7 Future Areas of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

189 189 191 191 192 194 196 199 201 201 204 206 207 208 208 209 210 210 211 212 213 214 215

Social Multi-Criteria Evaluation of Policy Options . . . . . . . . . . . . . . . . . Giuseppe Munda 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Democracy and Policy Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Down to Earth Social Multi-Criteria Evaluation Processes . . . . . . . . . . . 4 Mathematical Approaches in the Social Multi-Criteria Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

217

GIS Based/MCDA Modelling for Strategic Environmental and Social Assessment of Land-Use Planning Scenarios in Conflictual Socioecosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jean-Francois Guay and Jean-Philippe Waaub 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 A Brief Overview of One Complex Problem . . . . . . . . . . . . . . . . . . . . . 2.1 Legal Apparatus by the Expressions of Legislative Corpuses . . . . . 2.2 Territorial Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Complexities and Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 An Adapted Methodology Combining Tools . . . . . . . . . . . . . . . . . . . . . 4 Real-World Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

217 218 221 227 230 231

235 235 237 237 239 240 241 244

xviii

Contents

4.1

Selection of the RCM, Logic of the Planning Process, and Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Problem Structuring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Choice and Decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Concluding Remarks and Recommendations for Planners . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

244 247 253 258 260 262

What Are the Results of an MCDA Intervention? Some Reflections Maria Franca Norese

1 Introduction Decision aiding is the activity of a person who, through the use of explicit, but not necessarily complete formalised models, helps another person or group of people to obtain elements of responses to questions posed by actors involved in a decision process. These elements work towards clarifying the decision and usually towards recommending, or simply favouring, behaviour that will increase the consistency between the evolution of the process and the actors’ objectives and value system (cf Definition 2.2, Roy 1996). At least two actors are involved in a decision-aiding process, that is a client and an analyst, where the former describes a “problem”, and the latter tries to give him/her advice. The client, who may be a person, a group of people, or an organisation, is clearly the decision maker in relation to the decision aid intervention, its development and its results. The role of the client in a decision process may be central, minimal (e.g. a spectator with the aim of acquiring an active role in the process) or marginal because the process involves several organisations in an intricate way. Other actors in the decision process, who may have different concerns and stakes, may also be involved in the decision-aiding process (Tsoukiàs 2001). The use of formal methods, i.e. methods that reduce ambiguity, which is typical of human communication, characterise the MultiCriteria Decision Aid (MCDA) methodology (see the EURO Working Group MCDA web site http://www.cs.put. poznan.pl/ewgmcda/). MCDA adopts a constructivist approach to help a client build models based on the communicative conception of rationality, by means of a “discussion” between an analyst and the client to construct representations of the client’s problem (Genard and Pirlot 2002; Tsoukiàs 2007; Meinard and Tsoukias M. F. Norese (✉) Politecnico di Torino, Turin, Italy e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_1

1

2

M. F. Norese

2019). MCDA adopts this approach because a clearly defined problem often does not exist in real-life applications, and the way a problem is formulated cannot be totally objective, but it is expected to evolve throughout the decision-making process and above all throughout the decision aiding process. Following the path of constructivism consists in considering concepts, models, procedures, and results as suitable tools to develop convictions (Roy 1993), i.e. opinions and visions, knowledge pieces and working hypotheses, to allow them to evolve towards a recommendation, as well as to establish a dialogue based on these elements. There is not only one set of tools suitable for clarifying a decision, nor is there a single “best” way to make use of them (Roy 1993). From such a perspective, the outcomes of an MCDA process are not only a model and the results of a method that has been applied to the model; they are also associated with an appropriate use of interactive procedures that can be seen as tools which contribute to eliminating questions, solving conflicts, transforming contradictions and to destabilising certain convictions (Roy 1987). An MCDA intervention not only generates hard outcomes, such as models, method applications, and/or decision support systems, but also internal “soft” outcomes (Barcus and Montibeller 2008), which determine the decision aiding quality and the achievement of its ultimate aim (Norese 2016b, 2020a, b). Going back to Definition 2.2 (Roy 1996), the aim of such a process is to respond to a question posed by a client, even though a clearly defined question is not always possible. But what adequate uses of concepts, procedures, models, and methods produce results that make the response to a client appropriate? What conditions or behaviour generate these kinds of results? What are the different results of a Decision Aid (DA) intervention and its activities? What roles can MC models and methods play in generating these results? An analysis of these questions and some answers are proposed in this chapter, above all in relation to some direct experiences with different problems and organisations but also to the cases that are proposed in the book. The chapter includes, in the first section, an analysis framework, to distinguish between some situations of MCDA intervention and their main characteristics and to propose a synthetic definition of the DA “result” concept. The second section describes the main factors that influence a DA analyst’s behaviour and results, and it proposes some synthetic case descriptions (which are introduced by the symbol §), in relation to different factors and problem situations. The last section applies the framework to analyse the difficulties, the uses of specific tools and the (not always expected) results of the proposed DA interventions presented in this and the other chapters of the book. Some reflections on the meanings that can be associated to the word “result” are proposed in the conclusions.

What Are the Results of an MCDA Intervention? Some Reflections

3

2 An Analysis Framework Decision aiding develops in different ways that are closely connected to the nature of the decision problem and the main intervention aim (Norese 2016a, 2020a, b). A specific use of suitable tools, in relation to the approach adopted in a DA intervention (Meinard and Tsoukias 2019), can facilitate dealing with complexities and uncertainties and can ensure validity and legitimacy of an intervention. There are various types of decision problems, which are often multifaceted and not always clear at the beginning. To facilitate an analysis of a DA intervention and its results, a simple but useful distinction may be made between technical decisions and strategic decision problems (as they are defined in Mintzberg et al. 1976). Another distinction that may simplify this analysis is between the cognitive and operational aims of a DA intervention. A cognitive aim is mainly associated with strategic decision problems, whilst an operational approach is almost always required in relation to technical decisions. An operational aim is sometimes associated with a strategic problem, whilst a technical problem more rarely implies an intervention with a cognitive aim. The different lines that connect the Main intervention aim and the Decision problem in Fig. 1 indicate almost “standard” situations (continuous lines), the less frequent requests of an operational approach, above all of data structuring (black dotted line), and rare cognitive requests, above all of knowledge structuring, for technical decisions (grey dotted line). However, the analyst’s behaviour is also connected to certain factors, whose effects change in relation to the nature of a DA intervention. The time factor, the dynamics of the involved actors and the role that the DA analyst plays in the DA process are the most important. The nature of a strategic or technical decision problem is described hereafter in relation to DA intervention situations, and their main characteristics. The different results of a DA intervention are then analysed, and an application of the analysis framework is proposed, in relation to only one factor, that is, the role that the DA analyst plays in the DA process. The main factors that influence the behaviour of an analyst and the different hard and soft outcomes of a DA intervention are analysed in the next section.

Main intervention aim Cognitive

Operational

Main activities

Decision problem

Problem understanding

MC modelling and method application

Strategic

Technical

Fig. 1 Four different situations of DA intervention and main requests and activities

4

2.1

M. F. Norese

Strategic Decision Problems and DA Interventions

A strategic decision problem is described in (Mintzberg et al. 1976) as being “characterised by novelty, complexity, and openendedness, and by the fact that the organisation usually begins with little understanding of the decision situation it faces or the route to its solution, and it has only a vague idea of what that solution might be and how it will be evaluated when it is developed”. A strategic decision problem is often connected to introducing and drafting innovation or defining strategies that may imply change. The basic structure of an “unstructured” strategic decision problem (Mintzberg et al. 1976) includes three phases that recall Simon’s Intelligence-Design-Choice decision process model (Simon 1960, 1991). A DA intervention is often required in the Intelligence phase, where time limits are not very restrictive because a decision is not imminent. The analyst’s action is above all motivated by the cognitive aim of understanding a new and unstructured problem situation, clarifying and formulating the problem, and identifying an action space, in relation to constraints and opportunities. Knowledge acquisition and its use, in relation to the cognitive aim of a specific intervention, are important activities that require time and effort. The integration of different methods and the use of specific tools to facilitate interactions between analysts, actors, and knowledge sources are very frequent in such problem situations. A strategic decision process often implies the involvement of experts with different backgrounds and different languages. Reducing ambiguities and misunderstandings and generating a shared space that facilitates communication may be the first actions of a DA analyst. A DA intervention with an operational aim is possible in strategic problem situations when a large amount of data is available, but there is not enough knowledge to transform these data into information because the situation is new. An MCDA intervention can be required to give structure to these data and to generate different working hypotheses, without being involved in the strategic problem.

2.2

Technical Problems and DA Interventions

A technical decision often implies an improvement of existing procedures, policies, and/or strategies, the allocation of tangible or intangible resources, as well as recurring selection or prioritisation procedures that require operational support. A technical decision problem is sufficiently clear and at least partially structured, whilst information that is consistent with the problem is present, although not immediately available, in the organisation(s) involved in the decision process. A DA intervention with an operational aim is often required in the Design and Choice phases of Simon’s model, where a technical decision is pre-defined and may be urgent, or in a postdecisional situation, when at least one decision has already been made, but the resources may have to be identified and mobilised and the decision implementation

What Are the Results of an MCDA Intervention? Some Reflections

5

may require operational support. Knowledge about existing procedures, policies, and/or strategies facilitates a DA analyst in the required operational activities. The main activities of an analyst in relation to a technical problem are the organisation of knowledge elements, modelling, method application, and the analysis and validation of the results, in interaction with the problem owner(s). However, the DA process is not always linear in these situations, and non-essential but useful activities of knowledge acquisition and structuring can be implemented (see some of the cases presented in Norese 2020b). Operational complexity can sometimes be associated with a need for relationships with the sources of the knowledge/information, who, in certain cases, are also the actors in the decision process. Relationships require time and do not always produce results. This situation is not so frequent and becomes possible when the client proposes a well-structured problem and asks for structured knowledge, but the real problem is not so structured. Any result whatsoever depends on the relationship between the client and analyst as well as on the space of action of the analyst, who can (and should) use specific tools as soon as possible to identify the actual nature of the decisional problem and then decide how knowledge should be acquired and used.

2.3

Aims and Results of a DA Intervention

In a DA intervention, “it is important to be aware of the myth that such an intervention can provide a so-called ‘right’ answer, through an ‘objective’ analysis, which will relieve decision-makers from the responsibility of making difficult or complex judgements”. An intervention may instead be aimed “at helping decisionmakers and other involved actors learn about the issues and problems they are dealing with, as well as about their values and judgments, which have, of course, a subjective nature” (Bana e Costa et al. 2006). Important and innovative decisions are difficult to make because they are inherently complex. Managing such complexity is of critical importance, and formal modelling may provide valuable decision support. An essential outcome of an MCDA intervention is to shed light on objectives and values, and to help develop decision makers’ preferences—as well as to support further learning about the problem that they have to cope with (Barcus and Montibeller 2008). The expected results of a DA intervention are above all associated with the declared main aim of the intervention, whilst a multiplicity of secondary aims, which are not always explicit, can characterise an intervention and influence an analyst’s behaviour. In a simplified distinction between cognitive and operational aims, cognitive aims may be those that reduce uncertainty and orient knowledge acquisition and problem formulation, as well as analyse, validate, or improve knowledge and/or information, whilst operational aims may be associated with requests for outcomes that should be usable quickly (e.g. the best compromise between technical activities, a selection of candidate solutions, the choice of an

6

M. F. Norese

innovative product or system, and so on). A combination of cognitive and operational aims is not an outstanding aspect in DA processes. Some results are not necessarily associated with the main aim, and they can sometimes be unexpected. Certain outcomes are the conclusions of an analysis and recommendations of an intervention, whilst others need to be attained during the decision aid process. An application of a method to a draft model produces preliminary results that may play the essential role of legitimating the behaviour of the DA analyst, clarifying the problem situation, and allowing the intervention to be completed and the aims to be achieved. A result that often needs to be acquired is the activation of an organisational learning process for a client and the involved organisations. Another kind of learning process should be an essential result for a team of DA analysts. Such a process implies an evaluation of the robustness of an adopted approach, and not only of the proposed recommendations, and of the capability of coping with the main identified weaknesses and complexities (Norese 2016a and b; Marleau Donais et al. 2021). A failure or poor results may sometimes be more useful than positive ones, if an approach oriented towards reflection and learning is adopted, and the experiences are analysed and reflected upon to extract lessons for future interventions. Hard and soft outcomes should be recognised and analysed, even though a long-term horizon often makes such an analysis difficult. A distinction needs to be made between immediate and medium-long-term results in the decision and learning processes that an intervention facilitates or activates, and between the outcomes of a DA process at an individual level (the analyst), group level (the technical team of DA analysts and/or the involved experts and actors) and at an organisation level (the organisation context of the client, the involved organisations and the organisations and people that are impacted by the decisions) (see Ormerod 2014).

2.4

An Application of the Analysis Framework

A DA intervention requires the use of specific tools, procedures, and methods, whose application produces outcomes that then have to be validated and legitimated in order to be used. Some factors generate difficulties and require the specific use of tools to reduce and constrain the analyst’s action in a DA intervention. The effects of these factors are described in the next section in relation to a number of DA interventions. An application of an analysis framework is here proposed in relation to the main aim of the intervention, the nature of the decision problem, and to only one factor, that is the role that the DA analyst plays in a DA process (see Table 1). This factor is described in Sect. 3.2, and some details are given for three MCDA intervention cases. A simple distinction is here proposed between an analyst who is part of (IA) or external to (EA) an organisation that is involved in a decision problem. These different roles can facilitate or make the analyst’s interactions with the actors of

What Are the Results of an MCDA Intervention? Some Reflections

7

Table 1 DA interventions and the need for interactions and tools Main aim Cognitive

Operational

Decision problem Strategic

Internal (IA) or External Analyst (EA)/Interactions IA/ easily guaranteed

Strategic

EA/ difficult, but essential

Technical

Essential, with information sources and actors EA/ difficult

Strategic Strategic Technical

IA/ possible, but not always needed Minimal, mainly with decision maker(s)

Need for tools that generate some expected results To guarantee shared knowledge at an organisational level To facilitate interaction and communication To synthesise information and formally develop and evaluate possible actions To facilitate interaction and evaluate strategies To formally develop and evaluate possible strategies Application of an MC method to a model. Result analysis

the DA process and/or the sources of knowledge and information more difficult. Their different impacts on the intervention, in relation to its nature, orient the choice of tools and their use. Tools that generate a communication space, such as problem structuring methods and simulation tools, but also MC models and the application of MC methods (only explicitly indicated in the last line of the schema), can be used in relation to the different expected results listed in Table 1.

3 The Main Factors that Influence Decision Aid Interventions Any result that is reached by a DA intervention is a consequence of actions in relation to a client’s vision of a decision problem and some declared and often evolving intervention aims. A description of some factors that can influence an analyst’s behaviour, activities, and results is here proposed in relation to a few cases of DA intervention that are summarised in terms of the nature of the problem and the intervention aims, and of how the factors influenced the results.

3.1

Time Factor

Time is always an important factor in any MCDA intervention, and above all for technical decision problems that require an urgent solution. A great deal of time is required when operational complexity is associated with the need for relationships between the analyst and the sources of the knowledge/information of the involved organisation(s). This activity implies legitimation of an interaction pattern, which is not naturally associated with the operational aim of a DA intervention, and detailed

8

M. F. Norese

validation activities, although a decision is in general urgent. One DA process that can be considered as an example of such operational complexity was described in (Norese and Novello 2014 and in Norese et al. 2015) and is here synthesised in Sect. 3.1.1. Time is an ambiguous factor when a DA intervention is conducted in relation to a new and unstructured problem situation. A request for a DA action as quickly as possible is critical at least in terms of time because decisions are impossible without clarification, understanding, and uncertainty control. On the other hand, time becomes a precious and available resource when decisions are not considered to be so urgent, because an analyst is involved in the Intelligence phase of a decision process. Therefore, in such a situation, the time factor is not critical, and a DA intervention can be characterised by a system approach and by a careful consideration of the adopted practices, in relation to an agreement on these terms (and above all on the analyst’s timing) with the client organisation. Some interventions can take a particularly long time to be completed, and one possible consequence of this prolongment is the concomitant evolution of the decision system (which includes decision makers and a decision structure, with rules and formal relationships with other actors in the decision process) and of the DA process that requires a new problem formulation. An interesting case of this type was described in (Norese and Torta 2014) and is here synthesised in Sect. 3.1.2 in order to analyse how the time factor can be controlled in certain interventions.

3.1.1

A Collective Design of an Integrated Land Monitoring System §

On this occasion, the request was clear: a number of companies were involved in an innovation project and wanted information about the possible new uses and operational requirements of a technology, through the expectations of the potential new clients pertaining to possible improvements in their procedures as a result of the innovation. Structured information was available about the product, i.e. military Unmanned Aerial Vehicles (UAV) that the companies wished to use for civil applications (Norese and Novello 2014). The work started with the identification of a set of possible clients in order to acquire knowledge about their needs, proposals, and suggestions. In this case, the identification of possible actions (i.e. a complete list of possible civil UAV uses) was necessary to identify the specific requirements that should have been included in the company’s product improvement processes. Only 3 months were available for the delivery of the list, a period that was not sufficient to guarantee a complete and valid list. The delivery date was motivated by the companies’ need to start a compatibility analysis of their internal processes and the proposed product innovations. Negotiation with the company that coordinated the project produced two different deadlines, 3 months for the preliminary list and another 8 months for a complete, structured, and validated list. It was a race against time. The search for the possible civil UAV uses (the actions) required most of the 3 months and was made by questioning all the possible UAV clients in the different sectors of a Public Administration, that is the Province of Turin.

What Are the Results of an MCDA Intervention? Some Reflections

9

The inquiry was then extended to other territorial areas and public bodies, whilst the consistency of the identified actions with the available or possible performances of a UAV was verified and five classes of possible civil UAV uses were created using a multiple criteria clustering procedure (Norese et al. 2015). The completeness, reliability, and legibility of this set of possible civil applications were verified, by means of an integrated use of different tools (above all cognitive mapping and actor network analysis) that involved the potential users of the new technology, in order to develop a requirement model and to analyse it, in terms of feasibility, with the companies involved in the project (the client of this DA intervention).

3.1.2

How to Improve Financing Actions in the Public Sector §

The implementation of a financing law, at a regional level, is a process that is based on operational planning activities, the collection of tenders, evaluation and selection processes, admission, and granting of the financing, payment, and control. The evaluation and selection activities have the aim of choosing the projects that should be financed. These activities play an important role in the whole implementation process. In 2003, the Evaluation Division of the Piedmont Region (NUVAL) expressed the desire to have better knowledge of the competences developed in the Regional Administration offices, regarding the evaluation and selection of projects, and to understand how to intervene in order to fill possible gaps between the requirements and local expertise or between different directorates. The promoter was the president of NUVAL, a new Division that was created in Italy in the 2000s and is still present in the Ministry of Economics and Finance and in Regional Administrations (Norese and Torta 2014; Norese 2016b). In the beginning, the strategic decision problem was totally unstructured, but the final aim of the knowledge acquisition process was clear to the promoter/decision maker. He wanted to create a role for the new Division, by improving an important and widespread function in the Regional Administration: the implementation of Regional laws that allocate public resources to several sectors. The aim of the DA intervention was above all of a cognitive nature: analysing the project evaluation and selection procedures which the Regional directorates had activated and their results, and associating the need for a less diversified terminology, timing, and procedural constraints as well as the positive or negative experiences of the directorate with each procedure. An inquiry that involves several Regional departments and directorates, in order to acquire knowledge on the expertise of the Regional Administration offices, is not simple and implies long waiting times for all the interviews. However, surprisingly, the time factor was not critical in this case. Indeed, the NUVAL president’s very good knowledge of the organisation and good relationships with all the sectors allowed the analyst to easily get in contact with all the key figures in the whole Administration and was an essential element in activating an effective and quick inquiry.

10

M. F. Norese

When the study was completed, it was possible to make a distinction between the procedures of the different directorates, and the elements that indicated the best practices and local gaps became evident. However, the promoter of the study retired suddenly, and his retirement eliminated the possibility of creating an innovative decision system for the Region coordinated by NUVAL. Meanwhile, his retirement reduced the justification of the research and perhaps even the legitimacy of a cognitive intervention. Dissemination of the results was considered a risk, with the possibility of unwelcome organisational consequences, which needed to be avoided by NUVAL and the analysts. As a result, NUVAL did not use the outcomes directly, although the results were published in an internal publication, without any transparent indication of the directorates associated with the best practices or the local gaps. Moreover, a scheme that represented the different procedures that had been carried out and which indicated both the involved actors and their mutual relationships became a very useful tool. This scheme was similar to the one proposed in the Problem formulation methodology (Bowen 1983), and it was used during the enquiry to describe and validate the acquired knowledge and during workshops and training courses for Regional and Provincial Administration executives and officials in the years 2005–2009, where MC classification models and descriptions of all the procedures were anonymously proposed as positive or negative references and guidelines for project evaluation and selection procedures. In 2012, the Region reasserted its desire to strengthen the skills within the Public Administration in the complex activity of financing action planning and in the project evaluation and selection activities, in order to simplify the procedures and internalise the activities that had previously been delegated to external organisations. In this scenario, the knowledge acquired during the study and the subsequent workshops and training courses encouraged NUVAL to conduct a new analysis. The results of the earlier work of inquiry, which were structured as a rich and easily usable knowledge base, were considered essential in relation to the Region’s problem situation. In 2012, an MC model of the well-structured problem was developed and used in a system to support the design of innovative financing actions. A DA analyst who has the position of supervisor of a master’s or PhD thesis that has to be developed in relation to a real decision problem plays a particular role. An agreement with the client organisation often includes, in addition to the content of the research project and financial terms, a specific action space and time conditions that allow the participants to reflect on the interaction between theory and practice in relation to the problem. This case represented an interesting occasion to share a logic, in which a theoretical perspective and a collaborative protocol were combined, with the Action research field (see for instance Dash 1999 and Whyte 1991) and, above all, with the Intervention research in management field (see David and Hatchuel 2008, 2014). In this DA intervention, the theoretical perspective was associated with the basic issues of MCDA, and the collaborative protocol was based on a research-oriented partnership agreement between a Master Course on Public policy analysis and the Evaluation Division of the Piedmont Region.

What Are the Results of an MCDA Intervention? Some Reflections

11

The role of an analyst in a DA intervention is another important factor that influences the analyst’s behaviour and all the results.

3.2

The Possible Roles of a DA Analyst

The role of a DA analyst is a multifaceted concept that includes his/her official position, functions, expectations, responsibilities, duties, and so on. Moreover, an analyst can take on different roles in a DA intervention. An analyst can be part of or external to an organisation that is involved in a decision problem, that is functioning as a consultant in the latter case. The nature, social dynamics, language, and culture of an organisation should be known in the former case, whilst everything is new and has to be explored, each and every time, for an external analyst. The relationships with the decision maker(s) and actors in the process may be relatively easy in the former case, but an external analyst may look into the problem without any preconceptions. Two cases, both of the project selection types, have been chosen to analyse how external and internal analysts can interact. The first, which is synthesised in Sect. 3.2.1, was developed in a Public Administration and it generated very interesting results (see Norese and Viale 2002 and Norese 2020b). The second (presented in Sect. 3.2.2) was developed in an Italian company, where a remarkable integration was witnessed between internal and external analysts, and it generated effectiveness and interesting organisational results, but also certain risks (see Norese and Bono 2019). A DA analyst can also be involved in decision situations, such as in “roundtables of experts”, as an expert in methods. An example was described in Norese (2010), in which the behaviour of an analyst in a team of experts, where communication and interaction were essential to formulate the problem and then to structure and develop the model, was examined. This case is synthesised in Sect. 3.2.3.

3.2.1

A Project Selection Case Pertaining to the Public Administration §

In the public administration, evaluation and selection are activities that are aimed at choosing the projects that have to be financed, but they can take on a much broader meaning in a continuous improvement process. The outcomes of analyses carried out on any previous procedures enable the subsequent ones to be improved and are helpful in defining new calls for tenders and, in some cases, in planning the distribution of funds. In this specific case, a DA intervention (see Norese and Viale 2002) was required to improve the evaluation and selection processes that were conducted each year in the Department of Environment of the Piedmont Region, in Italy. The experience that the client had acquired over time facilitated MC modelling and the application of an MC method. Different validation activities facilitated the acquisition of legitimation and led to a stable relationship being set up between the client and the analyst.

12

M. F. Norese

An evolution of the decision aid process was consistent with the problem situation. The analyst was involved in new problem situations, for example when some projects resulted to be particularly interesting but so expensive that the entire funding was not sufficient to cover them, and obviously in other projects. The model and procedure were collectively analysed and partially changed in relation to the new scenario. An MC procedure was used for years and marginal evolutions, in relation to new scenarios, showed the procedure’s flexibility. A good relationship with the client also produced interesting results at the DA team level: learning the model evolution logics in relation to decisions repeated over time, involvement of the analysts in the definition of new calls for tenders, elaboration of sensitivity analysis procedures, useful to aid decisions in evolving scenarios of funds distribution, and involvement of the client in the development and testing of a software package and then of a Decision Support System that included different variants of the original MC method. In this case, the model generated in the MCDA intervention evolved consistency with the policy evolution, because a good relationship between analyst and client facilitated it, but also because the outcomes of previous procedures were used to improve the policy implementation process. An incremental process of this kind is not always possible, since it is difficult to acquire and transfer knowledge and competences, since the figures who cover decisional or operational positions are changed frequently and the acquired competences are thus dispersed; or, because reduced times and budgets do not allow the outcomes to be analysed and, therefore, the models and procedures to be improved. A lack of funds, time, and availability of skilled human resources often result in the externalisation of the activities or in a prompt implementation of the procedures that thus end up being unreliable and not very effective.

3.2.2

Project Selection in a Process of IT Innovation §

In 2012, an Italian company, with a work force of 3200 employees, activated an Information Technology (IT) innovation project. The company created an IT Demand Management office (which is hereafter referred to as DEMAND) with the task of soliciting projects from the different sectors of the company, in order to map the IT innovation needs of the whole company and to plan the implementation of the most “interesting” and/or urgent IT innovation requests. Over the years, the DEMAND office had created a set of indicators to allow all the possible benefits, which could be associated with each new IT project or improvement of previous IT applications, to be recognised, measured, and documented. As a result of these activities, the DEMAND office gathered approximately 1000 requests each year from 15 to 20 sectors of the company, and around ten million euros were allocated each year from 2012 to 2019 for IT interventions. After some years, in which the number of requests had grown but the budget had been reduced, some doubts arose about the reliability of the acquired data concerning the costs and benefits of each possible IT intervention, and therefore about the project selection

What Are the Results of an MCDA Intervention? Some Reflections

13

procedure. The need for more transparency and effectiveness in the IT implementation process stimulated the development of a new evaluation procedure, which DEMAND conducted in 2017. An in-depth ex post study of the new procedure and its results, in relation to the previous and the new data, was developed, with the support of an external team of MCDA analysts (see Norese and Bono 2019). The external team was integrated into the DEMAND office and participated in meetings with the other actors of the IT innovation management process. In this phase, the DEMAND office was open to criticism. Some weak points of both the indicators and the new procedure (introduced in 2017) were identified, from a methodological point of view. New data treatment procedures were proposed by the MCDA team, and some variants of the procedures were identified and tested. The DEMAND office considered the new proposals useful. Some improvements to the procedure, which had been introduced in autumn of 2018, produced interesting results and the activation of an organisational change process. Meanwhile, the DEMAND office reorganised its work in order to create an institutional and operational exchange with the different sectors of the company, during the whole IT innovation management process, and to increase the reliability and effectiveness of the new data acquisition procedure. The innovations were communicated during some meetings that involved the proponents of the improvements and of the IT projects, in the spring and summer of 2018, to explain the motivations behind and nature of each innovation, to create new relationships between the different sectors and the DEMAND office, and to facilitate the implementation of procedural and organisational changes that were to be made in the following autumn. The data acquisition procedure was changed, for the first time in over 7 years, and other offices that had competences in relation to data analysis and use procedures were involved in the process of IT innovation projects management. Cooperation between these offices and DEMAND improved the quality of the evaluation procedure and of the model. The involvement of all these offices in the meetings facilitated the communication of each change. Furthermore, the specific knowledge of each office was made available to the different sectors of the organisation to facilitate them in valorising the benefits of their requests. This activation and sharing of knowledge resources were oriented towards discussing, planning, and implementing a new IT demand management process, which was implemented in 2019. Some interesting results were produced, in relation to the IT demand: a 55% reduction in requests, an increase in the high-value requests, that is, from 36 to 43% and a reduction in the low-value requests from 64 to 27%. The new evaluation procedure reduced the workload of the IT Area, which continued to be involved in the cost estimation, albeit with less urgency and for a reduced number of proposals. The improved relationship between the internal offices, as well as the new valorisation of their knowledge, in a process in which they had not been involved in the past, were aspects that were greatly appreciated by the participants. The quality and the results of the work of the DEMAND office were improved, but the internal dynamics of the company also improved. This new use of organisational

14

M. F. Norese

knowledge, together with new interactions and the co-creation of value, may have activated an organisational change process, from a centralised activity to a shared communication space. However, such a new knowledge-sharing vision can also generate risks. In April 2019, the risks that had been observed were described in this way: the new transparency may also reduce the “bargaining” space and can generate criticism of any aspect of a report or an automatisation proposal of a process that now, for the first time, involves several sectors of the company. Some new model and procedure improvements will be tested this year, but the possible consequences of any new change need to be studied, not only in relation to the quality of the evaluation and selection procedure, but also considering any possible organisational effects (Norese and Bono 2019). In the summer of 2019, an “organisational storm” led to the DEMAND office being dismantled, the motivation being that “the process and its procedures are now so clear and well organised that the work of this office is no longer needed”. This result could be considered as the perfect success of the DEMAND efforts and marginally of the MCDA intervention. But it could also be considered a not totally positive consequence of the proposed transparency, which could reduce the informal bargaining space, and of an erroneous perception that a well-organised procedure does not necessitate any change.

3.2.3

A DA Analyst Involved in a Roundtable of Experts as an Expert in Methods §

At the end of an innovative monitoring of funded projects, in the Department of Tourism of the Piedmont Region, the decision maker (DM) activated a “roundtable of experts” to synthesise the monitoring results (Norese 2010). The roundtable activities included field experts, but also the author, as an expert in methodologies, some analysts who had monitored the development of the funded projects and a few coordinators of the funded projects. The DM’s request was considered to be too generic, so the roundtable of experts tried to obtain indications about the purposes of the monitoring process and the possible uses of the monitoring results, but the attempt did not produce any results. As a consequence of this DM’s unstructured request, the experts then conducted an analysis of the funding law, its implementation, and of the data from the monitoring process, which led to a debate between the experts in relation to the concepts they used to define criticisms and weaknesses. The generated conflictual ambit of work led to each problem formulation attempt being blocked, because each expert’s proposal was vetoed by another expert. The situation only changed when an MC structuring approach (see Norese 2020a) was proposed and accepted. The points of view, which the experts expressed in relation to their knowledge of the field, were organised in a classic MC model, with actions and criteria, which was then used as a draft model to reduce conflict between the experts, synthesise visions and uncertainties, in relation to the monitoring process and some possible future decisions, and to arrive at the final model and conclusions.

What Are the Results of an MCDA Intervention? Some Reflections

15

Only a few of the 333 funded projects were analysed in the first draft model, to stimulate and acquire judgements of the roundtable experts on the project development and evolution (consistent with the prevision or including interruptions, delays, renunciations, revocations, absent documentation, and other anomalies), which were considered useful to create a “picture” of the situation and to explain how different data could be dealt with in an MC model and how the acquired information could be synthesised and transferred to the policy making process. Some tentative criteria were modelled to synthesise the experts’ points of view and their use of the data acquired during the monitoring process. An MC method was used to evaluate and assign the few analysed projects to such categories as: ideal project implementation, acceptable implementation, presence of marginal criticalities, and presence of widespread criticalities. Different draft models were created, as in a learning lab, and were used to elaborate a shared and formal language that involved the experts and the other participants (the actors in the roundtable). Each different interpretation was discussed, and only shared elements were included in any new model. Some elements, which had been acquired during the monitoring process (data, interviews, critical events and their motivations, evolutions of the funded project or the funding process, and so on), were analysed and used to express hypotheses, i.e. a set of formulations of how the monitoring results could be used to make a decision. A collective model development action was facilitated by the shared language and collective testing of each hypothesis. In the end, the final MC model and the results of an MC method application were used not only to produce a clear distinction between good and bad practices, but also to formulate proposals to improve the funding law, to validate the monitoring database, and to use a reference procedure for future monitoring processes and decision activities.

3.3

Decision System and Its Dynamics

The outcomes of any DA intervention are influenced by the nature and dynamics of the decision system. Such a system includes decision makers and the actors involved in a decision process, as well as a decision structure that is formalised to a certain extent, with the actors’ roles being recognised by the involved organisations and specific rules being implemented in the process. The formalisation level of the decision structure may be very limited, and thus have a negative impact on the DA process (above all concerning uncertainty about the tasks and the action space). The identification of the decision makers, at different levels of the involved organisations, and their objectives and responsibilities can sometimes be difficult. A deep analysis of this specific topic and the relationships between DA analysts and decision makers and actors is proposed in the chapter (Spatial Decision Support to Reorganise the Swiss Postal Network) by F. Joerin.

16

M. F. Norese

A decision system can change over time. The retirement of a problem owner who had stimulated the DA intervention (see A. Fiordaliso, O. Pilate, and M. Pirlot in the fourth chapter of this book or the case synthesised in Sect. 3.1.2) obviously has a negative impact on the DA intervention and its outcomes. DA processes and their evolution are frequently impacted by such dynamics, and the consequences may be an unsatisfactory intervention, in spite of the acquired expertise and its positive impact on the analyst’s learning process. Instead, this expertise can generate a more careful approach to a new intervention or new results based on the previous intervention outcomes. Some DA interventions develop without having any direct relationship with a decision system or in relation to a decision system that has not yet been activated. The client of a DA intervention may know a decision system well, as he/she habitually interacts with this system to sell services or products, and aims at effectively proposing a completely new product or service The aim of such a DA intervention is to model innovative scenarios for the decision system, without having any relationship with the final decision makers. A problem situation that needed robust elements to generate a new decision structure was described in (Norese and Carbone 2014) and is here synthesised in Sect. 3.3.1. A decision system can be latent, because the decision process is in an initial pre-decisional state, in which there is a severe lack of knowledge, or there is a difficulty that needs to be overcome in order to pass on to a decisional state of the process. In these situations, the client may be an expert in a specific domain who is in charge of a research project that could activate a decision system, and only becomes a decision maker as far as the intervention and results are concerned, or someone who perceives the possibility of activating a new decision process. In these contexts, MC modelling, models and method applications can be used to give structure to the available data (see Balestra et al. 2001; Cavallo and Norese 2001; Norese et al. 2016, 2022; Norese 2020a). An MCDA approach to these problem situations is described in Sect. 3.3.2.

3.3.1

A New Service to Improve Airport Revenues from off-Flight Services §

A consultancy company, which supplies different services to the Italian Aviation Authority and to several Italian airport concessionaire companies, was the client of an MCDA intervention (Norese and Carbone 2014). The consultancy company wanted to analyse Italian airport revenues from off-flight services, in order to help its clients to improve their product offerings and identify new product value drivers. An incremental model structuring and result analysis procedure were developed, involving both the analysts and the client, and it was oriented towards the generation of “robust” models that were clear enough to facilitate the consultancy company’s interaction with the clients, that is the final users of the procedure. The procedure allowed a specific preference elicitation approach to be adopted, without the actual

What Are the Results of an MCDA Intervention? Some Reflections

17

Table 2 Performance results in relation with the contexts Performance capability High Medium High Medium Medium Low Low

Good

Naples

Good intermediate

Intermediate

Weak intermediate

Weak

Cagliari

decision makers, that is the clients of the consultancy company, i.e. the clients of the DA intervention client. A comparison of different and almost incomparable situations (some large airports and other very small ones) was difficult and not so convincing, above all in relation to revenues from off-flight services, a completely new topic in Italy at that time. Therefore, two different models were developed, the first in relation to a wellknown context for the consultancy company, as a capability model of the airport service structure, whilst the second was developed in relation to a new problem situation (in this case, about the lack of off-flight service performances) that required knowledge that was not easily available, above all in a structured form, without the involvement of the actual decision makers in the modelling process. The capability modelling, which was relatively easy, simplified the acquisition of a common language and made the approach to a more difficult second modelling activity easier. The ELECTRE Tri method, an outranking method that can handle different and almost incomparable situations (Roy and Bouyssou 1993; Yu 1992) was used to assign the Italian airports to specific categories of capability, in the first model, and of off-flight service performance in the other. At the end of the process, the integration of the results of the two models (see Table 2) facilitated a visualisation and understanding of each concessionaire’s performance, in relation to the related airport service structure (because a small airport cannot propose several off-flight services if there is a limited demand, but a lack of these services is critical in a medium-sized airport), and of the best practices (e.g. Naples) and critical conditions (e.g. Cagliari) of the Italian airports. The models and the results of the ELECTRE Tri application, i.e. the hard outcomes of the DA process, were generated to propose an understandable vision of the marginal and overall activations necessary for each Italian airport to make a profit through the organisation of public facilities and services that were different from the classic air navigation services on the ground. A soft outcome was a logical reasoning, with clear arguments that the client acquired and used to facilitate the definition and communication with its clients, of some improvement actions, in terms of new or better off-flight services.

18

3.3.2

M. F. Norese

An MCDA Process that Transforms Data into Information §

A client is sometimes an expert in a specific domain who has acquired a great deal of data in relation to a research project, but there is not enough knowledge to transform such data into information because the situation is new and complicated. This kind of client knows that a decision system could be activated in relation to the research project and its results. But this decision system has not yet been activated or its rules, actors, and relationships are not yet totally defined and depend on the research project results. Sometimes a decision system exists, but the intervention client knows that any relationship between the research project and the decision system has not yet been activated and the research results could generate this relationship. The operational aim of a DA analyst in these situations is to give structure to data by means of an MC model, where the few available knowledge elements, or the methodological proposals from experts or from the involved disciplines, are expressed by means of alternative actions and criteria. MC modelling, models, and method applications are used to first introduce a shared language and reduce uncertainty, ambiguities and misunderstandings, and then to analyse, validate or improve the structured syntheses of the data that should produce information. A draft of an MC model is structured, and some of the available data are used to develop a first application of an MC method to this conceptual model. The first application of an MC method produces temporary results that are critically analysed, in order to find some weaknesses in the results and verify their possible relations with some weaknesses in the model parameters or structure and therefore to introduce possible changes. The improvement proposals of some model parameters are then used in a cycle of method application and testing of the results. If these improvements are not sufficient, new cycles of model restructuring, MC method application, and result analysis are activated (a procedure of internal validation of the model). Model and results, after the improvements that the internal validation has proposed and activated, are then tested with other data, in relation to the same phenomenon, and/or presented to experts who are not directly involved in the research project, in order to activate an external validation procedure. Such as structured sequence of activities was described in (Norese 2016b) with reference to a specific MCDA process, which was called a Model-Based Process, and which was based on a new and conscious use of multicriteria models as communication spaces, to facilitate the identification of the key components of the studied phenomenon (associated with different contexts, such as landslides as a consequence of a flood, a pharmacological trial in relation to muscular dystrophy, modelling of concepts such as environmental value or disaster resilience) and the quantification of the importance of such components. Model development and validation, and therefore knowledge and information acquisition, are made possible thanks to these unconventional uses of MC models and methods, which not only facilitate communication between the client and the DA analyst but also between the analysts of the client’s team, who may have different points of view, backgrounds, and experience.

What Are the Results of an MCDA Intervention? Some Reflections

19

Such uses of MC models and the application of MC methods to these models, which are temporary expressions of knowledge, may be considered methodological tools that facilitate mutual understanding and knowledge development, i.e. soft outcomes adopted to achieve the operational aim of giving structure to data, or to facilitate the activation of a decision system and process, which is the main aim of a client (see, for instance, Norese et al. 2022).

4 MCDA Interventions: Some of the Main Difficulties and Results The outcomes of an MCDA process are often described in terms of the modelling phase and application of an MC method, together with an analysis of the results and validation procedures. Other important results concern the identification and control of uncertainties and difficulties, which are achieved by means of logical tools and methods oriented towards problem formulation and structuring, together with some specific uses of MC models and methods. The MCDA constructivist approach creates a common language that facilitates communication about a problem and its difficulties, and the activation of mutual learning processes, and it also contributes to the legitimation of the analyst. The framework proposed in the first section of this chapter is here applied to analyse the hard and soft outcomes of some MCDA interventions, as proposed in this and other chapters of the book. These results are a consequence of circumstances, premises, and assumed ideas, in relation to a decision problem, declared request, and difficulties or factors that interfere with a DA process.

4.1

Technical Decision Problems

MCDA interventions are often requested in relation to technical decision problems. MC modelling of a problem and the application of an MC method are the main activities, and they may be repeated until the requested result is achieved. The main intervention aim is in general of an operational nature and is often a procedure that has to be improved, or new components have to be included in a well know procedure. Knowledge about the existing procedures is present and often becomes the essential resource for an MCDA analyst. Previous experiences and the analysis of the obtained results, and implicit categories of priority, acceptability, risk, urgency, and so on, become a reference system that facilitates and orients an MCDA intervention. Information about the organisation and the specific problem situation can be completely or partially lacking and may require specific inquiry activities. Constraints in the use of previously acquired information often generate difficulties and a waste of time. It is normal to be in touch with imperfect

20

M. F. Norese

information, whereby elements of uncertainty are present and they have to be identified, analysed, and controlled in all the phases of the decision process and in the decision aiding processes. The expected results are associated with the operational aim, although some other outcomes may also be present, above all soft outcomes resulting from having dealt with specific complexities that characterise the problem situation, such as a change in the decision system or a difficult relationship with the actors of the decision process or with some information sources. A secondary cognitive aim sometimes emerges, for example because a decision problem that had been described as wellstructured presents unstructured components, or knowledge about the existing procedures is fragmented and occasionally contradictory. In these cases, an application of an MC method is mainly oriented towards validating the acquired and modelled knowledge. Only at the end of this kind of MCDA intervention, do the final model and results of an MC method application generate the expected outcomes. The time factor is almost always critical in interventions associated with technical decision problems. A form of control of this factor should be generated at the beginning of an intervention, during the interaction process between the analyst and client. The use of simple tools that facilitate collaborative model structuring and result analysis procedures can reduce the time and efforts (see some of the tools proposed in the seventh chapter (Contrasting applications in environmental planning and public procurement) by J. Pictet and D. Bollinger, and an example of how these tools can be used, as presented here in Sect. 4.1.1. A result that often needs to be acquired, in relation to a technical problem, is the involvement of the problem owner in the process that produces a procedure improvement or modification, to enable the client to manage the post-implementation and updating phases. The activation of this kind of organisational learning process can be considered as something between a hard and soft outcome. Such a process requires time, but it can produce interesting results, not only for the client, but also for the MCDA analyst. Some of these results are described in relation to a case in Sect. 4.1.2. A cognitive aim, in relation to a technical decision problem, is rarely associated as the main aim with a DA request. On this topic, I can only mention my lack of experience when one of the supervisors of my master’s thesis proposed an interesting theme. The supervisor was also the advisor of the Regional Planning Ministry, and the proposed theme was a comparison of some possible locations of buildings belonging to the “superior tertiary sector” in Turin, i.e. the central offices of FIAT, banks and financial groups, but also the regional departments of the Turin Courthouse and other public services. I created an MC model and used an MC method in my master thesis. The aim of my work was only cognitive, not operational, and the expected outcomes were the MC model and the potential application of an MC method, not the result of the application of the method. I started this activity by studying a great deal of documentation housed in the regional offices. When I discovered that there had been an official proposal from FIAT to locate its central offices in a different place from those listed as possible

What Are the Results of an MCDA Intervention? Some Reflections

21

sites, I was excited and immediately described this possible enlargement of the location sites to my supervisor. But he was not excited! In fact, he lost all interest in my work. I completed my thesis and included this site, but he refused to sign my thesis. Forty-five years later, these buildings are now almost completed, and I have realised that the FIAT proposal (two cubic metres of documentation) was only a real estate investment, as the central offices were eventually located in a totally different site, which was neither in the proposal nor included in the advisor’s list of possible sites. In this case, the reaction of my supervisor, the client of my work, was an important lesson for me. He had only wanted a comparison of the sites he had suggested. And I learned that any change in the work and behaviour of an analyst, after the problem formulation phase, should be proposed with a great deal of attention.

4.1.1

A Simple Tool to Reflect on and Structure an MC Model §

A simple way of facilitating a collaborative model structuring procedure and a timeconsuming activity is to use a logical tree in which all the elements that are proposed to generate criteria are included, but visually distinguished into conceptual elements, data, and criteria. Figure 2 proposes the second and the final structure of a model that was elaborated as a sequence of five different trees, that is different in terms of levels of conceptual model disaggregation (from 2 to 5), number of conceptual elements (only one at the beginning, the aim of the model, which then became four in the second structure, with an additional three dimensions, and nine in the last structure, with some sequential disjunctions and better definitions of concepts and, albeit only at the end, of the criteria) and of the use of the data. In the end, only one criterion required the elaboration of a great deal of data, whilst some evaluations were associated with only one indicator/judgement and others synthesised a few different data, mostly as a result of additions or combinations.

Structural concepts

Criteria

Fig. 2 Logical structures of an MC model

Data

22

4.1.2

M. F. Norese

Expected and Unexpected Results in a Process of Project Selection §

Project evaluation and selection procedures are very frequent in public administrations. These decision situations imply a clear definition of the projects that can be admitted to the selection process, and a transparent evaluation, selection, and communication of the results. The expected result in a particular MCDA case (see Norese 2016a) was the introduction of formal MC models and procedures to improve repeated project selection processes. The DA intervention included the classic activities that should be associated with a sufficiently clear decision problem, when information is available about the procedures, results, and about the experience acquired in the past: defining the intervention mode and the time limits, identifying the nature and weaknesses of the past procedures, formally developing a model, and the application of a method in relation to some new projects, which are, at the same time, evaluated and selected through the usual procedure. The results of the MC method application were compared with the results that the client had produced by adopting the usual procedure, and the model and results were analysed and validated by the analysts’ team and then with the client. The conclusions of the MCDA analyst were considered satisfactory, above all in relation to the projects associated with the main uncertainties. A multicriteria procedure was adopted and improvements of the model, in relation to evolutions of the selection problem and suggestions made by the client, were introduced for some project selection processes. The client’s understanding of the DA path and involvement in the updating of the model were directed towards generating a certain level of autonomy in using the MCDA procedure in the repeated decisions. However, an unexpected and sudden change in the personnel put the client in the difficult position of having to entrust a new employee with the task of evaluating the projects. This person was not trained in this field and, as a first duty, was asked to analyse and evaluate the projects. The new employee encountered some problems in this task, but she was able to correctly evaluate each project. At the end of this task, her difficulties in the evaluation process and the results were analysed, with the help of the previous employee, and no misinterpretation of the process or results was observed. In fact, some of her suggestions, about a clearer definition of evaluation states, were considered valuable and were used to improve the legibility and operationality of the MC model for future applications. Moreover, an unexpected result emerged after some years. At the start of the intervention, the client’s declared aim had been to facilitate a procedure that was becoming complicated, and which would be eliminated in the near future. However, after several years, the procedure was still being adopted, without any further involvement of the MCDA team. A graphical schema, which included the criteria, evaluation states, and reference profiles that were necessary to distinguish acceptable projects from others that had to be rejected, was created and used each time to evaluate the projects and to distinguish acceptable projects from those that needed to

What Are the Results of an MCDA Intervention? Some Reflections

23

be rejected, whilst the reasons for rejecting them clearly emerged in the schemata and were documented. Only at that moment was it evident that the logic of the method was clear and that self-sufficiency in the repeated decisions had been achieved. Moreover, the graphical and logical tool created by the employee resulted to be an interesting and unexpected positive outcome.

4.2

Strategic Decision Problems

A DA intervention is sometimes required in relation to strategic decision problems, motivated by the cognitive aim of clarifying and formulating a problem. The innovative nature of the problem generates complexity, above all when information, knowledge, and previous procedures do not exist. These sources of complexity may be associated with an insufficiently clear or structured request. The use of Problem Structuring Methods (PSMs) is always more frequent in these situations (see, for instance, Rosenhead 1989; Mingers and Rosenhead 2004; White 2009). MC methods are often combined with PSMs (see Belton and Stewart 2010 and the review by Marttunen et al. 2017). PSMs allow the problem to be seen and analysed in its total context, clarified and structured, and the main components of an MC model (alternative actions and scenarios, weights, and need of other parameters) to be identified. Concurrently, the proposed use of MC methods that accept any kind of data (not only quantitative and qualitative, but also textual and visual) reduces the concerns and issues of decision makers who are well aware of the limited availability of quantitative and structured data. The sequence of an MCDA application, after the structuring contribution of a PSM, produces interesting results and underlines the complementarity of these approaches (see, for instance, Belton et al. 1997; Bana e Costa et al. 1999; Montibeller et al. 2008; Norese et al. 2004, 2008; Stewart et al. 2010; Ferreira et al. 2011). Other Operation Research methods can be used with the same aims: understanding the situation, reducing uncertainty, and structuring the decision problem. The PERT method, which was used in (Norese 2011) as a communication space to elicit the actual attitudes of the teams involved in a strategic project, is an example. MC models and methods are sometimes used at the beginning of an MCDA intervention as tools that facilitate communication and structuring, when the problem is so new that the client’s request needs such activities as problem definition and formulation, knowledge acquisition and structuring, and organisational learning (see, for instance, Norese 2009). Some examples of MCDA interventions, in relation to strategic decision problems, are described in this book, and some of their difficulties and results are analysed in this section.

24

4.2.1

M. F. Norese

The Urban Postal Network Reorganisation in Switzerland §

A strategic decision process was activated by the Swiss Post, which involved a consultancy firm between 1999 and 2003. A reflexive analysis about an MCDA intervention, in relation to this decision process, is proposed in the second chapter (Spatial Decision Support to Reorganise the Swiss Postal Network) by F. Joerin, to underline how contextual elements make a public policy process complicated and how they condition any relationship between the processes of decision aiding and decision-making. The passage from a logistical strategy of cost reduction to the reorganisation of the post office network and the inclusion of a new dimension, that is of urban planning and territorial organisation, were the first outcomes of the study. The main difficulties were the identification of the different decision levels involved in the project and the involvement of the decision makers, at the different levels. An integrated use of statistical analysis, MC analysis, and spatial proximity analysis generated information and detailed arguments (rather than solutions), which significantly facilitated the negotiations of the project team with the local authorities of each city, as well as the implementation of the restructuring of the urban postal networks.

4.2.2

MCDA Analysts and the City of Quebec §

Two MCDA interventions, pertaining to strategic decisions contexts, which were conducted together with the City of Quebec, are described in the sixth chapter (Multicriteria Decision Aiding—An Overview of Some Challenges of Interventions in Real-Life Applications) by I. Abi-Zeid, F. Marleau Donais, and J. Cerutti, and a comparison of the situations is particularly interesting. The interventions were conducted in different fields, that is, transportation and urban planning, and the used methods, which were different from each other, represent two classes of MCDA methods. However, other important differences between the two interventions have been pointed out in the chapter. As far as the first intervention is concerned, the City of Quebec was ready for a socio-technical MCDA process, and the problem situation, which the City’s professional figures had changed before the intervention, was clear. The operational aim of the MCDA project was well defined, did not change during the intervention and produced successful results. The City administration went public with the developed decision support tool, and started using it and the maps that had been produced during public consultation meetings. The project, which became a flagship of an innovative urban laboratory, went on to receive many awards. However, another kind of result of this intervention should also be underlined. The transdisciplinary nature of the decision process, which was conducted in collaboration with stakeholders from different departments, was a new “approach” organised according to the City’s wishes. Several participants expressed doubts

What Are the Results of an MCDA Intervention? Some Reflections

25

about the project and, in general, had different views on the validity of the method (considering it too “rational” or too “subjective”). The different departmental cultures and knowledge backgrounds of the participants created a state of confusion, and a great deal of time was required to reach a consensus and formalise their experience into useful information for the model. The generation of a common vocabulary and consensus in this complicated organisation context were important results. A facilitator approach to the problem was able to ensure that equilibrium was reached during the workshops, and it also allowed the professionals to be more creative in their thinking, to gain confidence in themselves, and increase their trust in the process. The model was accepted during the intervention and its value also became evident thanks to the use of validation tests, which convinced the professionals and confirmed the use of the model in the decision support tool. The second case was also considered a success and was appreciated by the City’s administrators, but, in this case, the role of the analysts was different. The team was initially involved as assistants to an advisory committee in its task and as observers during the first meeting. Initially, no operational aim was defined, whilst the team was allowed to observe the issues and within power dynamics between members of the committee, and they were thus able to perceive the vague nature of the question that had to be resolved. An operational aim (the development of an MCDA model and methodological sorting approach) only became clear as a result of the analysts’ aid in problem formulation (first important result), and through the use of conceptual maps, as a basis for discussion, the visual organisation of some results and the orientation of the committee’s work (other important results). The success of this MCDA intervention is discussed in the chapter with an explicit reference to another possible role of the analysts, which was perceived by the involved actors, who also expected a policy analysis. This analysis, which is a qualitative opinion regarding the “why” of the results, was not an explicit (or implicit or secondary) aim nor a request during the intervention. It could instead be perceived as the almost natural evolution of a role of observer, facilitator, and then generator of results that aided decisions and “will be an excellent basis for further reflection”. Another new role, of policy analyst, was therefore considered “obvious” by the participants, although it was not explicitly expressed. This was an issue of expectations management.

4.2.3

Methodological Approaches to Public Contexts of Decision-Making §

In the seventh chapter (Contrasting Applications in Environmental Planning and Public Procurement), J. Pictet and D. Bollinger underline that MCDA cases are always different, but there are certain classes of cases that can be dealt with through the use of specific approaches and combinations of technical activities. The analysis of the outcomes of different methodological approaches to two public contexts of decision-making (strategic decisions in environmental planning and public procurement) is particularly interesting. A distinction is proposed in the chapter between the

26

M. F. Norese

acceptability of the adopted methodology and the practical usefulness of its application. The acceptability of the methodology is associated with a choice of MC methods that is consistent with the problem situation and, mainly, with the use of these methods and their integration with particular tools. These tools were created specifically and used to facilitate the interaction of the involved people that the methods required for their applications and the visualisation and understandability of the model component and method results. The practical usefulness of an MCDA application is not only associated with the implementation of the proposed recommendation, but also with some process results that can be used to generate new solutions or to modify policies, procedures or projects, and to activate learning processes. Some reasons for the non-implementation of certain proposed recommendations are described hereafter to clarify that the practical usefulness of an MCDA application is not reduced by a not accepted recommendation. I would also add that the proposed methodology and tools can be easily used in different countries and for different cultures, as I realised when I explicitly referred to the “Swiss methodological approach” in a case of environmental planning in Italy, and the more than 50 members of the committee agreed with the implementation of the methodology and the use of the different tools (see Norese 2006). The authors underline that a great deal of effort to promote MCDA in the new and apparently favourable context of public procurement has produced some good results, but not a sufficient diffusion and acceptability. In Italy, where a positive factor has been the explicit reference to MC methods, such as ELECTRE and AHP, in a National law on Public procurement (No. 415, 18/11/1998, known as Merloni ter), the situation has been the same, i.e. a missed opportunity, above all because the methods were perceived in a juridical context as an automatic generation of solutions and not as powerful tools in decision aiding contexts.

4.2.4

Decisions for Complex Policy Problems

Complex policy problems include, for example political constraints, interest groups, complicated responsibilities, collusion effects and sustainability issues (Daniell et al. 2016; De Marchi et al. 2016). Transparency should be an essential feature of public policy processes, and the tools that facilitate collective analysis, argumentation, debate, and communication should generate transparency. A public policy process should consider civil society and future generations along with policy objectives and the market conditions. An answer to the need of considering “the ethical obligation of taking a plurality of social values, perspectives and interests into account” is the methodological framework of public policy evaluation and conflict management that is described in the eighth chapter (Social MultiCriteria Evaluation of Policy Options) by G. Munda. The author indicates that not only the nature of a specific policy, but also the geographical and cultural contexts, generate the complexity of real-world problems, and only an integration of different scientific languages can deal with these essential components of complexity and the

What Are the Results of an MCDA Intervention? Some Reflections

27

closely related difficulties. Some lessons that emerged from interventions in different geographical contexts are proposed to underline that results are only possible if the conditions listed in the Munda’s chapter are verified. J.F. Guy and J.Ph. Waaub, the authors of the ninth chapter (GIS-Based/MCDA Platform for Strategic Planning and Complexity Management in Conflictual SocioEcosystems), had acquired a very good knowledge of regional planning processes and environmental assessment, as a result of their past experience in committees and/or when working for the Quebec government. Their methodological approach was developed to generate a precious opportunity for municipalities and regional county municipalities, as they often have to face constraints concerning their budgets and the resources available to implement the participated processes that are requested in many countries, even though they are not yet mandatory in Quebec. Any reader who would like to implement such an approach could proceed by referring to the proposed pilot case, where recommendations are given for each stage of the process, according to what the authors have learned from their activities in actual cases.

4.2.5

Research Projects and Strategic Decision Problems

Research projects are characterised by innovativeness, and therefore by complexity, and are often associated with a strategic decision problem. Some MCDA interventions are possible and useful in such problem situations and are motivated by the operational aim of giving a structure to data by means of an MC model, when some dispersed knowledge elements, or the methodological proposals from experts or from the involved disciplines, are expressed by means of alternative actions and criteria. The principles and techniques that have been proposed in the literature (see Roy 1996) to verify the coherence of a criterion family, and the internal conditions of a set of potential actions, can be used as rational guidelines for a collaborative modelling process. Such guidelines are above all appreciated when the involved actors have technical competences or have acquired structural modelling competences during the DA intervention. Problems of disagreement or micro conflict can be dissolved by means of this logical approach, in which the modelling/validation process (see Déri et al. 1993; Landry et al. 1983) becomes a rational reference frame that may be used to reduce misunderstandings and make points of view explicit. The strategic role of a research project sometimes motivates a long and complicated MCDA intervention. A Project that Had Been Explicitly Requested by Some Wine Industry Stakeholders § An interesting example is described in the fifth chapter (Multicriteria Decision Aiding for a Shift Towards Best Environmental Practices in Agriculture, with a Focus on Viticulture) by F. Macary, in relation to a project conducted to study the use of pesticides and their impacts on ecosystems in the Bordeaux wine-growing region.

28

M. F. Norese

The aims of the project were both cognitive (the acquisition and use of significant data was a fundamental component of the project) and operational (the definition, evaluation, and selection of new practices with the involvement of the stakeholders in the working group). The main difficulty was evident before the project even started, and it was associated with the subject of the project, i.e. the use of pesticides. This topic is always very delicate and requires great care and caution, with a negative impact on the time factor and relationships inside any project. A close collaboration with the stakeholders, and above all with the professional winegrowers, over the 5 years of the project was the essential tool adopted to face this critical difficulty and to legitimate the project and its results. The adopted approach was made possible because of the role of the MCDA analyst, who was (and is) a researcher at INRAE, the National Research Institute for Agriculture, Food, and the Environment, which coordinated the project and the agronomy, agroecology, and ecotoxicology tasks. INRAE applies research to be useful to the community, and MC models and methods are considered suitable tools to synthesise, present and discuss each step of a research project. Relationships between a Strategic Decision Problem and a Technical Problem § A technical problem, such as a hospital’s choice of the most appropriate patient classification system, once the current system had become obsolete, can be associated with a strategic decision problem, when the hospital is an actor of organisational change processes at a national and municipal level, and the technical systems must evolve in relation to changes in diseases (the COVID-19 pandemic is an example) and in the nature and needs of the population. M. de Vicente y Oliva describes this case in the third chapter (A Decision Aiding Methodology to Compare Patient Classification Systems). When a temporary working group was created at the Fuenlabrada Hospital in Madrid, a reflection process was required for the experts who worked in the hospital, and for the MCDA analysts, so the main aim of their work was therefore cognitive. Interest in research was shared by the group and partially changed the main aim of the group activity, which became operational. The working group not only presented their reasons for or against the new analysed systems, but also elaborated a decision support system, to be used for frequent technological decisions, and adopted transparency in the definition of all the elements of the proposed MC model, and of its present and future role in the system. The ex post analysis of this intervention led to some reflections, which are synthesised in the chapter. The interaction between experts from the hospital and the MCDA analysts (from a university) was not always smooth, but, nevertheless, the result was a success: shared knowledge about a methodologic approach to a problem, which would be useful for future cases. A negative outcome came to light, that is the critical relationship that emerged between the working group and the hospital director, that is the decision maker in relation to the specific “technical decision”, who did not accept the working group’s proposal and anticipated the end of the research project and the working group activities.

What Are the Results of an MCDA Intervention? Some Reflections

29

Decision aiding does not impose a decision and is not prescriptive, therefore the rejection of a technical proposal is obviously always possible. The critical element of this intervention was that the working group had no occasion to adopt a constructivist approach together with the decision maker during the 2 years of the analysis activities. A possible reason for this could be due to the nature of the context, a research project, which involved all the participants in a medium-long-term vision of the problem, whilst the hospital director had a short-time vision, consistent with the aims of another political level. Conversion of a Technical Problem into a Strategic Decision Problem § A. Fiordaliso, O. Pilate, and M. Pirlot analyse in their chapter (A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing) an intervention that started as a master thesis and then became a research project to develop a decision aid tool for a Ministry (MET). An engineer from the Belgian Road Research Centre (BRRC) identified a topic for his master’s thesis during his degree programme, which then became an opportunity for his organisation. The main aim of the intervention was of an operational nature (a new component had to be included in a well-known procedure), but the decision problem was more strategic than technical. The intervention became an occasion for the MET general manager to reduce the decisional autonomy of some members of the organisation. However, these final users of the new tools were not involved in the working group or informed about this innovation. Although the MCDA analysts underlined that this could constitute a criticality, nothing was done about it. The nature of the technical team was an aspect of strength in the intervention. The researcher at BRRC (who was also a student and then ex student at the University of Mons) played the key role of the link who facilitated the translation between languages and competences. He was ideally positioned to explain the decision aiding concepts to the MET working group (which included regional managers and engineer directors and was headed by the MET general manager), as well as the goals, constraints, and technicalities of the road management to the University team. The other members of the MCDA team first acted remotely as the thesis advisors and then integrated the working group and took an active part in the interaction with the experts in the working group. The involvement of the general manager and some directors in the working group is another element that is not so usual. The double role of the people who were involved, that is of decision makers and experts, may be seen as a clear orientation towards an operational context, which can be considered as a learning lab that can often generate organisational learning. After the retirement of the general manager, the arrival of a new direction who had different priorities is a classic example of actor dynamics, which, in this case, made the complete validation of the results difficult and impacted the legitimation of the decision aid intervention. The main technical difficulty was that the model had to be tailored to all the possible types of road works. Another kind of difficulty was that the analysts had to confront heavy time constraints, which made discussion and examination of any

30

M. F. Norese

inconsistency or uncertainty time-consuming and therefore impossible, or generated negative implications for the MCDA analysts in the preference elicitation and validation activities. The description of the intervention and its difficulties and the ex post analysis and evaluation of the case and its results produced a lesson for the future, which could be useful for the readers of the book.

5 Conclusions Checking whether the solution selected was indeed successful, when implemented, is inadequate to appraise the quality of an intervention (due to external factors impacting such a success) (Barcus and Montibeller 2008). An evaluation of internal outcomes that focuses on the quality of decision-making and decision aiding processes is still needed, but there seems to be a lack of attempts, in the literature, of appraising these ‘soft’ outcomes. When an MC model is used as a draft model to stimulate judgements, visions and criticisms, or to reduce conflicts between the involved actors, different internal (soft and hard) outcomes are produced: a shared and formal language that facilitates communication, reduces misunderstandings and makes points of view explicit; an integration of different scientific languages that interpret the complex, interconnected and multidisciplinary nature of real-world problems; a logical connection of fragmented data; the generation and testing of hypotheses on project implementation; the use of the results, and a final and shared model to apply an MC method and produce consistent recommendations. The application of an MC method to a draft model produces other internal and temporary outcomes, i.e. results that need to be critically analysed, in order to improve or validate the model, as well as to generate new solutions or the adaptation of the originally analysed policies, procedures and projects, sometimes to acquire and use knowledge, and to restructure or reformulate the problem. Some internal outcomes, which allow the modelling process to be activated, validated, and completed, are the creation and use of simple logical tools, to facilitate understanding and communication, and collaborative procedures of MC model structuring and result analysis, validation, improvement, and communication. These tools are hard outcomes that allow an MCDA intervention to be completed and its quality improved (essential soft outcomes), because the visualisation and understandability of the model components, used data and MC method results not only generate possible solutions but also detailed arguments, justifications, maps, and logical reasoning for the decision-making process. The organisation and governance of an MCDA intervention imply activities, whose results facilitate the modelling process and the use of an MC method. Some of the important outcomes of such activities are: identification of the social, political, and economic context; analysis of the decision structure and its components at different levels; proposal of decision rules; activation of a latent decision system;

What Are the Results of an MCDA Intervention? Some Reflections

31

development and control of the relationships between the DA analysts and decision makers, actors and experts, and between the external and internal analysts; control of the time factor and identification and control of any uncertainties and difficulties. Some DA activities are mainly oriented towards the involved organisations, their active participation in the DA process and positive consequences in the decision processes. Some soft outcomes of this logical approach can be: the activation of organisational learning processes at different levels and in different contexts; detection and, sometimes, dissolution of disagreements or micro conflicts; presentation and discussion of each step of a research project, which facilitate the acquisition of legitimation for the analysts, self-sufficiency in the repeated decisions for the client and its organisation, and reflection on the DA intervention. An a posteriori analysis of a DA intervention, or post-project evaluation, as it is called in (Marleau Donais et al. 2021), can be of interest for several stakeholders and, above all, for facilitators and MCDA analysts, who “are interested in learning about how the process they facilitated was perceived and what can be improved”. A specific action space and the time conditions that allow the participants to reflect on the interaction between theory and practice, in relation to the problem and the process, could be negotiated at the beginning of a project in order to monitor theoretical and practical problems, the team’s reactions and the results, the stakeholders’ perceptions and reactions, and so on. A research-oriented partnership agreement could clarify the attitude of an analyst who aims to learn from his/her experience (and also from mistakes), but mainly attempts to reduce and control uncertainty and time wasting, and to aid decision-making and action. Sharing the “intervention research” logic (see David and Hatchuel 2008, 2014) can improve a decision aid intervention and help document its soft outcomes for future practice, for the involved stakeholders and, by means of adequate reports, for “young” researchers and practitioners. The proposed analysis framework is very simple: only two main aspects (the nature of the problem and the intervention aims) are used to classify the situations. The distinction between strategic decision problems and technical problems is minimal, but sometimes not unambiguous, because a technical problem may be included in a strategic decision problem, or an intervention undertaken to solve a technical problem can activate new decision processes, and sometimes strategic decision processes. The time factor and the decision system dynamics have important effects on these complicated situations. The role of the analyst is another factor that impacts any intervention and its results. Other factors should be analysed. A suggestion is proposed by Belton and Hodgkin (1999) through the question: who are the final users and in what context will they use the outcomes of a DA intervention? A DA intervention is not only oriented towards the client but also towards all the participants, stakeholders, and final users, even though they may not be involved. Moreover, another factor that should be analysed is linked to the unicity or multiplicity of the required language(s) and communication tools, and to the context of use of the direct and indirect outcomes of the DA intervention.

32

M. F. Norese

References Balestra G, Norese MF, Knaflitz M (2001) Model structuring to assess the progression of muscular dystrophy. In: Colorni A, Parruccini M, Roy B (eds) A-MCD-A–Aide Multicritère à la Décision (Multiple Criteria Decision Aiding). European Commission Joint Research Centre EUR Report, Luxembourg, pp 31–46 Bana e Costa CA, Ensslin L, Correa EC, Vansnick J-C (1999) Decision support systems in action: integrated application in a multicriteria decision aid process. Eur J Opl Res 113(2):315–335. https://doi.org/10.1016/S0377-2217(98)00219-7 Bana e Costa CA, Fernandes FGF, Correia PVD (2006) Prioritisation of public investments in social infrastructures using multicriteria value and analysis and decision conferencing: a casestudy. Int T Oper Res 13(4):279–297. https://doi.org/10.1111/j.1475-3995.2006.00549.x Barcus A, Montibeller G (2008) Supporting the allocation of software development work in distributed teams with multi-criteria decision analysis. Omega 36(3):464–475 Belton V, Ackermann F, Shepherd I (1997) Integrated support from problem structuring through to alternative evaluation using COPE and VISA. J Multicrit Decis Anal 6(3):115–130. https://doi. org/10.1002/(SICI)1099-1360(199705)6:3%3C115::AID-MCDA140%3E3.0.CO;2-I Belton V, Hodgkin J (1999) Facilitators, decision makers, D.I.Y. users: is intelligent multicriteria decision support for all feasible or desirable? Eur J Opl Res 113(2):247–260. https://doi.org/10. 1016/S0377-2217%2898%2990214-4 Belton V, Stewart T (2010) Problem structuring and multiple criteria decision analysis. In: Ehrgott M, Figueira JR, Greco S (eds) Trends in multiple criteria decision analysis. International series in Operations Research & Management Science, vol 142. Springer, Boston, pp 209–239 Bowen K (1983) An experiment in problem formulation. J Opl Res Soc 34(8):685–694. https://doi. org/10.1057/jors.1983.162 Cavallo A, Norese MF (2001) Slope instability and erosion: hazard evaluation through the integration of GIS and multicriteria analysis. Informatica 12(1):25–44. https://doi.org/10. 3233/INF-2001-12102 Daniell KA, Morton A, Ríos Insua D (2016) Policy analysis and policy analytics. Ann Oper Res 236(1):1–13. https://doi.org/10.1007/s10479-015-1902-9 Dash DP (1999) Current debates in action research. Syst Pract Act Res 12(5):457–492. https://doi. org/10.1023/A:1022465506555 David A, Hatchuel A (2008) Collaborating for management research: from action research to intervention research in management. In: Shani AB, Mohrman SA, Pasmore WA, Stymne B, Adler N (eds) Handbook of collaborative management research. SAGE Publications, London, pp 143–162 David A, Hatchuel A (2014) Intervention research in management. The SAGE Encyclopedia of Action Research. SAGE Publications, London, pp 461–463 De Marchi G, Lucertini G, Tsoukiàs A (2016) From evidence based policy making to policy analytics. Ann Oper Res 236(1):15–38. https://doi.org/10.1007/s10479-014-1578-6 Ferreira F, Santos S, Rodrigues P (2011) Adding value to bank branch performance evaluation using cognitive maps and MCDA: a case study. J Oper Res Soc 62(7):1320–1333. https://doi. org/10.1057/jors.2010.111 Genard J-L, Pirlot M (2002) Multi-criteria decision-aid in a philosophical perspective. In: Boyssou D, Jacquet-Lagrèze E, Perny P, Slowinski R, Vanderpooten D, Vincke P (eds) Aiding decisions with multiple criteria: essays in honour of Bernard Roy. Kluwer Academic, Dordrecht, pp 89–117 Déri R, Landry M, Banville C (1993) Revisiting the issue of model validation in OR: an epistemological view. Eur J Opl Res 66(2):168–183. https://doi.org/10.1016/0377-2217(93)90311-A Landry M, Malouin JL, Oral M (1983) Model validation in operations research. Eur J Opl Res 14(3):207–220. https://doi.org/10.1016/0377-2217(83)90257-6 Marleau Donais F, Abi-Zeid I, Waygood EOD, Lavoie R (2021) A framework for post-project evaluation of multicriteria decision aiding processes from the stakeholders’ perspective: design

What Are the Results of an MCDA Intervention? Some Reflections

33

and application. Group Decis Negot 30(5):1161–1191. https://doi.org/10.1007/s10726-02109753-y Marttunen M, Lienert J, Belton V (2017) Structuring problems for multi-criteria decision analysis in practice: a literature review of method combinations. Eur J Opl Res 263(1):1–17. https://doi.org/ 10.1016/j.ejor.2017.04.041 Meinard Y, Tsoukias A (2019) On the rationality of decision aiding processes. Eur J Opl Res 273(3):1074–1084. https://doi.org/10.1016/j.ejor.2018.09.009 Mingers J, Rosenhead J (2004) Problem structuring methods in action. Eur J Opl Res 152(3): 530–554. https://doi.org/10.1016/S0377-2217(03)00056-0 Mintzberg H, Raisinghani D, Theoret A (1976) The structure of the unstructured decision process. Admin Sci Quart 21(2):246–276. https://www.jstor.org/stable/2392045 Montibeller G, Belton V, Ackermann F, Ensslin L (2008) Reasoning maps for decision aid: an integrated approach for problem-structuring and multi-criteria evaluation. J Oper Res Soc 59(5): 575–589. https://doi.org/10.1057/palgrave.jors.2602347 Norese MF (2006) ELECTRE III as a support for participatory decision-making on the localisation of waste-treatment plants. Land Use Policy 23(1):76–85. https://doi.org/10.1016/j.landusepol. 2004.08.009 Norese MF (2009) A multi-criteria decision aiding system to support monitoring in a public administration. Int J Decis Support Syst Tech 1(4):59–71. https://doi.org/10.4018/jdsst. 2009062604 Norese MF (2010) How to support decisions and guarantee robustness in multi criteria decision aid when the preference system is not “accessible”. In: Zopounidis C, Doumpos M, Matsatsinis NF, Grigoroudis E (eds) Multiple criteria decision aiding. Nova Science Publishers, New York, pp 1–16 Norese MF (2011) An application of MACRAME to support communication and decisions in a multi-unit project. Group Decis Negot 20(1):115–131. https://doi.org/10.1007/s10726-0109206-6 Norese MF (2016a) Decision aid in public administration: from evidence-based decision making to organizational learning. In: Zopounidis C, Doumpos M (eds) Multiple criteria decision making: applications in management and engineering. Springer, Heidelberg, pp 1–29 Norese MF (2016b) A model-based process to improve robustness in multi criteria decision aiding interventions. J Multicrit Decis Anal 23(5-6):183–196. https://doi.org/10.1002/mcda.1597 Norese MF (2020a) SISTI: a multicriteria approach to structure complex decision problems. In: Ben Amor S, Teixeira de Almeida A, de Miranda JL, Aktas E (eds) Advanced studies in multicriteria decision making. Taylor & Francis, New York, pp 57–80 Norese MF (2020b) Profiling analysts and actors in interaction: how behavioural aspects can positively affect the decision aid process. EURO J Decis Process 8(3):125–150. https://doi. org/10.1007/s40070-020-00113-w Norese MF, Montagna F, Vinardi FM (2004) Multicriteria modelling and rational use of waste. In: Meredith R, Shanks G, Arnott D, Carlsson S (eds) Decision support in an uncertain and complex world (proceedings of the 2004 IFIP international conference on decision support systems. Monash University, Prato), pp 598–607 Norese MF, Montagna F, Riva S (2008) A multicriteria approach to support the design of complex systems. Found Comput Decis Sci 33(1):53–70 Norese MF, Viale S (2002) A multi-profile sorting procedure in the public administration. Eur J Opl Res 138(2):365–379 Norese MF, Carbone V (2014) An application of ELECTRE tri to support innovation. J Multicrit Decis Anal 21(1-2):77–93. https://doi.org/10.1002/mcda.1508 Norese MF, Torta V (2014) A decision support system in order to facilitate new financing actions in the public sector. Territorio Italia 2:85–100. https://doi.org/10.14609/Ti_2_14_6e Norese MF, Novello C (2014) Understanding and profiling user requirements to support the conceptual design in the aeronautical sector. Int J Multicriteria Decision making 4(2): 135–153. https://doi.org/10.1504/IJMCDM.2014.060424

34

M. F. Norese

Norese MF, Novello C, Salassa F (2015) An integrated system to acquire knowledge and support decisions in complex innovation design processes. J Org Comp Elect Com 25(2):194–212. https://doi.org/10.1080/10919392.2015.1033949 Norese MF, Mustafa A, Scarelli A (2016) New frontiers for MCDA: from several indicators to structured models and decision aid processes. Newsletter of the European Working Group “Multiple Criteria Decision Aiding” 3(34):1–8 Norese MF, Bono GF (2019) A knowledge-based organisational solution to create value and facilitate IT innovation. In: Knowledge ecosystems and growth, Proceedings IFKAD 2019, Matera 5-7 June 2019, pp. 966-972 Norese MF, Rolando D, Curto R (2022) DIKEDOC: a multicriteria methodology to organise and communicate knowledge. Ann Oper Res. https://doi.org/10.1007/s10479-022-04711-6 Ormerod RJ (2014) Critical rationalism in practice: strategies to manage subjectivity in OR investigations. Eur J Opl Res 235(3):784–797. https://doi.org/10.1016/j.ejor.2013.12.018 Rosenhead J (ed) (1989) Rational analysis for a problematic world: problem structuring methods for complexity, uncertainty and conflict. Wiley, Chichester Roy B (1987) Meaning and validity of interactive procedures as tools for decision making. Eur J Opl Res 31(3):298–303 Roy B (1993) Decision science or decision-aid science? Eur J Opl Res 66(2):184–203. https://doi. org/10.1016/0377-2217(93)90312-B Roy B (1996) Multicriteria methodology for decision aiding. Kluwer, Dordrecht Roy B, Bouyssou D (1993) Aide multicritère à la décision: mèthodes et cas. Economica, Collection Gestion, Paris Simon HA (1960) The new science of management decision. Harper and Brothers, New York Simon HA (1991) Bounded rationality and Organisational learning. Organ Sci 2:125–139. https:// doi.org/10.1287/orsc.2.1.125 Stewart TJ, Joubert A, Janssen R (2010) MCDA framework for fishing rights allocation in South Africa. Group Decis Negot 19(3):247–265. https://doi.org/10.1007/s10726-009-9159-9 Tsoukiàs A (2001) What are real-life applications of decision support? Newsletter of the European Working Group “Multiple Criteria Decision Aiding” 3(4):6–7 Tsoukiàs A (2007) On the concept of decision aiding process: an operational perspective. Ann Oper Res 154(1):3–27. https://doi.org/10.1007/s10479-007-0187-z Whyte WF (1991) Participatory action research. Sage, Newbury Park, Ca White L (2009) Understanding problem structuring methods interventions. Eur J Opl Res 199(3): 823–833. https://doi.org/10.1016/j.ejor.2009.01.066 Yu W (1992) ELECTRE TRI: Aspects mèthodologiques et manuel d’utilization. Document du LAMSADE n°74. Université Paris-Dauphine, LAMSADE, Paris

Spatial Decision Support to Reorganize the Swiss Postal Network Florent Joerin

1 Introduction In 1999, Swiss Post noted that there were too many post offices and that the number of letters and parcels (at the counter) was decreasing. The management of the Swiss Post then requested a profitability analysis (cost/customers) which led them to consider major closures of post offices. However, this study generated an intense public and political debate at the national level (Le Temps, tenth of December 1999—see Annexe 1). The Swiss Post was then in a complicated and uncomfortable situation, under double pressure, partly contradictory. On the one hand, it was asked to reduce its operating costs and on the other hand, to maintain a high quality of service throughout the territory. In this difficult context, the company’s management decided to commit substantial resources to meet the challenge of the necessary reorganization of its postal network. A project team was set up within the public company to help the managers of each city to diagnose the spatial coverage of their territory, and to identify and evaluate alternatives to urban postal networks. Between 2001 and 2005, which corresponds roughly to the years of this project called “Réseau-Ville,” the number of post offices decreased from 3396 to 2531 units. Of these, 141 post offices were operated by private businesses, which was a novelty in Switzerland, at the time. The complexity of the Swiss Post’s situation is also revealed in its regulatory framework. Since 1998, the Federal Council (Swiss federal government) has imposed a profitability obligation on Swiss Post, an institution under public law. However, as a public service provider, Swiss Post is also required by law to ensure

F. Joerin (✉) HEIG-VD, Yverdon-les-Bains, Switzerland e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_2

35

36 Table 1 The postal network in 17 Swiss cities whose postal network was reorganized using the decision support method developed in 2000. The first eight cities were treated between 2000 and 2002. The next nine cities were processed between 2002 and 2005

F. Joerin Name Zürich Geneva Basel Lausanne Bern Alfalfa St. Gallen Lugano

Population 1,401,783 603,204 552,863 427,850 422,055 203,491 167,643 150,175

Name Winterthur Biel / Bienne Neuchâtel Thun Bellinzona Fribourg Schaffhausen La Chaux-de-Fonds Chur

Population 114,420 55,206 44,531 43,476 43,220 38,039 36,956 36,915 37,424

fair access to the service for the entire population. Since 2003, it has been required to maintain an infrastructure network that covers “the whole land and ensures that the universal service is available in all regions for all groups of the population within a reasonable distance” (Postal Ordinance 783.1 2003). In urban areas, the modification of the postal network is confronted with issues relating largely to the maintenance of neighborhood life and the role of the post offices as a social link, whose infrastructures constitute, for certain inhabitants, places of sociability that structure their daily activities. In rural areas, the same issues are at stake, but the emphasis is mainly on the economic survival of villages and the development of “peripheral” regions on the fringe of large urban centers. These tensions around the reorganization of the postal network in Switzerland echo much research on access to public services (Boyne and Powell 1991, Marsh and Schilling 1994, Wei et al. 2016, Tsou et al. 2005, Neutens et al. 2010a, b, etc.). For example, some researchers have focused on measuring spatial inequities (Geographic inequity) in access to health services (Schuurman et al. 2010; Apparicio et al. 2017). Others have looked at the inequalities generated by the spatial distribution of public spaces and parks in urban settings (Rigolon 2016). Furthermore, others have looked at the social impact of the location of education services (De Oliver 1998; Sharma and Patil 2022). Quebec (Canada) has also experienced a strong conflict over the maintenance of post offices in rural areas (Beaudry et al. 1998). This chapter has two main parts. The first describes the decision support approach that was used in this context to support the Swiss Post reorganizing its postal network. In the second part, a look in the rearview mirror is proposed to focus on the decision-making process in which this decision support approach was implemented. The use and role of information, often geographic, that has been used by the various decision-makers will be examined. This chapter is based on three complementary sources of information: the experience of the author, who was involved in the design of the decision support process and then its implementation in the first eight cities (2000–2002); a press review which made it possible to collect 44 articles from one of Switzerland’s major newspapers (Annexe 1); and finally, interviews conducted with local players from Swiss Post, directly involved in the reorganization of the postal network in 17 Swiss cities (Table 1).

Spatial Decision Support to Reorganize the Swiss Postal Network

37

Despite its imposing scope for a territorial decision support application, it is important to point out that this study focused only on a part of the general problem of the reorganization of the post office network. Indeed, it was limited to urban areas because, for internal organizational reasons, the Swiss Post has separated the studies of rural and urban areas.

2 The Decision Support Process The reorganization of post office networks in 17 Swiss cities (Table 1) was processed using an identical decision support approach. The decision support process was designed during a test application carried out for the postal network of the city of Lausanne. It is composed of a common core, identical for all cities, and of modular parts allowing to consider the particularities of the different urban contexts. The common core essentially includes: the general objective (of the reorganization) of the postal network, the definition, and the method of evaluation of the criteria for assessing the quality of the office locations and the network alternatives. The adjustable part includes some parameters of the criteria evaluation, as well as the set of parameters reflecting the preferences of the postal network managers in each city. The chosen multicriteria analysis method was Electre Tri-B (Roy and Bouyssou 1993) using the following preference parameters: criteria weighting, indifference, preference, and veto thresholds, as well as two reference actions, one bad, the other good (see Sect. 2.5).

2.1

Sources of Information

The main source of information used in this study was the census of population, buildings, and businesses (Geostat 1990). This source of information is provided by the Swiss Federal Statistical Office in the form of a 100-m grid covering the entire Swiss territory (Map 1). Thus, for each 100-m grid, the available data describe the population, the businesses, and the buildings. Concerning the population, the data used are notably the age category, the professional situation, or the mode of transport used to go to work. For buildings, the information used was mainly about the age of the buildings, their height, their surface, and their use (commercial, agricultural, industrial, etc.). For businesses, the data used describes the type of activity (secondary or tertiary) as well as the surface area of retail businesses. This representation in the form of a hectometric grid had a great influence on the decision support process, since all the reflections and analyses that have been carried out to evaluate the quality of the sites were then based on this elementary spatial unit of 100 m by 100 m. The phenomena considered, such as the density of inhabitants, jobs, or shops, were thus represented by continuous maps crossing administrative boundaries.

38

F. Joerin

Map 1 (Partial) representation of the hectometric grid used to describe the study areas. Each point is placed in the center of a hectometric grid cell. A first sorting of the meshes was done to eliminate those that are outside the urban fabric (forests, gravel pits, fields, non-buildable land, etc.)

As we will see, this database was supplemented by the results of a large survey of the Post’s customers (see Sect. 2.4).

2.2

The Main Steps

The decision support process consists of four main steps (Fig. 1). The first three aim to select, for each city, a strategic alternative for the reorganization of the postal network. The fourth step concerns the operational implementation of the chosen alternative. The first step was to carry out a survey of Swiss Post customers to identify their travel habits to urban post offices. This data, combined with spatial data provided by the Confederation, the Cantons, and the Cities, allowed to carry out, in the second step, a multicriteria analysis measuring the quality of the different locations in the urban area (quality of hosting a post office). This multicriteria analysis included all its usual components: definition of criteria, evaluation of the actions’ performances, definition of preferences, sensitivity analysis, etc. The multicriteria analysis thus essentially produced a suitability map showing the potential of each location to host a post office for the entire territory of each city.

Spatial Decision Support to Reorganize the Swiss Postal Network

39

Fig. 1 The main steps of the decision support process

This map was the starting point for the third stage, which aims to develop alternatives to post office networks. Several alternatives were thus considered for each city, varying the rate of reduction in the total number of post offices, but also the number of new offices opened (partly compensating for certain closures). On this basis, a strategic alternative was adopted for each city defining the order of magnitude of the final number of post offices, including the maintenance, closure, relocation, or opening of post offices. Since this alternative is mapped, these choices are located geographically in the different parts of the city. In the fourth stage, detailed analyses were carried out to specify the choices concerning each post office in each city. On this basis, the decisions to open, close, or relocate post offices were discussed with local stakeholders (not part of the public enterprise), essentially local public authorities (elected officials and administrative staff), and local associations. It should also be noted that in the event of disagreement between the choices made by the Swiss Post and the local stakeholders, the latter had the possibility of lodging an objection with a federal body (The Swiss Post Commission 2006), which then would study the arguments of both sides and makes a recommendation (PostComm 2022).

2.3

Definition of Objectives and Criteria

The definition of the general objective of the postal network was done in a working session with the managers of the postal network. This first step allowed us to quickly broaden the scope of thinking. Indeed, it appeared that if the reduction of operating costs was the main motivation, it was in itself an insufficient criterion, since the

40

F. Joerin

Table 2 Description of site evaluation criteria Criteria Accessibility from the places of residence Accessibility from the workplace Accessibility from commercial locations Accessibility from public facilities Accessibility from known locations or landmarks Public transportation

Ease of parking

Degree of visibility

Type of neighborhoods

Description Number of inhabitants “close” to the place considered. Number of employees “near” the location. Commercial surfaces (in m2) “close” to the place considered, all businesses combined. Number of public facilities (school, nursery, retirement home, hospital) “close” to the location. Number of landmarks, known places and heavily traveled places “near” the location. Quality of public transport service to the location. Depends on the number of lines, their frequency, and capacity. Number of car parking places potentially available at the location. Depends on the total number of places and the occupancy rate. Potential of frequentation of the place considered. This is measured by the number of passages in front of this place by private vehicles, public transport, or pedestrians. Importance of the post office in the life of the neighborhood where the location is located. A post office was considered to play a more important role in a neighborhood with fewer services.

Unit Number of inhabitants Number of jobs M2 of commercial space Number of public facilities Number of landmarks

Number of public transportation vehicle passages (nearby) Number of parking spaces

None (ordinal scale)

None (ordinal scale)

cheapest network is certainly not the best. The objective of the postal network was thus defined from three angles: (1) to offer easy access from the various places of activity of the population (home, work, shops, schools, etc.), taking into account the diversity of the modes of transport used; (2) to ensure good visibility of the postal offices from the main places of passage; (3) to maintain a role in the animation of neighborhood life. These general objectives were then used to define a set of nine criteria, common to all cities, to assess the quality of a potential location (Table 2).

2.4

Evaluation of Actions’ Performances

The post office network is a public basic service to the population that must respect a degree of social equity in access to the service offered. Table 2 shows that several criteria (4 out of 9) refer to the notion of spatial accessibility. These criteria propose a measure of the ease of access to postal offices from different places of activity

Spatial Decision Support to Reorganize the Swiss Postal Network

41

(residence, work, shops, and public facilities). By diversifying the measure of accessibility to the postal service and by not only considering access from the place of residence, it is possible to consider the needs of a larger proportion of the population. This section illustrates the spatial analysis procedures used to evaluate the location criteria. However, since it would take too long to detail all the procedures applied, it focuses on the accessibility criteria that are most common. The measurement of accessibility has been the subject of much research. A first proposal, formulated in 1959 by Hansen, is based on the analogy of the gravity model, where accessibility is perceived as a “force of attraction” (Hansen 1959). A few years later, Wachs and Kumagi (1973), insisting on the relevance of the study of accessibility in social studies, considered that accessibility (or its measurement) should incorporate not only the trip, but also the possibility of achieving the goal of the trip (e.g., compatibility of business hours), as well as the possibility of using a transportation (e.g., access to the household car) to achieve the trip (Wachs and Kumagi 1973). In a similar vein, Tagore and Sikdar (1995) posit that a person’s ability to get somewhere depends on the accessibility of the place and the mobility of the person. They thus propose a modification of the gravity model analogy to include mobility factors, such as access to public transportation or household income. In addition, several authors have studied the accessibility of specific services. Talen and Anselin (1998), for example, assessed accessibility (by different modes of transportation) to public parks in the city of Tulsa, Oklahoma. Rigolon (2016) also studies access to urban parks across socioeconomic and ethnic groups. Schuurman et al. (2010) consider kernel density estimation and a gravity model to measure potential spatial access to PHC physicians. They ultimately applied a modified version of the gravity model. Apparicio et al. (2017) compare discrepancies in results for the geographic accessibility of health services computed using six distance types (Euclidean and Manhattan distances; shortest network time on foot, by bicycle, by public transit, and by car), four aggregation methods, and 14 accessibility measures. Thus, the literature offers several alternatives for a general model, based on the gravity model first proposed by Hansen (1959), where the attractiveness of a place is contrasted with the effort required to reach it. These models are often distinguished by the way in which this effort is measured, also called “metrics” of distance. However, if the alternatives in the evaluation of these two dimensions (attraction, travel effort) are numerous, the mode of aggregation is quite common. It generally takes the form of a sum, weighted or not (Witten et al. 2003; Smoyer-Tomic et al. 2004). The accessibility measure used in this study is based on a fuzzy set approach (Joerin et al. 2001a). Accessibility preferences were derived from the results of a survey of about 8000 Swiss Post customers in about 50 post offices throughout the country. Accessibility evaluation considers a destination to be accessible if it is close to places of importance, i.e., highly frequented or significant in the city’s dynamics. The two main measures are therefore proximity (between the destination and the

42

F. Joerin

Fig. 2 Example of a proximity function. Distance is the spatial distance between the (potential) location of a post office and the places of origin of trips to this location (for a possible post office)

place of origin of the trip) and the level of importance of the place of origin of the trip. Proximity is measured by fuzzy sets defined by two thresholds: a complete proximity threshold (S1) and a no proximity threshold (S2) (see Fig. 2) (Theriault et al. 2005). Thus, a place is considered “close,” to a company, for example, if the distance is less than S1. Conversely, it is considered as “distant” if the distance is greater than S2. If the distance is between these two thresholds, the proximity will be “partial” (between 0 and 1, calculated by linear interpolation). The measure of distance was simply metric and Euclidean. Indeed, the constraints of the mandate did not allow for the consideration of actual distances on transportation networks, nor did it allow for issues related to the opening hours, cost, or availability of transportation modes. (These dimensions have indeed been considered in the four steps of the decision process, the implementation—Fig. 1). However, the consequences of this simplification seem in this case very acceptable, since on the one hand very few situations in the city require the use of a mode of transport to get to the post office, and on the other hand the places of activity considered (work, residence, and public facilities) are complementary in terms of schedules. The relative importance of the place (origin of the trip to the post office) was measured for each criterion in a specific way. In the case of the criterion of accessibility from workplaces, the relative importance of the location is, for example, measured by the number of employees. In other words, the accessibility unit is, in this case, the number of workplaces (jobs) near the destination. For the criterion of accessibility from known places or landmarks, the level of importance is expressed by a weighting, which is higher for the most known places or the most frequented by the inhabitants of the city (public squares, historical buildings or buildings of remarkable architecture, etc.). Accessibility criteria were so evaluated by combining the level of proximity (μj) and the relative importance (Nj) of the place (of origin of trips.) according to the following equation:

Spatial Decision Support to Reorganize the Swiss Postal Network

43

Fig. 3 Proximity functions: type of peripheral neighborhood motorized travel

Ai =

μj ∙ N j

ð1Þ

i

with: Ai: Accessibility level of destination i; μj: Degree of proximity to Nj: Relative importance of the place j. The proximity thresholds (S1 and S2) were calculated from the customer survey. First, a typology of clients was defined, based mainly on the type of postal offices usually used (downtown, residential, in a shopping mall, etc.) and the mode of transport used to get there. Then, for each type of customer and each mode of transportation, it was possible to observe the distribution of trip lengths to the post office (Fig. 3). Finally, a mathematical function (of the Gamma type, Eq.2) was fitted statistically (by the method of least squares) to the distribution of the distances travelled. The proximity thresholds, which are specific to the type of office, the type of customers, and the origin of the trip to the post office, were finally derived from the parameters of the Gamma function fitted on the distribution. F ðxÞ = xn ∙ eðmxþCÞ for x > 0 F ðxÞ = 0 for x ≥ 0

ð2Þ

n, m, and C are the constants that allow the fitting of Gamma functions on the displacement distribution. Note that the density of the office network in Switzerland was so high at the time of the survey that it was reasonable to assume that the observed trips do indeed express preferences.

44

2.5

F. Joerin

Synthesis by Multicriteria Analysis

The overall quality of a location was evaluated by performing a synthesis by multicriteria analysis of the nine criteria considered (Joerin et al. 2001b). The multicriteria analysis method used was the Electre Tri method (Roy and Bouyssou 1993; Yu 1992). This method compares each alternative, here locations or more precisely meshes of the hectometric grid, to alternatives called reference actions. In this implementation, two reference actions were considered: one good and one bad. To make it simple, alternatives that are similar or better than the “good reference” were classified as “good” and conversely, those that were similar or worse than the “bad reference” were classified as “bad.” The others were classified as “doubtful” or “uncertain.” The multicriteria analysis was carried out twice by modifying some of the calculation parameters (cut-off threshold) to incorporate a sensitivity analysis. The final result combined these two results and takes the form of a classification into five ordered classes: “good-good” (alternatives that were classified both times as good), “good-doubtful,” “doubtful-doubtful,” “doubtful-bad,” and “bad-bad.” The preferences of the decision-makers were thus taken into account in this synthesis by the definition of the reference actions, but also, for each criterion, by the definition of weights and various other parameters of the Electre Tri method (indifference, preference, and veto thresholds). Even if the reference actions can be freely defined, it is preferable, in a spatial problematic like this one, to choose reference actions that are real places (post offices) on the territory. In the context of this study, all the parameters expressing the preferences of the decision-makers, including the reference actions, were defined by working groups composed of the managers of the post office network in each of the cities. In order to define the reference actions, they were asked to rank a set of post offices between “good” post offices, i.e., well located in the city, and “bad” (poorly located). This possibility to define preferences through the choice of real locations was a very important advantage of the Electre Tri method, since the decision-makers were thus solicited within a familiar framework of reflection, in which they could rely on numerous points of reference, such as their own knowledge of the territory and of the qualities of the postal offices (The definition of a weighting, on the other hand, is much more abstract). In addition, to facilitate interpretation, the result of the multicriteria analysis was overlayed (transparently) on a “standard” topographic map, displayed in black and white (see Map 2). The map presenting the results of the multicriteria analysis was then submitted to the group of local decision-makers. The latter could thus rely on their global or local knowledge of the territory, which favors a geographical or urban interpretation. For example, a highly contrasting summary map where “good” and “bad” places are very close together is often not very satisfactory, since at the scale of the urban area, transitions are generally spatially more continuous.

Spatial Decision Support to Reorganize the Swiss Postal Network

45

Map 2 Example of a suitability map showing the result of the multicriteria analysis (level of location suitability for a post office)

The choice of reference actions required several iterations, which were easily accomplished using an Electre Tri module integrated within a Geographic Information System (GIS). The user could thus quickly choose two reference actions and directly observe the result on the synthesis map produced with Electre Tri (Joerin 1998; Joerin et al. 2001b). This fast and convenient mode of interaction were very important in preference building, as it allows the user to learn about the spatial structure of the city. After a few hours of practice, the user is usually able to anticipate the result of a multicriteria analysis, considering the location of the reference actions. Progressively, he learns that if this place is defined as good, then such other would be good too and such other would be doubtful. The decision support approach is thus completely in line with a constructivist approach (Meinard and Tsoukiàs 2019; Tsoukiàs 2008; Landry 1995). This learning process is, in our opinion, the essence of this multicriteria analysis, which in fact aims to produce a spatially coherent preference map. However, we will see in Sect. 3.1 that the user (of the multicriteria analysis) and the decision-maker are different people, which obliges us to consider with more nuance the reality of this learning process. Moreover, the notion of decision-maker also requires some clarification (see Sect. 3.1).

46

2.6

F. Joerin

Design and Evaluation of Network Alternatives

The synthesis by multicriteria analysis produced an overall suitability map covering the entire study area (Map 2). By overlaying this map with the current network, it was possible to diagnose the geographical location of each post office. However, the set of best locations does not constitute the best network, since the spatial coverage of the urban territory must also be considered. As shown in Map 2, the best locations are clearly concentrated in the city center, and it would be inefficient to concentrate there all the post offices. The network alternatives were therefore composed of relatively less favorable, but strategic locations in terms of spatial coverage. The map produced by the multicriteria analysis thus identifies locations that are not the best on the scale of the entire study area but are still the most favorable in their local geographic context. For each city studied, a set of network alternatives was defined based directly on the suitability map, without considering the current network of post offices. These alternatives (generally six) are progressively divided between the current network and a minimum network which generally included a number of offices equivalent to 75% of the actual number of offices. (The post office management has set a maximum reduction in the number of offices of 25%). It can be noted that even though the locations of the minimum network were chosen without reference to the current network, it appeared, by superimposing the two maps, that for about two-third of the locations of the minimum network, there was actually a post office. The design of the network alternatives was done empirically for two reasons. First, the use of an mathematical optimization seemed excessively cumbersome. Secondly, an empirical approach makes it possible to integrate into the design work the urbanistic knowledge, resulting from personal experiences in the city, which are important and significant, even if they could not find their place in the multicriteria analysis. The different network alternatives were finally evaluated by returning to the objectives set in the first step of the process. Thus, the criteria for evaluating the network alternatives were basically the same as the criteria for evaluating location suitability. However, the mode of evaluation was different, since it is no longer a question of considering each location separately, but rather the overall quality of the network alternatives. Accessibility from residential locations, for example, was evaluated using specialized software that simulates travel on the transportation network between home and the nearest office, taking into account the specific configuration of urban mobility networks. This calculation made it possible to evaluate the percentage of the population within 5, 10, or 15 min of walking or driving distance from a post office. Table 3 illustrates the form of the final result (the values are fictitious because the real results are the property of the Swiss Post). The process carried out for each city, which requires a very large amount of data and processing corresponding to almost 2 months of work, results in a table, on one page, presenting in a very synthetic way the advantages and disadvantages of a set of network alternatives, evaluated on

Spatial Decision Support to Reorganize the Swiss Postal Network

47

Table 3 Alternative comparison table, here for two alternatives Residents Average distance (m) Nearby (any mode of transportation) Travel time on foot ≤10 min ≤15 min Travel time by car ≤5 min ≤10 min People over 75 years old Less than 600 m away Jobs Nearby Poles of attraction Walking distance Close by car Parking Number of places close by Public transport Waiting time in minutes Visibility Scale from 1 to 5

Alternative 1

Alternative 2

600 58%

1000 46%

54% 78%

32% 54%

81% 99%

55% 87%

85%

80%

61%

58%

59% 85%

73% 96%

18

39

15

3

2

4

NOTE: These results are presented as an illustration. Indeed, in order to respect the property right on the results of the Swiss Post, some values are modified. They do not correspond to any alternative for any city

about 10 criteria. It should be recalled that these alternatives are strategic alternatives allowing decision-makers to have an opinion, on the one hand, on the adequate number of postal offices in the city under consideration and, on the other hand, on the sectors where the density of postal offices should be increased or reduced. However, even though relatively precise locations (100 × 100 m) were considered in the multicriteria analysis, the scale of the work (the whole city) does not allow for the consideration of local configurations that can have a great influence on the quality of a location. Finally, it is important to mention that the presentation of the results was also associated with a sensitivity analysis which allowed to measure the effect of uncertainties, the quality of the data, and some arbitrary choices sometimes necessary to carry out certain calculations. The sensitivity analysis carried out allows us to qualify the interpretation of the results obtained for the different alternatives. Thus, a difference between two alternatives was considered significant if, and only if, it was greater than the magnitude of the uncertainty evaluated by the sensitivity analysis.

48

F. Joerin

3 A Decision Support Approach in a Decision-Making Process The previous section presented the approach from a technical angle, that of information and territorial decision support tools (Joerin et al. 2001a, b; Graillot and Waaub 2006). As a reminder, the main tools used were statistical analysis, multicriteria analysis, and spatial proximity analysis. This new section moves away from these technical dimensions to focus on the decision-making process in which this territorial decision support approach is embedded. To do this, three sources of information are used. The first is the personal experience of the author, who at the time was part of the private consultancy firm that designed and implemented the approach for the first eight cities. The second source of information is a press review of 44 press articles (Appendix 1) taken from Switzerland’s main French language newspaper (Le Temps). These articles cover the period 1998–2006. The third source of information comes from three interviews conducted with actors, employees of the Swiss Post, directly involved in the reorganization project of the postal network of the 17 cities concerned. The proposed analysis attempts to answer the following question: By whom and how was the information (produced by the decision support approach) used to make decisions about the reorganization of the urban post office network and then to implement those decisions? Before going any further, it is useful to specify that the territorial decision-making process described above was part of a much broader social, political, and economic context. First, the need to reorganize the urban postal network is a component of the reorganization of the overall Swiss postal network (including outside cities), which is itself a component of the reorganization of the company itself (the Swiss Post). We can thus observe a great deal of media attention around the reorganization of the Swiss Post company, including but not limited to the reorganization of the postal network. Furthermore, several projects of reorganization of the postal network followed one another: “Réseau 2000” (Le Temps, December 6, 1998), “Optima” (Le Temps, August 23, 1999), then the project “Réseau-Ville” in which this decision support approach is included. The first concrete modifications to the network were made using the results of this last project. However, a few years later, the Swiss Post launched another project, called “Ymago” (Le Temps, July 7, 2004), which extends the reorganization strategy by further diversifying the delivery mode of the universal postal service. Another element of the context to be underlined is the importance of the social debate and the political struggles concerning the reorganization of the postal network. On several occasions, the (federal) minister in charge of these stakes intervened in the media to explain and defend the project (Le Temps, January 27, 2000, June 16, 2000, March 22, 2001, etc.). In addition, the federal parliament voted and narrowly refused a motion imposing a moratorium on the restructuring of the postal network (Le Temps, October 05, 2001). The unions also launched a popular

Spatial Decision Support to Reorganize the Swiss Postal Network

49

initiative “Postal services for all” which was also finally rejected in a democratic vote by the Swiss people (Le Temps, September 23, 2004). However, in order to anticipate and respond to this initiative, the Swiss parliament had previously adopted and implemented a counter-project (a law) that exhaustively list the services that fall under the universal service (including in terms of spatial accessibility) and introduced the obligation for the Swiss Post to produce an annual follow-up of the evolution of the postal network so that an external auditor could verify the quality of the service offered throughout the territory (Le Temps, August 11, 2004). These contextual elements are important because they put into perspective the role of the decision support approach in the more general decision-making process. Indeed, as we shall see (Sect. 3.2), the information produced considerably facilitated the implementation of the restructuring of the urban postal networks, especially during the negotiations with the local authorities of each city. Interview N02: “The study was with us all the time. All authorities received the results of the study. The maps were used regularly. That was the justification. It gave us very detailed arguments.” However, the outcome of these negotiations is certainly not only the result of the quality of the study itself, but also of a more general negotiation framework, including at the national level. In other words, the agreements that were reached locally were also fed by agreements built at higher levels, such as those concerning the definition and control of universal service.

3.1

Who Decides What?

Decision support is inconceivable without a decision-maker. However, identifying the decision-maker in a real process such as the one we are considering here is more difficult than it might seem. Flourentzou (2001), citing Roy (1985), notes that identifying the decision-maker is in fact a matter of specifying the objectives that the decision-maker must face, in other words, the object of the decision. In the case we are considering here, the decision concerned “the choice of the strategy for reorganizing the postal network.” However, this global or final decision was constructed by a series of small decisions that progressively gave it an orientation. The question “who decides what?” thus makes it possible to highlight this diffuse process of construction of the decision within an institution such as the Swiss Post. There are many definitions of the decision-maker. In the field of decision support, Bernard Roy (1985) considers that the decision-maker “designates (. . .) the entity which appreciates the possible and the finalities, expresses the preferences and is supposed to make them prevail in the evolution of the process.” However, in the situation dealt with here, this “entity” includes many people, because the various responsibilities are distributed among the hierarchical levels of the institution (the Swiss Post). In such a system, where actors, responsibilities and influence intersect, the notion of decision-maker becomes abstract, whereas it is essential for “decision support.” However, if a multitude of actors have a certain influence on the decision, they do not all bear the same responsibility for this decision. A citizens’ committee

50

F. Joerin

Table 4 Type of actors involved in the decision-making process, specifying their composition, their responsibilities, and the types of information used or produced in the decision process Hierarchical level

Designation Composition

Responsibilities

Information used or produced

Level 4 RéseauVille Design project office team 1 project 1 project manager, manager, 4 analysts 4 analysts Methodological proposals: • General decision support approach • Facilitation (definition of questions) • Criteria proposals • Method of criteria calculation (parameterization) • Multicriteria analysis method

Used • Customer survey data • Public Geodata Produced • Maps, reports, presentations

Level 3

City network manager 17 managers and their teams Weighting Ranking of a sample of post offices Validation of suitability maps Validation of network alternatives Negotiation with local actors Operational realization Used • Semiinteractive maps • Report • Oral presentations Produced • Preferences for each city

Level 2 National Network Manager 1 director and his team Validation of criteria Validation of the general approach Choice of the strategy network alternative Resources allocation for the study Resources allocation for operational implementation Used • Maps • Report • Oral presentations

Level 1

Head of the Swiss Post 1 director and 8 members of the board Validation of the resource allocation for the study Validation of the strategic network alternative Validation of resource allocation for operational implementation Public and political support

Used • Presentation of the approach and its results (20 min) • Others?

campaigning to maintain a post office may have a great deal of influence, but it bears no responsibility for the decision that will be made. On the other hand, some Swiss Post managers or directors may face economic or legal liabilities or lose their jobs or status if the consequences of the decision turn out to be negative. In the context of this study, we propose to distinguish four distinct types of actors involved in the decision-making process, each with a particular influence and responsibility on the decision (Table 4). At the first hierarchical level (L1, Table 4) is the director of the Swiss Post. He is the main political and media spokesperson for the general restructuring project of the Swiss Post as a whole, but also or especially for the reorganization of the postal network, which is particularly controversial. In 2000, the current director of the

Spatial Decision Support to Reorganize the Swiss Postal Network

51

Swiss Post resigned after only 17 months in office. According to the media, this resignation was mainly due to the difficulties encountered in the project to restructure the postal network (Le Temps, 12 January 2000). The subsequent director met the media very often to support and explain the project (73 mentions in the 44 articles collected). In particular, he was present at the various press conferences organized in each city to explain the restructuring of their urban postal network (Le Temps, 28 April 2001; Le Temps, 21 November 2001). At the second hierarchical level (L2, Table 4) is the person responsible for the postal network at the national level. This actor is the initiator and manager of the “Réseau-Ville” project. He personally defended the study and its results at a meeting of the board of directors of the Swiss Post, during which the decision to validate the decision support approach and its results were officially adopted. Following this decision, the managers of the post office networks in each city were given the task of implementing the general strategy. As for the director of the Swiss Post, the success or failure of the reorganization of the postal network certainly could have a direct influence on the appreciation of his or her leadership qualities. At the same hierarchical level, although this does not correspond exactly to the structure of the company, can be placed as the head of the regional postal networks in the Frenchspeaking part of Switzerland. This person has a similar level of responsibility to the previous one because the French-speaking part of the territory is delegated to him. It is worth mentioning this actor because he was often present in the French-speaking media. The third hierarchical level (L3, Table 4) is that of the postal network managers in each city. These actors had a great deal of influence, since they set the subjective parameters of the multicriteria analysis (such as weights and reference actions) expressing for each city, the preferences of the Swiss Post regarding the expected qualities of the future networks. Their level of responsibility is relatively high since they are responsible for the quality and performance of the postal network in each city. These actors only appear in the media during press conferences that directly concern the reorganization of their own urban postal network. The fourth hierarchical level (L4, Table 4) is made up of two groups that worked very closely together: the private consultancy firm and the “Réseau Ville” team that was formed within Swiss Post to accompany the project. The latter team independently conducted the study of the nine cities that followed the first eight carried out jointly with the consultancy firm. By working directly with the data, producing maps, tables, and reports, these two units certainly had an influence on the final results through certain technical choices (method of modelling the criteria, choice of the multicriteria analysis method, proposal of criteria, etc.). However, their responsibilities were limited to the quality of the method, the data, and the results. The three interviewees were at hierarchical levels three and four. These persons were so very directly involved in the decision-making process, had direct interactions with the higher hierarchical levels, but also with actors involved in the decision-making process who were not part of the Swiss Post, such as local actors (political and technical) who were part of the discussion tables.

52

3.2

F. Joerin

What Role for Information in the Decision-Making Process?

When a decision-maker is looking for the foundations of his decision, he/she first calls upon his own cognitive structures or representations (Flourentzou 2001). If these prove insufficient, he/she looks for new information to enrich his representations (Breux 2008; Charef 2010; Joerin et al. 2010). Either this new information is assimilated, integrating itself into the initial representation, which can thus be enriched, or the representation is accommodated (deconstructed, reconstructed) to integrate the new information (Espinasse 1994; Joerin et al. 2010). This enhancement of the cognitive model requires interaction between the decision-maker and the real system, or representations of the real system, which can take the form of texts, maps or indicator tables for example. This interaction allows the decision-maker to build an understanding of the system and ends with the stabilization of the cognitive model (model coherence) (Espinasse 1994). Thus, in order to understand the role of information, it is necessary to study the modes of interaction between this information and the actors involved. In the next section, the four hierarchical levels identified above are taken up with this in mind.

3.2.1

The Construction of a Complex Problem

Some signs seem to attest to an evolution in the cognitive model of the actors of the Swiss Post. Faced with the complexity of the issues related to the restructuring of the postal network, issues that are in part contradictory and that go far beyond those of economic profitability, the managers of the urban postal networks felt the need to build up their own vision of the problem to be solved. It was so for this purpose that they sought the help of a consultancy firm with skills in urban planning, geography, cartography, and decision support. Through multiple interactions, the decisionmakers of the Swiss Post (Fig. 4), essentially at the L4, L3, and L2 hierarchical levels (Table 4), gradually built up knowledge concerning the historical evolution of urban systems (including postal networks) and urban practices (including customer behavior). The problem of reorganizing postal networks has thus been progressively enriched, starting from issues of cost and profitability to include issues related to modes of travel, urban dynamics and practices, urban landmarks, population presence by time of day, etc. This enrichment is reflected in the press review. The first articles collected (from November 1998 to December 1999) essentially mention economic concerns about profitability and cost. However, as soon as the first announcements of possible closures of urban postal offices were publicly made, a change in vocabulary appeared. The comments of the representatives of the Swiss Post quickly incorporated new issues, particularly those raised by local actors (the social role of the post offices, neighborhood life, survival of villages, etc.). In a February 2000 press article, the director of the Swiss Post explains, for example, that: “It is wrong,

Spatial Decision Support to Reorganize the Swiss Postal Network

53

Fig. 4 The iterative and interactive process of multi-criteria analysis

however, to talk about office closures. We should rather talk about adapting the network. We want to put the post offices where the people are” (Le Temps, February 29, 2000). This evolution of the comments reflected by the media appears even more clearly during the public presentations of the results of the multicriteria decisionmaking process (Le Temps, April 28, 2001, Le Temps November 21, 2001), during which the nine criteria are the basis of the proposed strategic network alternatives were presented. The enrichment (complexification) of the problem to be solved is also reflected in the words of the interviewees: “The (decision support) study allowed for a complete paradigm shift. We understood that we had to be as interested in the places where people work as in the places where people live” (Interview N02). “The Swiss Post thus gave itself significant resources (mandate, purchase of data, training of a team, etc.) to have an argument that was not only based on a profitability analysis. (. . .) The study partly objectifies and complexifies the notion of public service (various modes of access, from various locations)” (Interview N01). It may also be noted that the reports produced for each city contained an important introductory section, generally much appreciated, which presented some background on the evolution of cities (metropolization, evolution of spatial mobility, emergence of network cities, etc.). This historical section certainly helped the Swiss Post’s decision-makers to situate the problem in a more general framework: the reorganization of the post office network is part of a general dynamic of

54

F. Joerin

reorganization of urban networks. They were so able to find explanations for their current situation, as well as a motivation to change previous practices.

3.2.2

Interaction with Information

As already mentioned, adopting a constructivist approach to decision support (Tsoukiàs 2008), the evolution of the cognitive model of the actors involved requires interactions with the information. However, as Table 4 shows, the higher the level of responsibility, the less direct interaction with the information produced can be observed. For example, the analysts who interact directly with the data and decision support tools did not participate in the negotiations with local actors, which were crucial to decision-making: “We prepared more precise maps for them, we explained them to them, but we did not participate in the discussions” (Interview N03). One may thus wonder how this construction; this complexification of the problem can still be done for the actors with the greatest responsibilities? At the highest hierarchical level (L1 and L2, Table 4), that of the director and the managers of the postal networks at the national level, direct interaction with the information collected and produced was relatively limited. Through short sessions or the reading of mid-term and summary reports, their role was essential to validate the approach (e.g., choice of objectives and criteria) and the overall outcome. This interaction probably took the form of reading synthetic alphanumeric documents (texts, maps, and tables) (about 20 pages). It is safe to assume that these interactions were complemented by discussions with lower-level managers. The lower hierarchical levels (L3 and L4, Table 4) were both involved in the production of the multicriteria analysis. The interaction with information was in this case much more important. We have already underlined in Sect. 2.5 the important role, on the construction of preferences, of the numerous iterations that took place during the production of the multicriteria analysis. Interactions with the available information, the decision support tools, but also between the analysts were almost daily at the fourth hierarchical level (Réseau-ville project team and the consultancy firm). With the third level (managers of postal networks in different cities), the interactions were approximately monthly during the 4 months that were generally necessary for the study of a city (at least for the first eight). It can also be noted that only people at the fourth hierarchical level were really using (operating) the decision support tools. From level 3 onward, interaction with information had always been about the (intermediate or final) results produced with the decision support tools (maps, tables, etc.) (Table 4). Indeed, even if the interactions (and iterations) with the decision support tools were rapid, they require sessions that would have been too long or too numerous for the managers of urban postal networks (the production of a synthesis map generally takes 1 or 2 days of work). The approach followed to progressively define the subjective parameters needed for the multicriteria analysis can be seen as an empirical version of an inference method (Dias et al. 2002; Cailloux et al. 2012). Initially, decision-makers answered simple, concrete questions that essentially consisted of ranking a sample of post

Spatial Decision Support to Reorganize the Swiss Postal Network

55

Fig. 5 Iterative process applied to set the preference parameters

offices constituted to reflect the diversity of situations specific to the city being treated. The rankings produced were accompanied by a moment of discussion to identify the reasons underlying the result produced. Decision-makers also answered questions aimed at defining three sets of weights (Fig. 5). Following these sessions, the decision support experts (L4, Table 4) conducted an iterative series of multicriteria analyses, modifying the preference parameters (indifference thresholds, preference, veto) to best replicate the rankings formulated by the decision-makers. These iterations resulted in an initial set of subjective parameters (Fig. 6). Approximately 1 week after the first session with the city’s postal network managers, a new meeting allowed them to assess the suitability map (Map 2). They could thus observe the conformities and discrepancies between the result of the multicriteria analysis and the ranking they had previously produced for a sample of post offices. They could also observe which classes of suitability were associated with all the post offices in the city. Often, this discussion identified results that were considered inconsistent or unsatisfactory. It was then possible to modify the rankings or weightings produced in the first session, which required entering a new stage of iterative multicriteria analysis before organizing a new session to present a new version of the suitability map. The cycle ended when the decision-makers are satisfied with the suitability map. Thus, even though the city’s postal network managers were not present when the multicriteria analyses were conducted, the discussions around different versions of the suitability map certainly fed a learning process. This form of interaction, from session to session, allows us to consider that the suitability map, which was used to design the network alternatives, did indeed express the preferences of the Swiss Post as to the best locations for postal offices.

56

Fig. 6 Example of preference parameters

F. Joerin

Spatial Decision Support to Reorganize the Swiss Postal Network

3.2.3

57

Credibility and Scientific Legitimacy of Decision Support

The decision support process that was carried out to reorganize the postal networks of the 17 Swiss cities provided a basis of scientific legitimacy for the Swiss Post. This was useful in two main ways: on the one hand as a general argument of scientific rigor and quality, and on the other hand as a detailed and complete expertise of the problem to be solved. The general “scientific” dimension of the approach was, for example, put forward by the Post Office officials during the public presentations of the strategic variants chosen for each city. The press article describing the very first of these, in Zürich, specifies that the variant presented was “(. . .) elaborated according to “purely scientific” criteria, following a method conceived, (. . .) by a private company specialised in the elaboration of decision support models (. . .)” (Le Temps, April 28, 2001). Similarly in Geneva a few months later (Le Temps, November 21, 2001): “To avoid doing anything questionable again, La Poste called on science. The questioning of some 18,000 people, a multicriteria analysis and no less than a dozen thematic maps were used to draw up the future “ideal network” of the city of Geneva”. It should be noted, however, that these arguments of scientific legitimacy seem to refer to a scientific objectivity that does not correspond to the decision support approach that was used. Indeed, as already mentioned, the latter is rather based on a constructivist approach that gives an important place to its subjective dimension. The other form of legitimacy produced by the decision support approach refers to the knowledge produced, which enabled the postal service’s stakeholders to have more detailed and complete information on all the issues. This form of scientific legitimacy became particularly useful when the Post Office set up discussion groups with local stakeholders to discuss the operational implementation of the chosen strategic variants. “The study and the method gave the Post Office solid and credible arguments to discuss with the authorities, especially on the basis of the criteria. The method and the data were reliable: inhabitants, days versus nights, public transport services, etc.” Interview N01. “We were going to present this in the neighbourhoods. We explained the process. It was very useful. Sometimes we asked the “RéseauVille” team to complete or clarify certain maps.” Interview N02. It is interesting to note that in the two previous quotations, the interviewees mention the criteria maps, but not the summary map resulting from the multicriteria analysis. This suggests that the multicriteria analysis contributed less significantly to the legitimacy of the approach than the more detailed and less aggregated information presented by the criteria maps.

58

F. Joerin

4 Conclusion This chapter described a decision support approach based on the use of geographic information. The originality of the approach used lies partly in the combination of different tools, such as spatial analysis and multicriteria analysis, which are still rarely used together in a decision support approach in a real situation. It is also to be found in the size of the study (relative to current practices in territorial decision support), which required the work of a dozen people over 2 years to deal with the first eight cities, and then five people over 3 years for the next nine. This chapter then focuses on the lessons learned from this decision support experience. A reflection is thus proposed on the role that the information produced took in the decision-making process. A general observation can be made first: the information produced by the decision support tools played an important role in the decision-making process that led to the choice of a strategy for the reorganization of urban postal networks in each city and its subsequent implementation. However, it has been also observed that the decision-makers with the highest responsibilities finally interacted little with the information collected, managed, and produced during the decision support process. This second observation somewhat challenges one of the main principles of decision support when it is approached from a constructivist perspective. According to this approach, decision support does not (or rarely) consist of proposing a recommendation to the decision-maker, but rather of helping the decision-maker to construct his or her preferences. The question then is: how can decision support feed a learning process if the decision-makers who have the most responsibility interact little with the information produced? However, based on interviews and the press review conducted 20 years later, it seems relevant to consider that the decision support process fed an institutional rather than an individual learning process. For some people in the institution, this learning process was more intense, for others probably much less so, but overall, a new conception of the problem posed by the reorganization of the postal network spread within the organization. Through the decision support process, the Swiss Post has broadened its view of the problem of reorganizing the postal network. A link was established between the problem to be solved and a more general problem of redefining cities. It thus appeared that the postal network had been built according to a logic adapted to a territorial organization that no longer existed. The sources of the problem are therefore partly to be found in the transformation of cities, the modification of travel patterns, and the reorganization of family structures. It also became apparent that even if the cost of operating the network of post offices was an indicator of dysfunction, the problem was not limited to an economic one. The problem initially considered as economic and logistical has thus gained a complementary dimension: that of urban planning and territorial organization. Thus, if the objective is no longer limited to “reducing the network of post offices” but rather to reorganizing it, it becomes possible to accompany the closure of poorly located post offices by the

Spatial Decision Support to Reorganize the Swiss Postal Network

59

opening of a few new post offices, thereby greatly reducing the negative impacts on accessibility to post offices of a much less dense network. Institutional learning is also revealed by the fact that the “Réseau-Ville” project team, formed during the study, has remained in place. This team produced particularly during the years the indicators demonstrating the maintenance of the quality of universal service throughout the territory. Finally, about areas for improvement, the study was not always understood as a decision support approach, particularly or certainly because of the difficulty, in a large institution such as the Swiss Post, of reaching the most influential decisionmakers. Thus, for some, the study was perceived as a relatively “traditional” expert study. It would certainly have been desirable to involve more directly in the decision support process the decision-makers with the highest responsibilities, especially during the sessions that were intended to establish preferences concerning the organization of urban postal networks. Their involvement would perhaps have made it possible to better anchor in the institution, on the one hand, the change of paradigm implicit in a constructivist decision support approach and, on the other hand, the importance of a better mastery of the company’s spatiality (spatial data and organization, territorial dynamics, etc.). Indeed, 20 years after the start of this project, awareness of the importance of the spatial dimensions of the enterprise is still being consolidated (Interview N01 and N03). Finally, this decision support experience has provided some more operational lessons that could be useful if a similar situation were to arise. Here are two final recommendations. The first recommendation is to keep decision-makers within their field of competence. In other words, care should be taken to provide them with a framework for thinking that is familiar to them and to avoid abstract issues as much as possible. In this experiment with the Swiss Post, we tried to ask them questions based on concrete and real situations. We made extensive use of cartographic representations so that the decision-makers could draw on their knowledge of the city and the post offices they visit regularly. The coherence of the results obtained is then defined in particular as spatial coherence. The second recommendation concerns the role of decision support or, more precisely, of the person providing this support. Indeed, in certain situations, relatively frequent in the field of territorial decision, the decision-makers do not ask for help to make a better decision, but to feel in position to decide. In other words, the decision support process must allow them to progressively enrich their understanding of the problem to be solved and the stakes of the decision to be taken. From this enriched understanding, they will be able to argue and explain their choice. They can also communicate transparently about the uncertainty or risk of the decision to be made because the decision support process has enabled them to reduce this uncertainty as much as possible. In this situation, the choice of the best alternative is probably less important or less useful to them than the reasons for the choice.

60

F. Joerin

Annexe 1 List of press articles collected in the daily newspaper “Le Temps” www.letemps.ch Date 12.11.1998 12.11.1998 6.12.1998 23.08.1999 9.06.1999 10.12.1999 13.12.1999 10.12.1999

12.01.2000

12.01.2000 18.01.2000 27.01.2000

29.02.2000 16.06.2000 4.11.2000 1.12.2000

11.02.2000

26.02.2000 1.03.2000

Title Une stratégie floue—La dimension du réseau postale n’est pas définie Comme Une lettre à la poste? La poste va modifier ses services et ses tarifs pour devenir rentable Pour Reto Braun, concilier service public et rentabilité n’a rien d’impossible La poste déterminée à comprimer ses coûts La poste envisage de supprimer la moitié de ses succursales dans les villes Alors que des centaines de postes vont fermer, les habitants d’Echarlens n’ont pu sauver la leur « Il faudra Une révolte des usages pour stopper cette tendance »—pour Michel Béguelin, le rôle de la Confédération doit être rediscuté Erreurs de communication—Décidément, l’entrée des ex-PTT dans l’ère de la concurrence se déroule dans la douleur Après dix-sept mois seulement, Reto Braun abandonne La poste à sa difficile mutation Pour remplacer son directeur général, La Poste se met en quête de la perle rare Des bémols à la rationalisation des bureaux de poste—Le conseil d’administration impose à La Poste de mieux respecter la diversité culturelle et économique de la Suisse. La nomination du socialiste Ulrich Gygi à la tête de La Poste rassure les syndicats Moritz Leuenberger: « L’Etat doit garantir le service public, mais il n’est pas obligé de l’effectuer lui-même » Diversifier et réorganiser: le patron de La Poste suisse réaffirme son credo Réseau postal: 500 millions de pertes par an—Le directeur de La Poste avance quelques pistes sur la restructuration prévue du réseau postal. Les réorganisateurs de La Poste affrontent la colère des syndicats: Le comité de pilotage veut limiter à une centaine les grands centres postaux urbains. Les 350 prévus au départ auraient été trop chers La Poste va nommer son directeur sous haute surveillance politique Entre service public et e-commerce, les défis qui attendent Ulrich Gygi

Author Bernard Wuthrich Bernard Wuthrich Bernard Wuthrich Agnès Wuthrich Paul Coudret Emmanuelle Brossin Willy Boder Bernard Wuthrich

Bernard Wuthrich

D.S Miéville Bernard Wuthrich François Modoux

Bernard Wuthrich Bernard Wuthrich et Ludovic Rocchi Alexandra Deruaz Paul Coudret

Denis Masmejan

Sylvain Besson Bernard Wuthrich (continued)

Spatial Decision Support to Reorganize the Swiss Postal Network Date 13.05.2000 24.05.2000 19.01.2001 25.01.2001

22.03.2001

28.04.2001 28.04.2001

2.07.2001

15.08.2001 18.09.2001

5.10.2001

21.11.2001

16.01.2002 14.02.2002 20.04.2002 22.04.2002 26.10.2002

22.01.2004 7.07.2004

Title Les communes devront-elles payer pour conserver leur bureau de poste? Concilier rentabilité et service public, une mission impossible pour La Poste Une réorganisation justifiée. Par Bernard Wuthrich—Au temps de PTT, le réseau postal profitait du généreux Les socialistes romands montent aux barricades—les élus socialistes commencent par contester la nécessité de la réforme de La poste. La réforme des bureaux de poste sévèrement critiquée— Pris sous un feu nourri au national, Moritz Leuenberger défend le plan de restructuration de La poste. Avec la réforme de La poste, la ville de Zurich perdra plus d’un office Sur quatre Timide réforme postale. Par Stéphane Zindel—Ulrich Gygi en a vu de toutes les couleurs depuis qu’il a annoncé les grandes lignes de la restructuration du réseau postal à la mi-janvier Etendre le réseau postal?—Une motion de Jean-Claude Rennwald (soc./JU) Est soutenue par la moitié du Conseil national. Prétextant de bons résultats, La poste refuse de retarder sa réorganisation Journée cruciale pour La poste—Une pléthore d’interventions parlementaires provenant de Tous les partis exigent un ralentissement plus ou moins substantiel du plan de restructuration du réseau postal. La réforme du réseau postal ne sera pas interrompue—Le Conseil national a renoncé à imposer un moratoire à l’entreprise Le réseau postal genevois se décolle des quartiers—but de la réorganisation: Des offices plus visibles, plus proches des transports et équipements collectifs. La poste estime que six de ses bureaux en ville de Lausanne sont de trop Lausanne et Genève à la rescousse de leurs petites postes La restructuration de La poste n’a pas braqué les usagers «Il faut garder des bureaux de poste où il y a de la demande» Focus. L’homme de la semaine. Ulrich Gygi, manager «poste-moderne»—Il ne fait pas bon être CEO et socialiste par les temps qui courent. Le régulateur postal suisse manque d’indépendance Poste: la commission d’arbitrage en activité: La poste reçoit le feu vert pour transformer deux bureaux en «agences» en dépit de l’opposition des communes concernées

61 Author Willy Boder Emmanuelle Brossin Bernard Wuthrich Yelmarc Roulet

Bernard Wuthrich

Stéphane Zindel, Stéphane Zindel

Stéphane Zindel

Bernard Wuthrich Stéphane Zindel

Bernard Wuthrich

Laurent Busslinger

Laurent Busslinger Philippe Simon Stéphane Zindel Bernard Wuthrich Bernard Wuthrich

Matthias finger Stéphane Zindel

(continued)

62 Date 10.07.2004

1.11.2006

8.09.2004

23.09.2004 19.06.2006 11.08.2004 12.12.2006

F. Joerin Title Un débat cacophonique s’engage Sur les conséquences d’un oui à l’initiative postale—les avis divergent du tout au tout Sur la portée de l’initiative « services postaux pour Tous » qui sera soumise au peuple le 26 septembre. Le réseau des offices postaux se réorganise. La poste poursuit sa mue en biffant 500 emplois—La réforme du réseau déclenche l’ire des syndicats. Ils sont déjà descendus dans la rue. La diminution des files d’attente tombe à pic pour La poste—L’entreprise affiche son souci d’améliorer la qualité du service public. L’initiative « services postaux pour Tous » Est un leurre Ymago ou les idées de La poste pour transformer son réseau Une bataille s’engage autour des conséquences financières qu’entraînerait l’initiative postale «on ne sait pas où Ulrich Gygi veut Aller»—Jean-Noël Rey sort de sa réserve pour dénoncer le projet de restructuration du réseau postal. Il collaborera au projet de révision de la loi.

Author Stéphane Zindel

Stéphanie Germanier

Stéphanie Germanier

Bernard Wuthrich Stéphanie Germanier Stéphane Zindel Stéphanie Germanier

References Apparicio P, Gelb J, Dubé AS, Kingham S, Gauvin L, Robitaille É (2017) The approaches to measuring the potential spatial access to urban health services revisited: distance types and aggregation-error issues. Int J Health Geogr 16(1):1–24. https://doi.org/10.1186/s12942-0170105-9 Beaudry R, Collectif de recherche de Saint-Clément PQ, Dionne H (1998) In search of a local community: a village territorial mobilization, the saint-Clément postal conflict. Editions TroisPistoles, Québec Breux S (2008) Représentations territoriales et engagement public individuel: premières explorations. Polit Soc 27(3):187–210. https://doi.org/10.7202/029852ar Boyne G, Powell M (1991) Territorial justice: a review of theory and evidence. Polit Geogr Quart 10(3):263–281. https://doi.org/10.1016/0260-9827(91)90038-V Cailloux O, Meyer P, Mousseau V (2012) Eliciting ELECTRE TRI category limits for a group of decision-makers. Eur J Oper Res 223(1):133–140. https://doi.org/10.1016/j.ejor.2012.05.032 Charef K (2010) La représentation numérique à l'épreuve de la complexité du projet de territoire. Université de Montréal, Montréal. https://www.collectionscanada.gc.ca/obj/thesescanada/ vol2/002/NR69207.PDF?is_thesis=1&oclc_number=778188106. Accessed 2 January 2023 Commission fédérale de la poste PostCom (2022) Offices postaux et agences. https://www. postcom.admin.ch/fr/service-universel/offices-postaux-et-agences. Accessed 15 June 2022 Confédération Suisse–Autorité de régulation postal PostReg (2006) Rapport d’activité 2005. Berne. https://www.postcom.admin.ch/inhalte/PDF/Jahresberichte/PostReg_Taetigkeitsbericht_2005_ F.pdf. Accessed 15 June 2022 De Oliver M (1998) Geography, race, and class: a case study of the role of geography at an urban public university. Am J Educ 106(2):273–301

Spatial Decision Support to Reorganize the Swiss Postal Network

63

Dias L, Mousseau V, Figueira J, Climaco J (2002) An aggregation/disaggregation approach to obtain robust conclusions with ELECTRE TRI. Eur J Oper Res 138:332–348. https://doi.org/10. 1016/S0377-2217(01)00250-8 Espinasse B (1994) A cognitivist model for decision support: COGITA project, a problem formulation assistant. Decis Support Syst 12(4–5):277–286. https://doi.org/10.1016/0167-9236(94) 90046-9 Flourentzou F (2001) Piagetian constructivism in decision support. Contribution to sustainable development in architecture. Thesis N0 2418, Ecole Polytechnique Fédérale de Lausanne, Lausanne Geostat (1990) Federal population census. Federal statistical office, Neuchâtel https://www.bfs. admin.ch/bfs/en/home/services/geostat.html. Accessed 15 June 2022 Graillot D, Waaub J-P (2006) Aide à la décision pour l'aménagement du territoire: méthodes et outils. Hermes-Lavoisier, Paris Hansen WG (1959) How accessibility shapes land use. J Am Ins Plan 25:73–76. https://doi.org/10. 1080/01944365908978307 Joerin F (1998) Deciding on the territory, Proposal of an approach using GIS and multicriteria analysis methods. Thesis n. 1755, Thesis n. 2418 Ecole Polytechnique Fédérale de Lausanne, Lausanne Joerin F, Thriault M, Villveuve P, Bégin F (2001a) A multi-criteria procedure for assessing accessibility to places of activity. Int J Geomatics 11(1):69–104 Joerin F, Léanza Y, Rondier P (2010) Information, social representations and decision support: elements of reflection and illustrations. Proceedings of the Colloquium “Les outils pour décider ensemble”, 3rd edition, Montpellier Joerin F, Thriault M, Musy A (2001b) Using GIS and outranking multicriteria analysis for land-use suitability assessment. Int J Geogr Inf Sci 15:153–174. https://doi.org/10.1080/ 13658810051030487 Landry M (1995) A note on the concept of 'problem. Organ Stud 16(2):315–343. https://doi.org/10. 1177/017084069501600206 Meinard Y, Tsoukiàs A (2019) On the rationality of decision aiding processes. Eur J Oper Res 273(3):1074–1084. https://doi.org/10.1016/j.ejor.2018.09.009 Marsh MT, Schilling DA (1994) Equity measurement in facility location analysis: a review and framework. Eur J Oper Res 74(1):1–17. https://doi.org/10.1016/0377-2217(94)90200-3 Neutens T, Schwanen T, Witlox F, De Maeyer P (2010a) Evaluating the temporal organization of public service provision using space-time accessibility analysis. Urban Geogr 31(8):1039–1064. https://doi.org/10.2747/0272-3638.31.8.1039 Neutens T, Schwanen T, Witlox F, De Maeyer P (2010b) Equity of urban service delivery: a comparison of different accessibility measures. Environ Plan A 42(7):1613–1635. https://doi. org/10.1068/a4230 Postal Ordinance 783.01 (2003) Berne, Switzerland. Available at https://www.fedlex.admin.ch/eli/ cc/2012/586/fr Rigolon A (2016) A complex landscape of inequity in access to urban parks: a literature review. Landsc Urban Plan 153:160–169. https://doi.org/10.1016/j.landurbplan.2016.05.017 Roy B (1985) Méthodologie multicritère d'aide à la décision. Economica, Paris Roy B, Bouyssou D (1993) Aide multicritère à la décision: méthodes et cas. Economica, Paris Sharma G, Patil GR (2022) Spatial and social inequities for educational services accessibility-A case study for schools in Greater Mumbai. Cities 122, Article 103543. https://doi.org/10.1016/j. cities.2021.103543 Schuurman N, Berube M, Crooks VA (2010) Measuring potential spatial access to primary health care physicians using a modified gravity model. Canadian Geographer/Le Geographe Canadien 54(1):29–45. https://doi.org/10.1111/j.1541-0064.2009.00301.x Smoyer-Tomic KE, Hewko JN, Hodgson MJ (2004) Spatial accessibility and equity of playgrounds in Edmonton, Canada. Canadian Geographer/Le Geographe Canadien 48(3):287–302. https:// doi.org/10.1111/j.0008-3658.2004.00061.x

64

F. Joerin

Theriault M, Desrosiers F, Joerin F (2005) Modelling accessibility to urban services using fuzzy logic: a comparative analysis of two methods. J Prop Invest Finance 23(1):22–54. https://doi. org/10.1108/14635780510575085 Tagore MR, Sikdar PK (1995) A new accessibility measure accounting mobility parameters. Proceedings of the Seventh WCTR Conference, Sydney Australia, 1 Talen E, Anselin L (1998) Assessing spatial equity: an evaluation of measures of accessibility to public playgrounds. Environ Plan A 30(4):595–613. https://doi.org/10.1068/a30059 Tsou KW, Hung YT, Chang YL (2005) An accessibility-based integrated measure of relative spatial equity in urban public facilities. Cities 22(6):424–435. https://doi.org/10.1016/j.cities.2005. 07.004 Tsoukiàs A (2008) From decision theory to decision aiding methodology. Eur J Oper Res 187(1): 138–161. https://doi.org/10.1016/j.ejor.2007.02.039 Wachs M, Kumagi TG (1973) Physical accessibility as a social indicator. Socio-Econ Plan Sci 7(5): 437–456. https://doi.org/10.1016/0038-0121(73)90041-4 Wei C, Cabrera-Barona P, Blaschke T (2016) Local geographic variation of public services inequality: does the neighborhood scale matter? Int J Environ Res Public Health 13(10):981. https://doi.org/10.3390/ijerph13100981 Witten K, Exeter D, Field A (2003) The quality of urban environments: mapping variation in access to community resources. Urban Stud 40(1):161–117. https://doi.org/10.1080/00420980220080 Yu W (1992) ELECTRE TRI (aspects méthodologiques et manuel d'utilisation). Document n°74, LAMSADE, Université de Paris-Dauphine, Paris

A Decision Aiding Methodology to Compare Patient Classification Systems María A. de Vicente y Oliva

1 Introduction “The Spanish Constitution establishes the right to health protection and health care for all citizens. The main criteria that enable the exercise of this right are as follows: – – – –

Public financing, universality, and free health services at the time of use. Defined rights and duties for citizens and public authorities. Political decentralization of health care to the autonomous communities. Provision of comprehensive health care with high levels of quality, duly evaluated and monitored. – Integration of the different public structures and services at the service of health in the National Health System.”.

For a detailed explanation see “Sistema Nacional de Salud de España 2010” (http://www.msps.es/organizacion/sns/librosSNS.htm). Health care in Spain is a non-contributory benefit financed through taxes and is included in the general financing of each Autonomous Community. The competences of the Autonomous Communities in health matters are health planning, public health of their citizens, and management of health systems. The issue we are dealing with here is hospital management, which, as we have just mentioned, is under the jurisdiction of the Autonomous Communities in the Spanish case. A central point for management models is the quantification of the costs of products’ hospital (Ward et al. 2014). This task is complicated when hospital products/services are people, so that each person really is a final product. To solve this problem, so-called patient classification systems (PCS) are used to assign a cost

M. A. de Vicente y Oliva (✉) Department of Financial Economics and Accounting II, Faculty of Economics and Business Sciences, Rey Juan Carlos University, Madrid, Spain e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_3

65

66

M. A. de Vicente y Oliva

to each group of patients, thus reducing the number of end products to a more easily quantifiable number. Patients are grouped thanks to a PCS into the so-called Diagnostic Related Groups (DRGs), see 3M (2014) for more information. The original DRGs were designed for the US healthcare system and have evolved to adapt to different countries and their respective healthcare systems. Several countries in Europe have adopted DRGs (Busse et al. 2011; Schreyögg et al. 2006). In Spain, DGRs implementation became generalized in 1997 through a project of the Ministry of Health and Consumer Affairs to establish the average weights of the DRGs, although some autonomous regions had already been using it since the early 1990s (Chordá and Soler 2011). It is interesting to mention that in Spain, and therefore in all the Autonomous Communities, healthcare is public and that PCSs, specifically those developed by 3 M, were originally designed for private healthcare systems. International Classification of Diseases (ICD) coding systems generate various DRGs (World Health Organization 2012). Based on the ICD-9, the All Patients Diagnostic Related Groups (AP-DRGs) were used in Spain (Avrill et al. 2003) . In January 2016, Spain adopted ICD-10. One of the collateral damages of this transition was the AP-DRG PCS, which did not accept information coded with ICD-10. A new system had therefore to be adopted. A decision had to be made at a hospital level, in order to complete the change and determine which PCS is more appropriate as a for substitute the AP-DRG, because each health system has its own specificities, and the chosen patient classification system should try to reflect them. To this purpose, it is necessary to have a rigorous analysis framework to understand the advantages and disadvantages of any new PCS, both from the clinical point of view (different pathologies and patients in terms of case mix of the hospital) and from the point of view of the costs. This is the role of the DRGs, groups of patients who have associated pathologies with similar costs. To reach a decision regarding which PCS to adopt, an analysis was carried out at the Fuenlabrada Hospital (Community of Madrid), where two main candidates were considered: International Refined Diagnostic Related Groups (IR-DRG) and All Patients Refined Diagnostic Related Groups (APR-DRG). A working group, composed of members of the Fuenlabrada hospital and researchers from the Universidad Rey Juan Carlos, was activated in 2013 and completed in March 2015, thanks to funding from a research project (PI12/02039 “Evaluation of the APR_GRD and IR-GRD classification systems in the measurement of the case-mix of a hospital,” Funding entity: FONDO DE INVESTIGACION SANITARIA). It is important to note that this study was not commissioned by the hospital’s management but was initially a work of reflection conceived by the hospital’s experts and a team of analysts and experts in multicriteria decisionmaking methods. The chapter is organized as follows. Section 2 presents a brief description of the problem and the organizational context. Section 3 focuses on the problem formulation and the model structuring. Section 4 deals with the implementation of an MC method and its results. Section 5 is devoted to some concluding remarks.

A Decision Aiding Methodology to Compare Patient Classification Systems

67

2 The Problem and the Organizational Context The working group of the Fuenlabrada hospital proposed to carry out a study of the consequences on hospital management of the implementation of a new PCS and thus to present to the hospital’s management a recommendation regarding the PCS that is best suited to the specific casuistry of patients at Fuenlabrada hospital. The team who carried out the analysis consisted of experts in charge of patient management (doctors of the Fuenlabrada Hospital), experts from the hospital economic management department and experts in Multi Criteria Decision Aiding (MCDA) from the Rey Juan Carlos University (the Fuenlabrada Hospital is affiliated with this University). The main objective of the decision aid intervention was to facilitate a process of reflection on what would be the best choice of a new PCS for the specific problematic of the hospital, and to submit the conclusions to the hospital direction in the hope that they would be taken into account. A change in the encoding of the information on which PCSs are based generates a technical problem, on what would be the most appropriate PCS once the current system had become obsolete because it did not accept information based on the new International Classification of Diseases ICD-10 (Quan et al. 2005). It is easy to realize that other changes, either to the ICDs or to the PCSs, will have to take place in the future. By their nature, ICDs and PCSs must evolve by adapting to changes in diseases (just take COVID-19 pandemic as an example) and in population nature and distribution. In fact, DRGs have evolved over time by adapting to medical needs and considering more and more relevant factors for grouping patients. Table 1 lists some of the most relevant DRGs. Therefore, a system designed for this occasion could be a Decision System that would last over time. This second objective of the decision aid intervention, almost as important as the first, was to make the hospital team of experts and the direction of the hospital aware of the need to implement a Decision System that could help them make decisions whenever a new change of ICDs or PCSs or both occurred. This would allow them to intervene with justified arguments in the decision-making processes of the Community of Madrid. If we focus on the main objective of the intervention, “the AP-DRG, based on the ICD-9, had to be replaced by a new PCS belonging to the ICD-10 coding system” with two possible alternatives whose performance evaluations were agreed upon by all those involved in the analysis. However, if we go to the secondary objective, the implementation of a decision support system for the comparison of new possible alternatives for any change in the PCSs or ICDs, then the organizational context and the problem may be described as evolutionary. The direction of the hospital changes from time to time and the medical experts of the hospital also change. The economic management experts of the hospital—because of the type of contractual relationship with the hospital—are perhaps more stable. And of course, alternatives and performance evaluations evolve in relation to the different organization contexts and

68

M. A. de Vicente y Oliva

Table 1 Evolution of main DRGs DRG system Medicare DRG

All patient DRGs (AP-DRG) All patient refined DRGs (APR-DRG)

International refined DRGs (IR-DRG)

Explanation and purpose DRGs are used by Medicare (Vladeck 1984) and measure the typical resource use of an inpatient stay. DRGs also include complications and comorbidities (CC). In the USA, Medicare is a national social insurance program, administered by the US federal government since 1966. AP-DRGs are like DRGs, but also include a more detailed DRG breakdown for non-Medicare patients, particularly newborns and children. The APR-DRG incorporates severity of illness subclasses into the AP-DRGs. The APR-DRGs expand the basic DRG structure by adding four subclasses to each DRG. The addition of the four subclasses addresses patient differences relating to the severity of illness and risk of mortality. The 3 M APR-DRGs (Shen 2003) have been updated annually to include ICD-9 code changes. 3 M has also updated them for ICD-10. IR-DRGs classify patients across the continuum of care, including inpatient, outpatient, clinics, and rehabilitation. The IR-DRG system uses the same logic and structure as the AP-DRG and APR-DRG systems, but if AP-DRGs and APR-DRGs are based on diagnostics IR-DRGs are based on procedures. They incorporate the same severity of illness adjustment using secondary diagnoses. The IR-DRG does not include multiple complications and comorbidities because 3 M identified most international datasets do not contain more than two secondary diagnoses. IR-DRGs are procedure-driven, IR-DRG also conforms to multiple versions of ICD: ICD-10, ICD-9-CM, and ICD-9.

Source: Oliva et al. (2018). A decision aiding methodology to compare patient classification systems. International Journal of Multicriteria Decision Making, 7(3–4), 177–194

problem formulations. The source of data, both for the main and secondary objectives, must always be the hospital itself. The decision aid intervention in relation to the main objective was developed with a clear connection with the secondary: the definition of the alternatives became a complete and clear example for any future decision problem and the modelling of the performance evaluation was elaborated in strict connection with the experts and documented as a guideline for future modelling activities. Transparency adopted in the modelling and method application, in relation to the main objective and technical problem, was considered a precondition for the development of a decision system.

2.1

Relationships and Communications

The relationships of the MCDA analysts with the experts of the working group were direct and fluid, although not very frequent and fluent with the hospital direction. A shared interest in research facilitated communication between analysts and experts and the tools used by the analysts in the decision aid intervention were viewed with interest and open-mindedness by the experts.

A Decision Aiding Methodology to Compare Patient Classification Systems

69

Fig. 1 Structure of DA interventions

However, the relationship between analysts and experts on the one hand and the hospital direction on the other was difficult, above all because the points of view were different. The direction of a hospital is political in nature and seeks short-term results in line with the ideology of the ruling party in the Community. The hospital employees, in charge of patient management and economic management, have a long-term vision based on medical knowledge and experience. In the case study, the Community of Madrid was immersed in a reorganization of the health care system based on a public–private system that aimed to privatize the economic management of hospitals. Some new hospitals, such as the Fuenlabrada hospital, were chosen as flagships of this system. The hospitals had to implement changes quickly (so that the results could be seen in the current legislature) and this did not favor in-depth discussions about technical decisions (the choice of the most appropriate PCS for the Fuenlabrada Hospital, in this case). The aim of the working group was to provide the hospital direction with a robust outcome of the research project so that both medical and economic interests of the hospital would be considered, in the current decision and in the future. However, the hospital direction was busy with a rapid organizational change of the health care system, oriented toward privatization of public medicine. Therefore, the main relationship of the hospital direction was with the PCS suppliers and not with the working group. Two different processes were developed, in relation to the technological change. The first decision process involved the Community of Madrid, the hospital direction and the technology suppliers. The other process, which was developed in the working group, can be called a reflection process (see Fig. 1). Communication between hospital experts and analysts was very satisfactory. They collaborated in the reflection process and could work together in the future (see Fig. 2).

3 Problem Formulation and Modelling Process The problem formulation and the structuring of a model were elaborated jointly by experts and analysts of the working group, with the hospital management, caregivers, and specialists. A multicriteria evaluation model was structured, to propose a ranking of the proposed alternatives, or of different alternatives in the future. The data were generated by the economic management department, responsible for using PCS and e elaborating the analytical accounting of the hospital.

70

M. A. de Vicente y Oliva

Fig. 2 Components of the different parts of the DA interventions

Two candidates had been considered for the new PCS, here called APRs (All Patients Refined Diagnostic Related Groups) and IRs (International Refined Diagnostic Related Groups), by the hospital direction. The first step in model structuring was a transparent definition of a set of alternative scenarios for the PCS use. These alternatives corresponded to the different combinations of the APRs and IRs with the possibility of considering or not the degree of severity and choosing between the fare paid by the hospital (UCH: hospital unit complexity fare) or a new fare called “public fare.” Eight alternative scenarios were therefore generated, four in relation to the technology APRs and four for IRs (see Table 2). The set of alternatives was created by the hospital’s group of experts. A family of criteria is required in multicriteria analysis (Roy 1993, 1996) to evaluate different scenarios. The process of criteria elaboration was conducted jointly with the experts from the Fuenlabrada hospital. First, hospital managers were consulted. They provided a list of control variables used by the department

A Decision Aiding Methodology to Compare Patient Classification Systems Table 2 Set of alternative scenarios of PCS use

E1 E2 E3 E4 E5 E6 E7 E8

71

APRs without severity UCH APRs without severity public fare APRs with severity UCH fare APRs with severity public fare IRs without severity UCH fare IRs without severity public fare IRs with severity UCH fare IRs with severity public fare

Source: Oliva et al. (2018). A decision aiding methodology to compare patient classification systems. International Journal of Multicriteria Decision Making, 7(3–4), 177–194

of economic management of the hospital. This list was discussed with the hospital managers and several key caregivers (mainly physicians) to build and validate the family of criteria. One of the properties of a coherent family of criteria is non-redundancy and a discussion with experts from the hospital management department was above all necessary to detect and eliminate possible redundancies. Three aspects of hospital management were considered to analyze the required funding to and ensure the non-existence of imbalances regarding the groups funding. Equity in the use of resources needed also to be considered. The distinction was reflected in the composition of the family of criteria: • The APRs or IRs groups (criteria C1, C2, and C3). • The patient as the final subject of the hospitalization (criteria C4 and C5). • The hospital management at a general level (criteria C6, C7, and C8). Here it is important to make an observation to better understand the model. When looking for balances and imbalances, it is necessary to do so both in the groups and in the number of discharges. The groups refer to identified pathology and cost profiles. In this way, it is necessary to ensure that the groups are in balance in terms of their financing. It is of little use if overall funding is good if, for example, psychiatry is underfunded and pediatrics is overfunded. The financing of pathologies must be correct; we must not find pathologies that are underfinanced and others that are overfinanced. But at the same time, the discharges of patients, who will belong to all the groups, must also be correctly financed. In this case we are dealing with patients at a general level, without taking into account pathologies (Table 3). Data used for the analysis corresponded to the 2009–2013 period. Patients who were grouped in AP-DRGs in that period were converted and grouped in APRs or IRs, in relation with the eight scenarios. Given the nature of the problem under study, it was proposed to the experts to use a non-compensatory multicriteria method to avoid choosing a PCS that performed excessively poorly on any of the retained criteria. Since the final objective was to carry out a reflection process to choose the PCS, it was considered that a ranking of the alternatives as the final solution would be the most appropriate. The method chosen was ELECTRE III. To know more about the use of this outranking method in this DA intervention see (Figueira et al. 2016).

72

M. A. de Vicente y Oliva

Table 3 Set of criteria

C1 C2 C3 C4 C5 C6 C7 C8

Name of the criterion Underfunded groups Groups in equilibrium Overfunded groups Underfunded discharges Discharges in equilibrium Underfunded cost Cost in equilibrium Balance funding/cost (difference between funding and cost)

Measurement unit [preference direction] Number of groups [min] Number of groups [max] Number of groups [max] Percentage [min] Percentage [max] Percentage [min] Percentage [max] € [max]

Source: Oliva et al. (2018). A decision aiding methodology to compare patient classification systems. International Journal of Multicriteria Decision Making, 7(3–4), 177–194

In relation to the chosen outranking method, some parameters had to be set with the experts. These parameters were the weights and veto thresholds, the inter-criteria information to implement the concordance and non-discordance principles of the outranking relation, and the indifference and preference thresholds, to reduce the negative effect of information and preference uncertainties on the result. All criteria were quantitative. The source of data being the same for all the criteria, experts considered to express the indifference and preference thresholds as a percentage of the difference between the maximum and the minimum of the performances of the eight alternative scenarios. It should be mentioned at this point that the experts of the hospital’s economic management department had considerable reluctance in setting the indifference and preference thresholds. This is curious because the data for the criteria were all derived from the hospital’s accounts. In the end they opted for percentages as a less risky or more aseptic choice. For indifference and preference thresholds a consensus was reached: indifference threshold would be “5% of the difference between the maximum and the minimum performance on the criterion,” while preference threshold would be “20% of the difference between the maximum and the minimum performance on the criterion.” It is important to note at this point that the thresholds are intended here to reflect the most pessimistic situation. These patient classification systems were initially designed for the United States where healthcare is private. That is to say, a very different healthcare model from the Spanish public healthcare model. The difference between public and private health systems is something that will influence all patient classification systems that have not been specifically adapted to a specific health system. The experts could not reach a consensus about the criteria weights; in fact, they wanted to know how different degrees of relative importance of the criteria (the meaning of weight in ELECTREIII) could affect the solution. The experts also expressed the need to analyze results with and without veto thresholds. The intention was to study possible situations and their consequences, to present clear and robust conclusions to the direction of the hospital.

A Decision Aiding Methodology to Compare Patient Classification Systems

73

The difficult definition of weights and veto, which are decision parameters in ELECTRE III, seemed consistent with the expert roles, internal experts in a working group that should facilitate reflection and not decision-makers.

4 The Implementation of the ELECTRE III Method: Analysis and Validation of the Results Despite different solutions being obtained with different sets of parameter values, E5, a special type of IR (IRs without severity, UCH fare), stood out as the best PCS option (see Table 4). The sensitivity analysis could produce some clarification regarding the multicriteria model structure, where the two main aspects were related to the condition of equilibrium and the other of non—equilibrium. In 2016, the ICD-10 was introduced, and the hospital’s decision was to change to APR-DRG. The hospital’s decision was based on the recommendations of the company that owns the software (3 M). APR-DRG was also the PCS that was being widely implemented in hospitals around the world using this technology. APR-DRG is a PCS valid for all populations, from newborns to the elderly, and at all levels of severity and risk. As 3 M remarks: “The 3 M APR DRG methodology classifies hospital inpatients according to their reason for admission, severity of illness and risk of mortality” (see: https://www.3m.com/3M/en_US/healthinformation-systems-us/drive-value-based-care/patient-classification-methodolo gies/apr-drgs/). Our study led to recommend an IR-DRG system. Experts from 3 M (the owner of the PCS software) explained that AP-DRGs (the old PCS used at the Fuenlabrada Hospital) were closer to IR-DRGs than to APR-DRGs. APR-DRGs were more inherently linked to clinical aspects whereas cost aspects are more important to AP-DRGs and IR-DRGs. The hypothesis put forward by the experts of the working group was de facto that the PCS used in the hospital was well suited to the patient case-mix they had from a financial point of view and from a clinical point of view, or at least well enough (Fetter et al. 1980). The change of PCS was only due to the change from ICD-9 to ICD-10. With these premises, it was logical that the result of the reflection process was the choice of an IR-DRG, since it was the one that most closely reproduced the grouping of the AP-GDRs. APR-GDR has a different approach to the AP-GDR and the IR-GDR and consequently moves further away from the patient grouping the hospital was working with. In other words, only economic criteria were taken into account in the decisionmaking process. Perhaps, if some clinical criteria had been included, the solution would have been different. The decision-makers at the hospital based their decision on the trend that was emerging in other countries using these PCS and on 3 M’s recommendations. This is a priori logical, but it must also be considered that the Spanish case (and specifically

74

M. A. de Vicente y Oliva

Table 4 Situations and parameter values considered for the sensitivity analysis and their respective results Criteria weights Veto threshold: 50% of the difference between the maximum and the minimum performance on the criterion Criteria weights: All equal to 1 but 0.5 for C8

Results

Criteria weights No veto threshold Criteria weights: All equal to 1 but 0.5 for C8

Criteria weights: Equilibrium weights twice the non-equilibrium Weights. 0.5 for C8

Criteria weights: Equilibrium weights twice the non-equilibrium weights. 0.5 for C8

Criteria weights: Non-equilibrium weights twice the equilibrium weights. 0.5 for C8

Criteria weights: Non-equilibrium weights twice the equilibrium weights. 0.5 for C8

Results

Source: Oliva et al. (2018). A decision aiding methodology to compare patient classification systems. International Journal of Multicriteria Decision Making, 7(3–4), 177–194

A Decision Aiding Methodology to Compare Patient Classification Systems

75

the case in Madrid) was not necessarily the same as in other countries. Once again, the great difference between public and private health systems must be considered.

4.1

Main Difficulties Encountered During the DA Intervention

Three major difficulties were encountered during the DA intervention, the first around the relationship of the working group and the hospital direction, the others internal to the working group. The main difficulty was the unwillingness of the hospital direction (the decisionmaker) to decide on the PCS jointly with the expert and with the analysts of the working group and, consequently, to use the outcome of the reflection process. to interact with the experts and analysts of the working group and, consequently, to use the outcome of the reflection process. The decision of the hospital direction did not take the reflection process into account. The decision was made in advance (this means that the decision made by the hospital before the decision process was finalized) based primarily on medical clinical considerations, while the PCS have been designed to form patient groups from a medical and economic point of view. Although it is wholly recognized that the decision must be adapted to the specific case study problem context, while in this case the Fuenlabrada hospital followed the supplier recommendations and general decisions, in Spain and elsewhere. The process of reflection on the PCS change was very fruitful in terms of interaction between experts and analysts, during the intervention and for the future. Nonetheless, some procedural difficulties have to be underlined. The definition of subjective information, to apply the ELECTRE III method to the model (thresholds, weights, and vetoes) was difficult for experts with economic and medical expertise. This is a common difficulty in decision processes and, above all, in interventions with experts and methodological experts who are not decision-makers but have to propose their visions to the decision-makers. The parameters of the model help to define each decision problem as a unique case (we would not have the same problem in one hospital as in another, even if they were in the same city, nor would we even have the same decision problem for the same hospital at two points in time). The establishment of such parameters assumes a thorough knowledge of the problem, and this is not always the case. Therefore, it is important to carry out a sensitivity analysis to study and analyze the variations in the final recommendation produced by variations in weights, thresholds and vetoes. In our case, as it is essentially a process of reflection, the difficulty in setting values for the model parameters was not an added difficulty. In fact, the aim was rather to see how variations in weights, thresholds, and vetoes affected the final recommendation.

76

M. A. de Vicente y Oliva

The “understanding and analysis” of the results by the experts was another difficulty and took some time. But the result was favorable and led the experts to a better understanding of the consequences of choosing one PCS or another. Unfortunately, there was no willingness on the part of the hospital direction to share this knowledge. Just as important as finding a recommendation at the end of a decision process is knowing how to convey the meaning and scope of the recommendation. The first step must be for the experts to understand it, bearing in mind that before the recommendation is proposed there has been a sensitivity analysis and that different variations of the problem’s parameters will have led to different decision-making proposals. Once the experts have understood and analyzed the possible solutions that have led to the final recommendation, it is time to transfer the recommendation to the decision-maker. In our case, when the time came to pass on the recommendation to the decisionmaker, we found that the hospital direction was not interested in our research. The decision was based not on the result of a process of research and its reflections, but on the trends imposed in the healthcare world using the patient classification systems developed and marketed by 3 M. In short, they would follow 3 M’s recommendations.

5 Concluding Remarks The ex post analysis of this intervention produced some reflections about the behavior of the decision-maker and analyst. What lies at the heart of case is a mistrust of the research team and of the research process itself. This is because a decision-maker usually finds it difficult to engage with and understand a research process. In decision aid processes based on methodologies such as the one applied here, an outranking method, the decision-maker’s feeling is that he or she must bring information to the process that is not easy to clarify. There is therefore some insecurity that the outcome of the analysis, based on their decision-maker’s preferences and information, will not be correct. In many situations, if the recommendation derived from the analysis does not coincide with “the general trend” or “the one imposed by higher authorities,” the decision-maker tends to reject the recommendation generated by the analysis mainly because he/she fears that he/she has not provided sufficiently clear or accurate information., that at some point of the decision process, someone (the analyst or the experts) had not included or used correct information. An easy introduction of new scenarios and criteria or definition of specific alternatives and associated criteria (perhaps not strictly economic criteria) could help the hospital make decisions whenever a new change of ICDs or PCSs or both occurred. Information should be provided by both medical and financial management experts of the hospital. Problem formulation and model structuring should be proposed and agreed upon by the hospital’s experts. And the information regarding

A Decision Aiding Methodology to Compare Patient Classification Systems

77

the assessment of alternatives for the criteria would come from the hospital internal data. Without going to the extreme conclusion of this case, it is common for the decision-maker to take into consideration not only the so-called “technical” aspects, but also other considerations that cannot or must not be included in the model but are in the decision-maker’s mind. Consequently, when the proposed final recommendation clashes with any of these decision-maker’s considerations it is difficult to obtain a clear acceptance of the proposed recommendation. From the point of view of the analyst, a good collaboration between the members of a working group is an essential condition in MCDA interventions. Any involvement of decision-maker(s) is very important and often this is a critical requirement. Communication has to be activated and managed during all the intervention steps, without any certainty of successful involvement. The organizational context can evolve, and a clear and complete documentation of the methodological approach and its results is essential. A reconsideration of the alternatives and/or the criteria should always be possible. Sometimes, problem formulation needs to be changed.

References 3M (2014) GRD Y CRGS. Available at: http://solutions.productos3m.es/wps/portal/3M/es_ES/ Healthcare-Europe/EU-Home/Products/HealthInformationSystems/ProductosYServicios/ SoftwareDeAgrupacion. Accessed 27 December 2022 Avrill R, Goldfield N, Hughes JS, Bonazelli J, McCullough EC, Steinbeck BA (2003) All patient refined diagnosis related groups (APR-DRGs) version 20.0: methodology overview. 3M Health Information Systems, Wallingford Busse R, Geissler A, Quentin W, Wiley M (eds) (2011) Diagnosis-related groups in Europe. Moving towards transparency, efficiency and quality in hospitals. Open University Press Chordá VMG, Soler MLM (2011) Grupos de pacientes Relacionados por el Diagnóstico (GRD) en los hospitales generales españoles: variabilidad en la estancia media y el coste medio por proceso (Diagnosis-Related Patient Groups (DRG) in Spanish general hospitals: variability in average length of stay and average cost per process). Enferm Glob 10(24). https://doi.org/10. 4321/S1695-61412011000400011 Fetter RB, Shin Y, Freeman JL, Averill RF, Thompson JD (1980) Case mix definition by diagnosisrelated groups. Med Care 18:1–53 Figueira JR, Mousseau V, Roy B (2016) ELECTRE methods. In: Greco S, Figueira J, Ehrgott M (eds) Multiple criteria decision analysis. Springer, New York, pp 55–185 Oliva MADVY, Amor SB, Bassa JM, Fando ICL, Delgado ÁG, Bohoyo PDM (2018) A decision aiding methodology to compare patient classification systems. Int J Multicriteria Decis Mak 7(3–4):177–194. https://doi.org/10.1504/IJMCDM.2018.094383 Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA (2005) Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 43(11):1130–1139. https://doi.org/10.1097/01.mlr.0000182534. 19832.83 Roy B (1993) Aide Multicritère à la Décision: Méthodes et Cas. Economica, Paris Roy B (1996) Multicriteria methodology for decision aiding. Springer Science & Business Media, Berlin

78

M. A. de Vicente y Oliva

Schreyögg J, Stargardt T, Tiemann O, Busse R (2006) Methods to determine reimbursement rates for diagnosis related groups (DRG): a comparison of nine European countries. Health Care Manag Sci 9(3):215–223. https://doi.org/10.1007/s10729-006-9040-1 Shen Y (2003) Applying the 3M all patient refined diagnosis related groups grouper to measure inpatient severity in the VA. Med Care 41(6):II103–II110. https://doi.org/10.1097/01.MLR. 0000068423.39715.CE Sistema Nacional de Salud de España (2010) Madrid. Ministerio de Sanidad y Política Social (Spanish National Health System 2010. Madrid. Ministry of Health and Social Policy.). Instituto de Información Sanitaria. Available at: http://www.msps.es/organizacion/sns/librosSNS.htm. Accessed 27 December 2022 Vladeck BC (1984) Medicare hospital payment by diagnosis-related groups. Ann Intern Med 100(4):576–591. https://doi.org/10.7326/0003-4819-100-4-576 Ward MJ, Marsolo KA, Froehle CM (2014) Applications of business analytics in healthcare. Bus Horiz 57(5):571–582. https://doi.org/10.1016/j.bushor.2014.06.003 World Health Organization (2012) International classification of diseases (ICD). Available at: https://www.who.int/standards/classifications/classification-of-diseases. Accessed 27 December 2022

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing Antonio Fiordaliso, Olivier Pilate, and Marc Pirlot

1 Introduction This is the story of an intervention in the development process of a decision-aiding tool for the Public Service of Wallonia - Mobility & Infrastructure, formerly called the Ministry of Equipment and Transport (MET). This administration is responsible for the management of the road network in Wallonia, the French-speaking, Southern part of Belgium, a region covering 55% of the Belgian territory. The tool was intended to help the engineers managing road works in Wallonia to make the best possible choice of road pavement and surfacing. We first outline the historical and institutional context of the intervention. Then we describe how the model underlying the evaluation of the different possible road pavements and surfacings and their adequacy to different types of road works (RW) was elaborated and validated. This model was implemented as a software tool aiming to facilitate decision and proposed to the end users (i.e., the field engineers responsible for road works). We discuss all along the issues raised in the course of the intervention. Finally, we look back on the entire process and highlight some difficulties encountered in the intervention and some issues that remain open.

A. Fiordaliso Service Public Fédéral Economie, Statbel-Statistics, Brussels, Belgium O. Pilate · M. Pirlot (✉) Faculté Polytechnique, Université de Mons, Mons, Belgium e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_4

79

80

1.1

A. Fiordaliso et al.

Context of the Intervention

A large variety of road pavements and surfacing (RPS) are in use in Wallonia and the choice of an RPS for each road works (RW) was made by the engineer in charge of managing it. The project of building a software tool for improving the choice of an appropriate RPS in RW in Wallonia was launched at the initiative of the general manager of the MET in 2003. A working group was set up • Headed by the MET general manager (GM). • Assisted by two economic attachés (in charge of writing calls for tenders for public procurement). • Including four Regional Managers (RM) and four Engineers Directors (ED) responsible for the supervision of RW in Wallonia’s sub-regions. • An engineer from the Belgian Road Research Center (BRRC) who is a co-author of this chapter (OP), joined at a certain point in the process, by the two other authors (AF, MP), who are experts in decision-aiding methods. The task of this working group was to elaborate a decision-aiding tool to be used by the Field Engineers (FE) responsible for RW in Wallonia in order to assist them in choosing the most appropriate RPS.

1.2

The Working Group

The GM led all meetings of the working group. Not all 14 members of the group attended each group meeting. The GM, the four RMs and the four EDs had an extended practical experience of RW and a good knowledge of the properties of road pavements and surfacing. They had the expertise to assess the adequacy of an RPS to an RW. They were also well aware of the concrete problems met by the Field Engineers (FE) managing RW. Each FE operates in a sub-region of Wallonia and hierarchically answers to the RM or the ED supervising this sub-region. The working group occasionally called upon external expertise, namely that of a member of the Belgian Road Research Center and that of two experts in concrete road pavements from FEBELCEM (Federation of the Belgian Cement Industry). It is important to note that the four RMs and the four EDs hierarchically answer to the GM. The Field Engineers in turn answer to an RM or an ED and, ultimately, to the GM. The Belgian Road Research Center (BRRC) is a research center associated with the road construction sector in Belgium (both the road construction industry and the public sector in charge of road management and planning). This research center develops and disseminates its expertise on all aspects of road construction. For instance, in relation with the present case, the BBRC had developed tools for dimensioning roads, i.e., what is the required thickness of the road pavement for a given intensity of the traffic). One co-author, OP, a construction engineer, was

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

81

working for the BRRC. At the same time, OP was pursuing a degree-granting program (with staggered hours) at the Mons Faculty of Engineering, Belgium. In this program, he attended a course on multiple criteria decision methods and proposed to write his master thesis on the topic of the choice of RPS aided by multi-criteria analysis. He managed to generate interest for this approach from the direction of the BRRC and from the direction of the MET. He thus joined the working group. In his master thesis, OP reports on the activity of the working group, up to the development of a decision model based on an additive value function and the validation of this model on a dozen of cases corresponding to real RW in the Walloon region. Initially, the authors AF and MP acted remotely as OP’s master thesis advisors. When it came to the model’s development phase, they integrated the working group and took an active part in the interactions within the group. After the thesis was defended, a contract was signed between the Mons Faculty of Engineering and the MET in order to implement the model into a software tool, with a user-friendly interface. During this phase, the model was further refined and amended. The intention of the MET management was to ask the FEs managing road works to use this software when deciding which RPS should be chosen. The software interface was designed to enable the user to compare the pros and cons of the top-ranked RPSs.

1.3

Goal of the Management of the MET

The initial goal of the GM with this working group was to determine the best option for the choice of an RPS in each of the various RW contexts that may appear in the Walloon Region and to impose strong guidelines regarding this choice to the FEs managing RW (who are supervised by the RMs or EDs). The authors argued in favor of a less prescriptive perspective. Our main reasons were twofold. First, considering the future software tool as a decision aid rather than as a norm would facilitate acceptance by the end-users. Second, no model is perfect. So there might be cases in which the solution recommended by the model might not be the best option, for reasons not taken into account in the model, but of which the RW manager might be aware. It was finally accepted that the FEs should use the system in all cases but could derogate from the system-recommended solution, in which case they should justify their choice explicitly to their superiors.

82

A. Fiordaliso et al.

2 Highlights of the Modeling Phase Initially, only five different types of RPSs were considered. Later, when it came to the evaluation of these five alternatives with respect to the various criteria, it appeared that the evaluation of these alternatives on some of the criteria substantially depended on the exact specification of the upper layer (“wearing course”) of the road surfacing. It was the case, e.g., for the “noise” criterion. The road noise level may be much higher with some wearing courses than with others. It was thus decided to consider a more detailed list of alternatives composed of 22 combinations of the five basic types of RPS with different wearing courses that can be used with each of them (Appendix A). Initially, 11 items were retained as having an impact on the choice of an RPS. These are: (1) heavy truck traffic; (2) disturbances due to works; (3) discomfort to users due to maintenance; (4) ride comfort; (5) safety; (6) insertion of inlets (sewer openings, gulleys, inlet cover, manhole cover, etc.); (7) current pavement type; (8) RPS of adjacent road sections; (9) importance of the RW; (10) cost relative to life cycle; and (11) type of road. It was soon realized that these items cannot all, as such, be used as criteria. For some of them, their impact heavily depends on the type of road and the location of the works. For instance, for works in a small street in town, considering “heavy truck traffic” is irrelevant because inexistent, while it has a strong impact on a highway, because it causes rutting and cracking, hence reducing the service life of an RPS. Similarly, the impact of bad or medium performance with respect to “safety” (item 5) or “ride comfort” (item 4) is different on a highway and a local road or a parking lot. So, the best choice for an RPS depends on the type of road and the location of the RW. A categorization of RW types was established by the working group (see Sect. 2.2). The preference model had thus to be adapted to the type of works. The option taken was to develop a model suited to the most important category of works and adjust its parameters to reflect the different impact of some criteria in other categories of works. As for the criteria, they remain the same in all types of RW.

2.1

The Criteria

The list of criteria finally retained is the following. It is composed of five quantitative and seven qualitative criteria. The evaluation of quantitative criteria has an objective character; it results from real data or physical simulation measures or from using computer simulation programs (as those developed by the BRRC) based on real data. The quantitative criteria are: 1. Cost (€/m2): estimation based on document (MET 2002) and discussions in working group; depends on works characteristics (intensity of heavy truck traffic); to be minimized.

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

83

2. Rutting (orniérage): measured (mm) through laboratory simulation in standardized conditions; to be minimized. 3. Skid resistance (résistance au dérapage): measure of the tangential friction coefficient (dimensionless) by a measurement device called odoliographe (CRR 2019); to be maximized. 4. Cracking rate (taux de fissuration): computed by an external “dimensioning module” (depends on the traffic load); dimensionless; to be minimized. 5. Noise generated by the vehicles running on the RPS: measured (dB) in standardized conditions in the lab; to be minimized. The qualitative criteria are the following: 1. 2. 3. 4. 5. 6. 7.

Drainability. Perturbation due to the works: depends on the work location. Ease of inlets insertion (sewer openings, gulleys, inlet cover, manhole cover, etc.) Suitability given the current pavement of adjacent zones. Suitability given the work location (independently of the current pavement). Maintenance requirements. Ease of implementation (given the current pavement).

The GM and three persons among the RMs and EDs in the group assessed these aspects for the different RPSs. The choice of the criteria scales is discussed in Sect. 3.1. We refer in the sequel to quantitative (resp. qualitative) criteria by their number preceded by QUANT (resp. QUAL): QUANT1 to QUANT5 (resp. QUAL1 to QUAL7).

2.2

Categorization of Road Works

Two types of attributes characterize RW. The first is the type of road. These have been categorized into the following nine types, referred to as RW1 to RW9 in the sequel. 1. 2. 3. 4. 5. 6. 7. 8. 9.

Crossing. Agricultural access road (chemin de remembrement). Parking lot. RESI: roads connecting cities and villages (allowed speed 3” in the rightmost column in Table 1. Discussion revealed for instance that some experts have included some aspects belonging to other criteria into their evaluation. Since discussing all such

A. Fiordaliso et al.

88

Table 1 RPS evaluations on criterion QUAL1 (drainability) by four different experts. “Max Diff” stands for “Maximal difference”

large discrepancies would take time, it was provisionally decided to proceed with the current evaluations and work with their average (column labeled “Mean” in Table 1). Table 2 shows the average evaluations of all WPS w.r.t. the 29 combinations (qualitative criterion x RW category and characteristics) described above. Depending on the RW category and the other RW parameters, the related evaluations for the seven qualitative criteria will be used in an additive value function model. For example, in the case of the works of category RW5 on the E411 (described at the end of Sect. 2.2), for assessing the first RPS, which is RBIT/BB-1, we extract the relevant evaluations of the qualitative criteria from Table 2; they are shown in Table 3 and justified by the characteristics of this RW.

3.2

Quantitative Criteria

We first give some detail about the (objective) evaluation of the five quantitative criteria. Some of them depend on the RW category. QUANT1: Cost The construction cost of an RPS chiefly depends on its thickness. The latter is mainly determined by the traffic of commercial vehicles (heavier than

Perturbation / RW2

7.0

7.0

7.0

7.0

7.0

7.0

7.0

7.0

6.8

6.8

6.8

6.8

6.8

6.8

3.1

Drainability

2.5

4.8

3.8

5.8

5.3

8.0

5.3

2.8

4.8

3.8

5.3

5.3

8.0

4.5

4.7

Alternatives

RBIT / BB-1

RBIT / RMD

RBIT / SMA

RBIT / RMTO

RBIT / RUMG

RBIT / ED

RBIT / Enduit superficiel

RBIT / RBCF

REME / RMD

REME / SMA

REME / RMTO

REME / RUMG

REME / ED

REME / Enduit superficiel

BCOM / RMD

4.0

6.9

6.9

6.9

6.9

6.9

6.9

7.1

7.1

7.1

7.1

7.1

7.1

7.1

7.1

Perturbation / RW3

2.4

5.9

5.9

5.9

5.9

5.9

5.9

6.1

6.4

6.4

6.4

6.4

6.4

6.4

6.4

Perturbation / RW4

2.6

6.0

6.1

6.1

6.1

6.1

1.9

5.9

5.6

5.6

5.6

5.6

5.6

6.0

6.0 6.1

6.0

5.8

5.8

5.8

5.8

5.8

5.8

Perturbation / RW6

6.5

6.5

6.5

6.5

8.5

6.5

6.3

Perturbation / RW5

2.8

5.5

5.8

5.8

5.8

5.8

5.8

6.0

5.8

5.6

5.6

5.6

5.6

5.6

5.6

Perturbation / RW7

3.0

6.4

6.4

6.4

6.4

6.4

6.4

6.5

6.4

6.4

6.4

6.4

6.4

6.4

6.4

Perturbation / RW8

Table 2 Evaluations of 22 RPs w.r.t. 29 combinations of qualitative criteria and RW categories

1.6

5.5

5.3

5.3

5.3

5.3

5.3

5.8

5.6

5.4

5.4

5.4

5.4

5.4

5.5

Perturbation / RW9

2.0

4.8

4.8

4.8

5.3

5.3

5.3

6.0

5.5

5.5

5.5

6.0

6.0

6.0

6.0

Ease of inlets insertion

6.3

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

7.8

Suitability / Adj. Section RBIT

7.0

7.8

7.8

7.8

7.8

7.8

7.8

6.3

6.3

6.3

6.3

6.3

6.3

6.3

6.3

Suitability / Adj. Section REME

6.3

4.8

4.8

4.8

4.8

4.8

4.8

4.0

4.0

4.0

4.0

4.0

4.0

4.0

4.0

Suitability / Adj. Section BCOM

4.0

2.5

2.5

3.0

3.0

3.0

3.0

2.5

2.5

2.5

2.5

2.5

2.5

2.5

2.5

4.0

2.0

2.0

2.5

2.5

2.5

2.5

2.0

2.0

2.0

2.0

2.0

2.0

2.0

2.0

Suitability / Adj. Section BAC

(continued)

Suitability / Adj. Section BBIC

Perturbation / RW2

3.1

3.1

3.1

3.1

3.1

3.3

2.9

Drainability

3.8

5.3

5.3

8.0

4.5

4.0

3.8

Alternatives

BCOM / SMA

BCOM / RMTO

BCOM / RUMG

BCOM / ED

BCOM / Enduit superficiel

BBIC (Béton bicouche)

BAC (Béon armé continu)

Table 2 (continued)

3.4

4.0

4.0

4.0

4.0

4.0

4.0

Perturbation / RW3

1.6

2.1

2.4

2.4

2.4

2.4

2.4

Perturbation / RW4

2.0

2.4

2.8

2.6

2.6

2.6

2.6

Perturbation / RW5

0.8

1.3

2.0

1.9

1.9

1.9

1.9

Perturbation / RW6

2.1

2.5

2.6

2.8

2.8

2.8

2.8

Perturbation / RW7

2.4

2.6

3.0

3.0

3.0

3.0

3.0

Perturbation / RW8

1.0

1.3

1.8

1.6

1.6

1.6

1.6

Perturbation / RW9

1.5

1.8

2.0

2.0

2.0

2.0

2.0

Ease of inlets insertion

1.8

1.8

6.3

6.3

6.3

6.3

6.3

Suitability / Adj. Section RBIT

1.8

1.8

7.0

7.0

7.0

7.0

7.0

Suitability / Adj. Section REME

4.5

4.5

6.3

6.3

6.3

6.3

6.3

Suitability / Adj. Section BCOM

7.5

8.0

3.0

3.0

4.0

4.0

4.0

8.5

7.5

3.0

3.0

4.0

4.0

4.0

Suitability / Adj. Section BAC

(continued)

Suitability / Adj. Section BBIC

Suitability / Location RW3

5.3

4.8

5.0

4.8

5.0

6.7

4.5

4.5

3.8

4.3

RBIT / BB-1

RBIT / RMD

RBIT / SMA

RBIT / RMTO

RBIT / RUMG

5.0

6.5

6.8

1.8

3.3

BCOM / RMD

5.3

3.7

REME / RUMG

5.3

6.5

3.3

REME / RMTO

3.0

5.3

4.0

REME / SMA

3.7

6.5

4.0

REME / RMD

REME / ED

4.3

5.7

RBIT / RBCF

REME / Enduit superficiel

4.8

5.0

3.5

4.3

RBIT / ED

RBIT / Enduit superficiel

Alternatives

Suitability / Location RW2

Table 2 (continued)

5.5

5.0

5.5

6.8

6.0

6.5

6.8

6.5

6.5

6.3

7.5

6.8

7.3

7.5

7.3

5.8

Suitability / Location RW4

8.8

7.8

4.3

6.0

5.8

5.8

7.0

6.3

2.8

2.8

3.8

3.5

3.5

4.8

4.0

3.5

Suitability / Location RW5

3.5

3.5

5.0

6.3

6.3

6.3

6.3

6.3

7.3

5.8

7.8

7.8

7.8

8.0

7.8

7.8

Suitability / Location RW6

7.5

7.5

4.8

5.0

6.3

6.3

6.3

6.3

3.8

3.8

4.0

4.0

4.0

4.0

4.0

3.8

Suitability / Location RW7

3.0

3.0

4.7

5.0

5.0

5.0

5.0

5.0

6.3

6.3

6.5

6.5

6.5

6.5

6.5

5.8

Suitability / Location RW8

1.8

1.8

4.0

5.3

5.3

5.3

5.3

5.3

7.3

5.8

7.8

7.8

7.8

8.0

7.8

7.0

Suitability / Location RW9

5.8

5.5

3.0

4.0

3.5

4.0

4.3

4.0

2.0

2.0

3.0

2.3

3.0

3.0

2.8

3.0

Maintenance requirements

5.0

5.0

5.8

5.8

5.8

5.8

5.8

5.8

6.5

6.5

6.5

6.5

6.5

6.5

6.5

6.5

Ease of implement. if RBIT

6.0

6.0

6.7

6.7

6.7

6.7

6.7

6.7

4.7

4.7

4.7

4.7

4.7

4.7

4.7

4.7

Ease of implement. if REME

4.0

4.0

6.3

6.3

6.3

6.3

6.3

6.3

5.0

5.0

5.0

5.0

5.0

5.0

5.0

5.0

Ease of implement. if BCOM

3.0

3.0

6.0

6.0

6.0

6.0

6.0

6.0

5.0

5.0

5.0

5.0

5.0

5.0

5.0

5.0

Ease of implement. if BBIC

(continued)

3.0

3.0

6.0

6.0

6.0

6.0

6.0

6.0

5.0

5.0

5.0

5.0

5.0

5.0

5.0

5.0

Ease of implement. if BAC

5.8

6.3

5.8

6.3

1.8

2.0

3.8

4.0

BCOM / ED

BBIC (Béton bicouche)

BAC (Béon armé continu)

6.8

2.0

BCOM / RUMG

BCOM / Enduit superficiel

5.8

Suitability / Location RW3

2.0

Suitability / Location RW2

BCOM / RMTO

BCOM / SMA

Alternatives

Table 2 (continued)

4.3

4.0

3.0

5.3

4.3

5.0

Suitability / Location RW4

8.8

8.3

5.5

8.0

7.3

7.3

Suitability / Location RW5

3.3

3.5

3.8

3.5

3.5

3.5

Suitability / Location RW6

8.3

8.0

6.0

6.3

7.5

7.5

Suitability / Location RW7

2.3

2.7

2.8

3.0

3.0

3.0

Suitability / Location RW8

1.5

1.8

2.0

1.8

1.8

1.8

Suitability / Location RW9

8.0

8.0

4.3

5.3

5.0

5.5

Maintenance requirements

5.8

5.8

5.0

5.0

5.0

5.0

Ease of implement. if RBIT

3.7

3.7

6.0

6.0

6.0

6.0

Ease of implement. if REME

3.3

3.3

4.0

4.0

4.0

4.0

Ease of implement. if BCOM

2.7

3.7

1.7

1.7

3.0

3.0

Ease of implement. if BBIC

3.7

3.7

1.7

1.7

3.0

3.0

Ease of implement. if BAC

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

93

Table 3 Evaluation of qualitative criteria used for RW on E411 Criterion QUAL1 (drainability) QUAL2 (perturbation due to works) QUAL3 (ease of inlets insertion) QUAL4 (suitability given adjacent zones) QUAL5 (suitability given works location) QUAL6 (maintenance requirements) QUAL7 (ease of implementation)

Eval 2.5 6.3 10.0 2.0 3.5 3.0 5.0

Justification Independent of RW characteristics Type of RW is RW5: RGG DescrRW4: No inlets RP in adjacent zones is RP5: BAC Type of RW is RW5: RGG Independent of RW characteristics Current RP is RP5:BAC

3.5 tons). Given the characteristics of the RW, the cost of any RPS can be computed by using the data in the MET database. The cost evaluation is expressed in €/m2. The useful range of the cost criterion is the interval [5,40]. QUANT2: Rutting The rutting resistance of an RPS is computed by using a traffic simulator with standard traffic conditions (available from BRRC). The result is a rut depth expressed in millimeters. The useful range is the interval [0,20]. This criterion is to be minimized. The danger threshold is a 16 mm rut depth. From this value on, the road surfacing must be repaired. QUANT3: Skid Resistance This corresponds to the (dimensionless) “transverse friction coefficient,” which can be measured using an appropriate apparatus (“odoliographe”). The useful range is [0.40; 1.00]. This criterion is to be maximized. The statement of work for roads in the Walloon region stipulates that the friction coefficient should be at least 0.45. QUANT4: Cracking Rate The MET’s Design Software, DimMET© (SPW Mobilité 2012), is used for evaluating the (dimensionless) cracking rate taking the traffic load and the type of road structure into account. When the cracking rate reaches 50%, the RPS must be replaced. The useful range is [0, 50], expressed in %. This criterion is to be minimized. QUANT5: Noise The noise generated by a tyre rolling on an RPS is measured in the lab (CRRB) under standard conditions. The unit is the decibel (dB). The useful range is [70, 85]. This criterion is to be minimized. Note that the evaluations of QUANT1 and QUANT4 depend on the parameters of the RW, while those of the other quantitative criteria do not. Marginal Value Functions The shapes of the marginal value functions are displayed in Appendix B. These shapes have been determined during a session of the working group. The process was supported by showing the shapes of the functions on a screen in the room. Experts could react and ask for changes. The reasonings that led to these marginal value functions are outlined below. For all quantitative criteria, we tried to build a function that represents differences of preference on the criteria. In the additive value function model, it is meaningful to compare preference differences on each criterion without specifying common values

94

A. Fiordaliso et al.

on the other criteria. Reaching a consensus on the shapes or the marginals, based on comparisons of value differences, was relatively easy. Note that all marginal value functions represented in Appendix B are non-decreasing even those corresponding to criteria to be minimized. In these cases, reasonings have been held in terms of value loss. A negative tradeoff is assigned to them when computing the overall additive value function. QUANT1: cost (first figure in Appendix B). The marginal value function for the cost is linear in the logarithm of the cost. This corresponds to comparing costs in relative values (e.g., “RPS X is 10% more expensive than RPS Y”). A cost increase of a given percentage is represented by the same value difference independently of the initial cost. In particular, whenever cost doubles, from 5€ to 10€ to 20€ and to 40 €, the losses in marginal value are equal. Therefore, the marginal value losses associated with 5, 10, 20, and 40 are respectively 0, 33.3, 66.67, and 100 on the [0,100] value scale. QUANT2: rutting (second figure in Appendix B). It is known that from 16 mm rut depth, the road surfacing needs repairing. Therefore, the disutility for 16 mm is maximal (100). The experts felt that the marginal value function is concave. They estimated that around 8 mm (resp. 12 mm) rut depth, the road surfacing has left 75% (resp. 95%) of its marginal value. From 0 to 8 mm, they judged that the loss of value was linear. A concave curve fitted the points (8, 75), (12, 95), and (16, 100). QUANT3: skid resistance (third figure in Appendix B). It was estimated that 80% of the value lies in the interval [0.45, 0.65]. A friction coefficient less than 0.45 is not allowed for roads in the Walloon region. Therefore, the marginal value is 0 below 0.45. For lack of further insight, the working group assumed that it is piecewise linear on the rest of the domain, i.e., on the interval [0.45, 1], with a breaking point at (0.45, 80). QUANT4: cracking rate (fourth figure in Appendix B). It was estimated that 85% of the marginal value of an RPS is lost when the cracking rate reaches 15%. For lack of further insight, the working group assumed that the loss is piecewise linear on the domain, i.e., on the interval [0, 50] %, with a breaking point at (15, 85). QUANT5: noise (fifth figure in Appendix B). For lack of insight, the working group assumes that the marginal value function is linear on the domain, i.e., on the interval [70, 85] dB.

3.3

Tradeoffs

DMs would have been ready to assess criteria weights by reasoning in terms of criteria importance. This would have been “the most common critical mistake” ((Keeney 1992), p. 147). Criteria weights or tradeoffs were estimated by means of indifference judgments made by four experts from the working group. The reference criterion is the cost (QUANT1). For example, the question for eliciting the tradeoff for the rutting resistance (QUANT2) is:

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

95

Which price am I willing to pay for improving a RPS1 costing 10 € with 8 mm rut depth into a RPS2 with 4 mm rut depth (all other evaluations being equal)? RPS1 : ð8 mm, 10 €Þ  RPS2 : ð4 mm, ?€Þ Similar questions were raised for assessing the tradeoffs for quantitative criteria (with reference to the cost criterion). For qualitative criteria, the question has to be adapted. For instance, for drainability (QUAL1), the question reads: Which price am I willing to pay for improving a RPS1 costing 10 € with weak surface drainability into a RPS2 with excellent surface drainability (all other evaluations being equal)? RPS1 : ð2, 10 €Þ  RPS2 : ð5, ?€Þ (We recall that 2 (resp. 5) is the value associated to weak (resp. excellent) surface drainability.) Each tradeoff (except that of the cost criterion which was set to one) was determined by the answer to one question. Tradeoff estimation Consider for example the tradeoff between cost and rutting resistance. Assume that an expert declares he/she is willing to pay 20 € instead of 10 € for reducing rut depth from 8 mm to 4 mm, all other evaluations being equal. In the additive value model, this means that: k cost ucost ð10Þ þ krut urut ð8Þ = kcost ucost ð20Þ þ krut urut ð4Þ Using the values ucost(10) = 100 - 33.3, urut(8) = 100 - 75, ucost(20) = 100 66.67, urut(4) = 100 - 37.5 (that can be read from the first two figures in Appendix B) and setting kcost = 1, we get: krut = 33:34 37:5 = 0:89. Standard Conditions Of course, such indifference judgments can be polluted by the RW category and other parameters of the works. For instance, the tradeoff between cost and noise is likely to be different whether the expert has in mind RW in an urban environment (RW9) or in the countryside (RW2). Therefore, the experts were asked to consider a standard situation, namely that of a highway (RW5), more specifically works on the E411 road (for which the other parameters are familiar to the experts), for all their indifference judgments in assessing tradeoffs. Results At the end of a working group meeting, four members (the GM and three out of the RMs and EDs) were assigned the task of making the required indifference judgments for the next meeting. The four experts worked independently. The corresponding computed tradeoffs are shown in Table 4 (the reference tradeoff for cost is set to 1).

96

A. Fiordaliso et al.

Table 4 Tradeoffs obtained through the indifference judgments made by four experts Rutting resistance (Orniérage) Skid resistance (Adhérence) Cracking rate (Taux de fissuration) Noise (Bruit) Drainability (Drainabilité) Perturbations due to works (Perturbations dues aux travaux) Inlets (Impétrants) Suitability adjacent zones (Adéquation sections adjacentes) Suitability works location (Adéquation emplacement voirie) Maintenance (Entretien) Ease of implementation (Facilité de mise en œuvre)

JW 0.338 0.317 0.155 0.132 0.422 0.153

ZK 0.756 0.317 0.501 0.293 0.422 0.359

ML 0.522 0.220 0.287 0.293 0.293 0.422

BS 0.433 0.317 0.223 0.293 0.422 0.293

Mean 0.512 0.292 0.292 0.253 0.390 0.307

0.153 0.293

0.293 0.153

0.153 0.293

0.153 0.293

0.188 0.258

0.153

0.483

0.153

0.153

0.236

0.541 0.293

0.359 0.293

0.652 0.153

0.652 0.153

0.551 0.223

Discrepancies between tradeoffs assessed by experts can be large. Due to time constraints, it was not considered realistic to try to reach a consensus of the experts about the tradeoffs. It was decided to use the average value of the four tradeoffs in the model. The working group has considered that the judgment of each expert, even divergent, had to be taken into account, hence using an arithmetic mean (instead of discarding “extremes,” for instance). These tradeoffs will be used for determining which is the best RPS for works on the E411 road and more generally on highways (RW5).

3.4

Revision

The parameter values previously elicited were finally reviewed and some revisions took place. In particular, Table 2 was modified into Table 5. Changes were made within a technical working group composed of OP and three experts in the road domain (from BRRC and FEBELCEM, the Federation of the Belgian Cement Industry), who had not been involved previously. These changes were endorsed by the MET working group. The main adjustments made are the following: 1. The RPS “crossing” was removed (as not specific). 2. The list of alternatives was revised (see Table 5, first column: DG stands for “dalles goujonnées” (“dowelled slabs”)). 3. Some combinations “criterion x RW” (corresponding to columns in Table 2) were aggregated: “suitability to pavement of adjacent zones x {BBIC, BAC}” need not be distinguished; they are substituted by “suitablility to pavement of adjacent zones x Concrete”; also, “Easiness of implementation x {RBIT, REME}” can be

Table 5 Revised evaluations of RPS w.r.t. 27 combinations of qualitative criteria and relevant RW category

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing 97

98

A. Fiordaliso et al.

merged together; finally, “Easiness of implementation x {BBIC, BAC}” are substituted by “Easiness of implementation x Concrete.” 4. Adding a new criterion named frost susceptibility proved necessary (see below). 5. Some evaluations in Table 2 were modified. 6. Some RPSs may not be used in certain categories of RW. This is mainly the case for RW2 (agricultural access roads). These incompatibilities are represented by black cells in Table 5. All revised evaluations are displayed in Table 5. QUAL8: Frost Susceptibility Some types of RPS are not recommended in regions exposed to intense frost during long periods. The experts assigned a (qualitative) frost susceptibility value (from the [0,10] interval) to each RPS. The associated marginal value function, obtained by means of indifference judgments, is represented in Appendix C. The weight (tradeoff) assigned to this criterion depends on the work location. It is computed as a linear function of the frost index of the city in Wallonia that is closest to the works (referred to as DescrRW7 in the list of works descriptors at the end of Sect. 2.2). A table reports the frost index of the most important cities in Wallonia. This index (expressed in °C day) is maximal in Bastogne and minimal in Tournai. The criterion tradeoff varies from 0.138 for works close to Tournai to four times as much (i.e., 0.555) for works close to Bastogne.

4 Tuning Based on RW Type Since tradeoffs were elicited using a standard situation (works on E411 highway), the corresponding model will not fit well in all other RW situations. Initially, three other road works cases were studied. The ranking obtained using the “standard” model was confronted to the experts’ judgment. On this basis, – The tradeoffs attached to some criteria were modulated depending on the RW category (and perhaps the parameters of the RW). – The evaluations of some criteria were revised. Finally, not only the RW category had to be taken into account but also other works characteristics (see list of six parameters in Sect. 2.2, to which the position of the works has to be added in order to determine the works frost index).

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

4.1

99

Computing the Value Associated to an RPS for a Given RW

The value u(x) of an RPS x is a sum over qualitative and quantitative criteria of the corresponding marginal values of the RPS weighted by the tradeoff associated to the criterion and the RW type: RW kRW i ui g i ð x Þ

uð x Þ = i2QUANT[QUAL

where – The evaluations giRW (x) of x w.r.t. qualitative criteria are taken from Table 5. Some values not only depend on the RPS but also on RW characteristics, namely, type of RW (RW2 to RW9), Current road pavement (DescrRW1), Current road pavement of adjacent zones (DescrRW2); the marginal values ui for qualitative criteria are linear except for the frost sensitivity criterion (see Appendix C). – The marginal values ui(giRW (x)) of RPS for quantitative criteria are computed using the functions illustrated in the five figures in Appendix B. – The weights or tradeoffs kiRW depend, in general, both on criterion i and on characteristics of the RW. We detail below how these tradeoffs are assessed.

4.2

RW-Dependent Tradeoffs

The tradeoff values in Table 4 were assessed for the RW5 case; they are adequate for works on highways. Since these tradeoffs were assessed by indifference judgments in terms of cost differences, it was asked to estimate the fraction of the cost difference corresponding to other types of RW. The reference tradeoffs were thus modulated using these estimated fractions, according to the type of road works and to characteristics of the works. Rutting The tradeoff related to the “Rutting resistance” criterion (QUANT2) is modulated depending on the intensity of heavy truck traffic. The tradeoff in Table 4 (0.512) is multiplied by the coefficients in Table 6 according to the number of trucks per day on the road (this parameter is specified as DescrRW5 in the work description). Table 6 Multiplicative coefficient to be applied to the weight of the “Rutting resistance” criterion according to the number of trucks per day Number of trucks per day (NT) NT < 250 250 ≤ NT < 2000 2000 ≤ NT

Coefficient 0.3 0.6 1

100

A. Fiordaliso et al.

Type of Road According to the type of road (RW2 to RW9) on which the work is located, a multiplicative coefficient is applied to the reference tradeoffs. Table 7 displays these coefficients. The value “0” indicates that the criterion is irrelevant for the type of RW. Frost Susceptibility The tradeoff value is a linear function of the frost index of the city closest to the works (see Sect. 3.4). The tradeoff thus ranges from 0.138 (works close to Tournai) to 0.555 (works close to Bastogne). The validity of these coefficients was checked on real RW examples (which sometimes led to adjusting previous values of the coefficients). The process of validating the model on cases is described in the next section.

5 Validation We distinguish two aspects in the validation process. One is related to the analysts (i.e., the participant(s) who is (are) expert(s) in MCDA methods) and another is related to the members of the working group that are experts in RW.

5.1

Logical Validity

It is a basic requirement of decision-aiding processes that the analyst(s) should be convinced that the chosen method(s) is used properly, which implies that the way of eliciting the method parameters should not betray the spirit of the method. That is what we call here “logical validity” (see, e.g., (Bouyssou, Perny, et al. 1993) and (Bouyssou, Marchant, et al. 2006)). We made every effort to ensure that questions posed to the experts did not cheat with the concepts of additive value function theory. Tradeoffs have been elicited as tradeoffs, not in terms of criteria importance. The construction of marginal value functions is consistent with the model. The value function can meaningfully be used to rank order the alternatives; in addition, differences of preferences on each criterion can be meaningfully compared because the marginal value functions have been constructed with a view to represent the ordering on single criterion preference differences.2 For eliciting the marginal value functions, we could not use the most rigorous methods, such as building standard sequences by means of indifference judgments 2 We do not assume that overall preference differences are correctly represented by value differences (as in the measurable value function model (Dyer and Sarin 1979), because we felt unlikely that the experts could reliably compare overall preference differences. In any case, no attempt has been made to test the validity of the representation of overall value differences by the value function model. Only the validity of the representation of single criterion preference differences has been checked and then used in the case-based validation process described in Sect. 5.2.

RW2: Agricultural access road RW3: Parking lot RW4: RESI roads connecting cities RW5: RGG highways, main roads RW6: Roads in commercial zone RW7: Roads in industrial zones RW8: Local road RW9: Urban roads, streets

1

1 1

1

1

1

0 1

1

1 1

1

1

1

1 1

0.5 1

1

1

1

0.5 1

1

1 1

1

1

1

1 1

1

0.5 1

0.5

0.5

1

0 1

0.5

0.5 1

1 0.5 2

2

2

1

1

1

0.5 1

0.5

0.5 1

0.5

1 1

1

1

1

0.5 1

0.5

1 1

1

1

1

0.5 2

1

3 3

3

3

3

3 3

3

0.5 1

1

1

1

0.5 1

0.5

1 1

1

1

1

1 1

1

0.5 1

1

1

1

0.5 1

1

Rutting Skid Cracking Suitability Suitability Ease of Cost resistance resistance rate Noise Drainability Perturbations Inlets Adjacent location Maintenance implementation Frost

Table 7 Multiplicative coefficients to be applied to criteria tradeoffs in the different types of RW

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing 101

102

A. Fiordaliso et al.

((Von Winterfeldt 1986), Sect. 8.1 or (Bouyssou and Pirlot 2016)), due to practical reasons (size of the working group, time constraints). Nevertheless, we paid attention to correctly represent preference differences on each criterion. Moreover, questions regarding tradeoffs were formulated in a rigorous manner. Of course, due to the large number of assessments required from the experts, one cannot expect that all assessments are accurate. So, even though the process seems logically consistent with additive value function theory, there was a need to confront the model to a number of decision situations that are well understood in a holistic manner3 by the experts. This is described in Sect. 5.2.

5.2

Validation by Cases

A dozen real RW examples with representative characteristics and specified location were analyzed. The computed values of the RPS allowed to rank them from the most desirable to the least. The experts in the working group checked whether the RPSs ranking produced by the model in the cases was in agreement with the experts’ experience. In most of the cases, the RPS rankings were found to be correct by the experts in the working group. Detailed histograms such as in Fig. 1 were helpful to analyze the contribution of each criterion in the value assigned to the alternatives. It allowed, in particular, to validate the fact that an alternative is better ranked than another. Some anomalous inversions in the RPS rankings were detected and analyzed, mainly on the basis of Fig. 1. Parameter adjustments were made, especially regarding the multiplicative coefficients applied to tradeoffs for taking the RW category into account (Table 7 contains the values eventually adopted). Example The characteristics of works on the E411, the reference case used for assessing criteria tradeoffs, are the following: – – – – – – – –

Type of road: RGG (RW5). Current road pavement: BAC. Current road pavement of adjacent zones: BAC. Length and surface of the works site: 5000 m. Inlet insertion: no. Traffic load: 5280 trucks/day. Number of working days: 365. Frost susceptibility: works close to Bastogne.

The corresponding RPSs ranking is illustrated in the histogram represented in Fig. 1. The RPSs values are ranked in decreasing order (the larger the better). Each 3

By this we mean that the experts are familiar enough with these cases so that they are able to judge whether the ranking of the RPS by the model makes sense in such a context. In other words, they are able to detect anomalies in the ranking.

e é e 1 é A A A D D D D CF ED TO D el G G O O G M M BM / E RM SM RM / E uch nud MT SM MT fici M ud ch M r U / R RB T / RM / S / B / / o én icou RU RU M é / / e E R R R I c / i / / D D / / E M M E M O up T / BIT IT RB IT / BIT BIT M BC EM EM O RE / BB / B ME CO OM it s / B BB E I R B R R RB C C / EM CO G u RB R BC E R B R C d G A D R R B B n B BA D E / T BI R Alternatives

0,0

50,0

100,0

Coût

Perturbation Drainabilité Bruit Taux de fissuration Adhérence Résistance à l'orniérage

Adéquation Rev. / Secton adj. Facilité pour pose d'impétrants

200,0

150,0

Facilité de mise en œuvre Entretien de la voirie Adéquation Rev. / Emplacement

Sensibilité hivernale

250,0

300,0

Fig. 1 Values of the RPSs ranked in decreasing order in the case of works on E411

Value

350,0

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing 103

104

A. Fiordaliso et al.

bar is decomposed in smaller bars representing the weighted marginal value of each criterion. Users can appraise at a glance the contribution of each criterion to the value of the RPS. They can compare these contributions in different RPSs and track the reasons why an RPS is ranked better than another. Such a graphic tool enables to identify the strong points (long segment corresponding to the contribution of a criterion in a bar) and weak points (short segment). The results raised the following comments: “The alternatives BAC/BDénudé and BBicouche appear on top of the ranking and well ahead, as expected. Their advantage over the other solutions is consistently reflected in the lengths of the colored bars. In the 3rd and 4th places arrives RUMG as surface layer, which is correct. RBITs are logically relegated to the bottom of the ranking.” A similar analysis was performed in working group for all validation cases.

6 The Decision-Aiding Software EVAL-MET The development of a decision-aiding software implementing the additive value function model was the object of a contract between the MET and the University of Mons. The software was written in Java. The data was stored in a database containing the list of alternatives (RPS), that of road works types (RW), the evaluations of the alternatives w.r.t. the criteria (including the auxiliary models used to compute evaluations whenever they depend on RW characteristics such as trucks traffic load), the marginal value functions, the tradeoffs, and the multiplicative coefficients used to adapt tradeoffs to RW. The program also makes calls to external technical modules, such as the design software dimMET© (SPW Mobilité 2012), to compute the thickness of the RPSs and the cracking rate as a function of traffic load and characteristics of the RPS layers.

6.1

Interface

Figure 2 shows the user interface of the decision-aiding software devised to help the RW manager choose the best possible and most suitable RPS. – The user enters the characteristics of the RW in the bottom left part of the screen (“Paramètres du chantier”). – The criteria weights, normalized and un-normalized, appear in the upper part of the screen (“Poids des critères”). The weights (tradeoffs) determined by the working group are the default. The user may enter his/her own weights and experiment with them. – Below is a sheet where the RPS evaluations w.r.t. all criteria are displayed (“Score des revêtements”). They are computed internally, taking into account the RW characteristics. They cannot be modified by the user. The last two

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

105

columns show the values (or scores) of the RPS computed using the model; the last (resp. last but one) column displays the scores computed using the weights entered by the user (resp. the default weights). – The bottom right part is a graphical representation of the scores (“Plots/Scores détaillés”). The histogram in Fig. 2 represents the scores (and their decomposition w.r.t. the criteria) for the user’s weights. The same representation is available for the default weights (by selecting “Pds par défaut” in the upper part of the screen). The RPSs can be sorted in decreasing or increasing order of their scores (default or user defined). In the histogram in Fig. 2, one can see that the yellow part (corresponding to criterion “Rutting resistance” or “Orniérage”) is dominant in all bars. This is a direct consequence of the large weight assigned by the user to the criterion “Rutting resistance”). By selecting two rows in the spreadsheet in Fig. 2 (one blue and the other red), one may contrast the corresponding two RPSs. The bar diagram in Fig. 3 is visible in the “Deltas scores” tab. It shows on which criteria the alternative BCOM/RMTO has an advantage w.r.t. alternative RBIT/ED (for a given RW and the default weights); the score difference on a criterion in favor of BCOM/RMTO are represented by the height of the red bars. The blue bars show where and how much RBIT/ED has an advantage over BCOM/RMTO. This may help the user understand why an alternative is ranked before another. It may start a process calling the model’s parameters into question.

6.2

Using the Decision-Aiding Tool EVAL-MET

Indications on how EVAL-MET should be used were given by the direction of the MET. The RW manager must use the software to choose the most adequate RPS for a given RW. As soon as he/she introduces the parameters of the RW, the program computes the value of each RPS by using the default tradeoffs (those elaborated by the working group). The RW manager can use the software to: – Compare an alternative to another and examine the strengths and weaknesses of one as compared to the other. – Modify the default tradeoffs and examine the changes in the RPSs ranking produced. – Depart from the best solution(s) recommended by EVAL-MET provided he/she writes an explicit justification for another RPS and reports to the MET about his/her decision; arguments justifying the choice of another RPS may rely on using the software with tradeoffs chosen by the user. Of course, using different tradeoffs requires justifications that can be challenged by the hierarchy. In a first step, it is planned to dispatch the program to the RW managers on a CD. To ease the maintenance of the software and of the database, as well as the program evolution, it is advisable to move to a decision-aiding tool installed on a

Fig. 2 A view of the interface of the decision-aiding software EVAL-MET

106 A. Fiordaliso et al.

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

107

Fig. 3 Score difference per criterion: in favor of BCOM/RMTO in red; in favor of RBIT/ED in blue

central server; the users would access the program through an internet connection. This evolution was not implemented.

6.3

The Acceptance Issue

The model and the decision-aiding tool were presented to the users, i.e., the Field Engineers managing RW, in a meeting organized by the MET (February 22, 2006). The decision-aiding tool was not warmly received. According to some participants, choosing a pavement or a wearing course is the engineer’s role and function; a software cannot help an engineer because it cannot consider all the cases and contextual issues. At this stage, the mission which was the object of the contract was complete. The contract did not involve either users training or model or software maintenance. The BRRC could have taken over these tasks. OP continued to work for BRRC until August 2007. He provided assistance to use the model and the decision-aiding tool upon request from Field Engineers. In subsequent months, the GM changed the rules for preparing public contracts by adding the use of Eval-MET into the specifications. He received complaints from some Field Engineers (we do not have precise information on the substance of these complaints). When he retired in October 2008, the use of this software was made

108

A. Fiordaliso et al.

optional by the new GM. The project of providing assistance to the Field Engineers for choosing appropriate RPS in RW fizzled out. The BRRC did not pursue in this direction either. Historically, the BRRC’s role was to provide scientific expertise on road techniques to the actors, public and private, involved in maintaining and developing the road network in Belgium. Developing or maintaining a decisionaiding tool was probably a bit unusual for the BRRC which focuses on technical expertise. Despite the criticisms about Eval-MET, the pavement of a major road project was chosen using this software. The engineer in charge of managing this RW was convinced of the great usefulness of this multi-criteria decision-aiding tool. Therefore, he used the EVAL-MET software, with the help of OP, to choose the RPS of the Couvin bypass, a 14 km long road with a twin layer-continuously reinforced concrete pavement (BAC/BBicouche).

7 Looking Back on the Process We briefly analyze the process of this intervention in terms of strengths, difficulties, and procedural approaches to face process weaknesses. We finish with some conclusions.

7.1

Strengths

Our main assets in this intervention were twofold: – The involvement of OP who was the project linchpin as a researcher at BRRC and student or ex-student at the University of Mons. OP played the key role of a translator-interpreter. He was ideally positioned to explain decision-aiding concepts to the MET, and the goals, constraints, and technicalities of road management to the team at UMONS. – The commitment and leadership of the MET top management. Its determination in regulating and optimizing the decisions regarding the choice of RPS in RW was the project driving force. Not all RMs and EDs members of the working group were convinced by the approach; some were skeptical and were mildly resisting the top management’s will. It was useful to understand dynamics and internal logics of the organization. The involvement of the working group experts in the validation cases was essential to detect and analyze some anomalous inversions of the alternatives in the rankings.

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

7.2

109

Difficulties and Process Weaknesses

The main difficulties were the following: – Field Engineers (FEs) had not been involved in the project elaboration process. We underlined the risk associated with this decision without any result. – Being forced to use a normative tool such as the EVAL-MET software reduces the FE’s freedom and autonomy in decision. We anticipated resistance. The initial views of the top management were that the program would determine the best choice, which should be followed by the FE in charge of the RW. We insisted the right concept was decision-aiding, not decision-making. During the working group sessions, we negotiated that the software should be used by the RW manager, but the recommendation issued by the software could be challenged and dismissed by setting out rational arguments, reported to the MET administration. This was accepted. However, it seems that little has been done in advance to convince the Fes that using the software could be beneficial. – The model and the software were validated in several ways before they were presented to the users but not tested in real situations. It is quite likely that feedback by the users of encountered problems (e.g., inconsistencies in rankings issued by the model) would have led to reworking the model and the software. – Software maintenance issues could not be anticipated. Alternatives (i.e., RPSs) change with technical progress and experience. Evaluations probably need revision. Feedback from the users must be examined and should trigger changes in the model and software. To avoid the proliferation of different versions, it will be necessary to provide a regularly updated official version available through internet on a server. Some process weaknesses needed specific approaches and sometimes generated model fragility: – There was no consensus among the experts in the working group regarding the evaluations of alternatives w.r.t. certain criteria, and regarding tradeoffs. The dispersion of experts’ assessments was sometimes substantial, and heavy time constraints limited discussion and analysis of specific misunderstandings. Averaging the experts’ assessments was compulsory but it could blur judgment variability. – The multiplicative coefficients applied to the reference tradeoffs dispensed from the time- and energy-consuming task of eliciting tradeoffs for each RW type (which was not an option). But the resulting models might be only a rough approximation of the models that would have resulted from eliciting the tradeoffs. – The validation and fine-tuning of the models was based on a dozen of real RW cases. Despite this nonexhaustive validation, the fact that no ranking sounded strange to the experts helped build trust in the model.

110

7.3

A. Fiordaliso et al.

Success or Failure?

It is difficult to state a final judgment on this intervention since it could not be fully developed due to the priority shift of the MET top management. The latter sort of event is not uncommon in decision-aiding interventions (Brown 2009). In our case, the reasons for that shift are the arrival of a new direction after the retirement of the general manager. The involvement of the FEs in the project elaboration process could perhaps facilitate the complete development of the intervention. Had the priorities of the new direction remained the same, the process could have been pursued. The key issue at that stage was to develop ownership of the model and decision-aiding tool by the FEs. It was necessary to allocate resource and time to let the model and the software evolve interactively with the FEs in charge of RW. In that way, acceptance of the tool and of its usage rules could probably be reached. Assuming that the necessary resource was dedicated to the users’ acceptance process, the model’s maintenance would have required continuous attention. Moving to a centralized version is a vital necessity. It is also imperative to rely on internal resource and skills (either in MET or BRRC) to maintain and evolve the model and the software, as well as to manage the interactions with the users. A monitoring committee should also have been constituted to supervise the evolution of the tool. Occasionally, experts in decision models could be consulted to ensure the logical consistency of the evolving model.

Appendix A. Alternatives. The first column in Table 8 specifies the type of road pavement (5 types, RP1 to RP5); the second column specifies the possible surfacings (wearing course) for each type of road pavement.

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

111

Table 8 Types of road pavements Type RP1: Asphalt concrete (RBIT: revêtement bitumineux)

RP2: High modulus asphalt concrete (REME: revêtement bitumineux avec enrobé à module élevé)

RP3: Composite concrete (BCOM: béton composite)

RP4: Two-layer concrete (BBIC: béton bicouche) RP5: Continuously reinforced concrete pavement (BAC: béton armé continu)

Wearing course BB-1 RMD SMA RMTO RUMG ED Surface dressing (Enduit superficiel) RBCF RMD SMA RMTO RUMG ED Surface dressing (Enduit superficiel) RMD SMA RMTO RUMG ED Surface dressing (Enduit superficiel) ---------

ED porous asphalt; RUMG ultra-thin, grained coating; RMTO ultra-thin, open-textured coating; RMD thin, open-textured coating; BB bituminous concrete; RBCF slurry seal; SMA stone mastic asphalt Table 9 Alternatives RP4 and RP5 RP4: Dowelled slabs (DG: Dalles goujonnées) RP5: Continuously reinforced concrete pavement (BAC: béton armé continu)

Nude Two-layer Nude Two-layer

The latter two groups of alternatives (RP4 and RP5) were restructured in the revision phase (see Sect. 3.4) as they are described in Table 9. An example of form filled by each expert for each of the qualitative criteria is presented in Table 10.

112

A. Fiordaliso et al.

Table 10 Form to be filled for the evaluation of qualitative criteria. Example: QUAL1, drainability 1. Drainabilité du Revêtement 2= 5= 8= 0

drainabilité de surface faible drainabilité de surface excellente drainabilité dans la masse satisfaisante 6 1 2 3 4 5

7

8

9

10

0

1

6

7

8

9

10

RBIT / BB-1 RBIT / RMD RBIT / SMA RBIT / RMTO RBIT / RUMG RBIT / ED RBIT / Enduit superficiel RBIT / RBCF REME / RMD REME / SMA REME / RMTO REME / RUMG REME / ED REME / Enduit superficiel BCOM / RMD BCOM / SMA BCOM / RMTO BCOM / RUMG BCOM / ED BCOM / Enduit superficiel BBIC / Béton bicouche BAC / Béton armé continu 2

3

4

5

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

113

B. Marginal value functions for the quantitative criteria

Marginal Value Function for Cost (Fig. 4)

COST marginal value function (semi-logarithmic)

40; 100

100 90 80

Value

70

20; 66.67

60 50 40 30 20 10 5; 0

0 1

100

10 Cost (€)

Fig. 4 Marginal value function for cost (logarithmic scale for cost) on the [5,40]€ cost interval

Marginal Value Function for Rutting Resistance (Fig. 5)

RUTTING RESISTANCE marginal value function

100 16; 100

12; 95

90 80 8; 75

70

value

60 50 40 30 20 10 0 0

2

4

6

8

10

12

14

16

rutting depth [mm]

Fig. 5 Marginal value function for rutting resistance on the [0,20] mm interval

18

20

114

A. Fiordaliso et al.

Marginal Value Function for Skid Resistance (Fig. 6)

SKID RESISTANCE marginal value funtion 100 90 80 0.65; 80 70

value

60 50 40 30 20 10 0.45; 0 0 0.4

0.45

0.55

0.5

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Tangential Friction Coefficient [-]

Fig. 6 Marginal value function for skid resistance on the [0.45, 1] interval

Marginal Value Function for Cracking Rate (Fig. 7)

CRACKING RATE marginal value function 100 90 15; 85

80 70

value

60 50 40 30 20 10 0 0

5

10

15

20

25

30

35

Cracking rate [%]

Fig. 7 Marginal value function for the cracking rate on the [0, 50]% interval

40

45

50

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

115

Marginal Value Function for Noise (Fig. 8)

NOISE marginal value function 100 90 80 70

value

60 50 40 30 20 10 0

0

75

80 Noise [dB]

Fig. 8 Marginal value function for the noise on the [70,85] dB interval

85

116

A. Fiordaliso et al.

C. Marginal value function for Frost Susceptibility (Fig. 9) FROST SUSCEPTIBILITY marginal value function 100 90 80 (5; 66,7 )

70

Value

60 50 40 30 20 10 0 0

1

2

3

4

5 6 Frost susceptibility

7

8

9

10

Fig. 9 Marginal value function for the frost susceptibility qualitative criterion

References Belton V, Stewart TJ (2002) Multiple criteria decision analysis. An Integrated Approach, Springer, Berlin Bouyssou D, Pirlot M (2016) Conjoint measurement tools for MCDM. In: Ehrgott M, Figueira JR, Greco S (eds) Multiple criteria decision analysis. State of the art surveys, Springer, New York, pp 97–151 Bouyssou D, Perny P, Pirlot M, Tsoukiàs A, Vincke P (1993) A manifesto for the new MCDA era. J Multicrit Decis Anal 2(3):125–127. https://doi.org/10.1002/mcda.1508 Bouyssou D, Marchant T, Pirlot M, Tsoukiàs A, Vincke P (2006) Evaluation and decision models with multiple criteria: stepping stones for the analyst. Springer, New York Brown RV (2009) Working with policy makers on their choices: a decision analyst reminisces. Decis Anal 6(1):14–24. https://doi.org/10.1287/deca.1080.0134 CRR (2019) Odoliographe-Mesure de l'adhérence des chaussées. Fiche 6, Bruxelles: centre de Recherches Routières (CRR) Dyer JS (2016) Multiattribute utility theory (MAUT). In: Ehrgott M, Figueira JR, Greco S (eds) Multiple criteria decision analysis. State of the Art Surveys, Springer, New York, pp 285–314 Dyer JS, Sarin RK (1979) Measurable multiattribute value functions. Oper Res 27(4):810–822. https://doi.org/10.1287/opre.27.4.810 Keeney RL (1992) Value-focused thinking. Harvard University Press, Cambridge MA MET (2002) Revêtements hydrocarbonés et en béton armé continu sur les autoroutes. Comparaison économique. Les Cahiers du MET-Collection Techniques, n°19, Direction générale des

A Decision-Aiding Tool for the Choice of Road Pavements and Surfacing

117

Autoroutes et des Routes, Ministère de l'Equipement et des Transports, Namur, Belgique: Service Public de Wallonie Roy B, Bouyssou D (1993) Aide Multicritère à la Décision: Méthodes et Cas. Economica, Paris SPW Mobilité (2012) Le logiciel de dimensionnement DimMET. Service public de Wallonie Mobilité. http://qc.spw.wallonie.be/fr/qualiroutes/dimmet.html. Accès le 12 25, 2022 Vincke P (1992) Multicriteria decision-aid. Wiley, Chichester Von Winterfeldt D, Edwards W (1986) Decision analysis and behavioral research. Cambridge University Press, Cambridge UK

Multicriteria Decision Aiding for a Shift Towards Best Environmental Practices in Agriculture, with a Focus on Viticulture Francis Macary

1 Introduction One and a half centuries of agriculture has seen an intensification of agricultural practices and an increase in chemical treatments: fertilisers to improve yields, pesticides to control pests and diseases, and weed control to manage competition from harmful weeds (Barriuso 2004). Whilst this has led to a significant increase in the productivity of all annual and perennial crops, it has also resulted in the simultaneous degradation of ecosystems (Bockstaller et al. 1997; Girardin et al. 2000; Payraudeau and Van Der Werf 2005). Moreover, the excessive use of pesticides has led to water and soil contamination, biodiversity reduction (Bengtsson et al. 2005), impoverishment of organic matter in soils (Van Der Werf 1996), and human health problems (Bohnen and Kurland 1995; Baldi et al. 2012). Many studies have shown that there is a relationship between the long-term exposure of humans to pesticides and the development of acute and chronic diseases, such as neurological diseases and cancers (Inserm 2021). There is also evidence that different components of the environment are contaminated with pesticide residues, especially surface water and groundwater (Leonard 1990) (Hildebrandt et al. 2008) (Granoulis 2009; Boithias et al. 2012). Pesticides have short-term effects on non-target organisms (Roubeix et al. 2010; Chaumet et al. 2019) and long-term effects on landscapes, habitats, and the food chain (Berendse et al. 2004). It has been established for more than a decade that pesticide use must therefore be urgently reduced and more sustainable production systems need to be developed (Surgan et al. 2010). This situation applies to orchards and vineyards in particular, which aim to produce satisfactory yields and highquality grapes for the production of quality wines (INAO, IFV 2017). Their fruit is F. Macary (✉) INRAE, ETTIS, F-33612, Cestas, France e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_5

119

120

F. Macary

more susceptible to pests and diseases than are large-scale crops, whether for direct consumption or for transformation into wine or cider, for example. It is, therefore, necessary to protect such crops, regardless of the type of cropping system applied; i.e. conventional (intensive or more or less sustainable), or organic (without synthetic chemical inputs, and completely natural, organic, or mineral) (Madelrieux and Alavoine-Mornas 2012; Merot and Wery 2017). However, due to growing public opposition to chemical inputs (i.e. pesticides) and stronger regulations for protecting natural habitats and human health, the agricultural sector is modifying its practices and production methods in order to shift towards more sustainable agriculture, respecting the three pillars of sustainability: environment, economy, and society (European Commission 2002; Angevin et al. 2017). To these, a fourth pillar can be added: the notion of knowledge transfer to future generations in the case of family farms, mostly in Europe, as well as in Africa, Latin America, and Asia. Therefore, modern agriculture is facing new challenges arising from the negative impacts of conventional agriculture on the environment and human health. Meeting these challenges by reducing pesticide reliance, improving biodiversity, adapting to climate change, and reconciling farmers and especially winegrowers with consumers would pave the way to more sustainable agriculture (Gliessman 1998). Decision aiding can have an important role in relation to changes that may involve the current management practices in fields, orchards, and vineyards which include heavy use of pesticides (Aouadi et al. 2021). New concepts of production have been developed for more than 20 years. Amongst them, this paper is focusing on agroecology (Altieri 1995), which is considered to be a scientific discipline involving systemic analysis by integrating human and social sciences, a set of practices integrating the principles of ecology in agronomy (working with nature), and a social movement (Gliessman 2015; Méndez et al. 2016; Hatt et al. 2016; Fao 2018). In viticulture more specifically, different authors have highlighted the importance of developing agroecological vineyards which combine management innovations and land-use planning at plot, farm, and landscape levels (Macary et al. 2020). However, before they can change their practices and their production methods, farmers and winegrowers need information, in particular, technical and economical ones (Renaud-Gentié et al. 2014; Métral et al. 2015; Aouadi et al. 2021). A few years ago, some wine industry and professional organisation stakeholders requested that a group of researchers at Bordeaux University and three French public research institutes in environmental and life sciences (INRAE, CNRS, IFREMER) develop a project to study different aspects of pesticides within a vineyard in the Bordeaux region. The professional expectations of these stakeholders were on the use of pesticides and their impacts on ecosystems (e.g. on soil, water, and biodiversity), ecotoxicology, the consequence of the implementation of good agrienvironmental practices, as well as to explore other alternative concepts, like agroecology. This project, called “PhytoCOTE” (Macary and Devier 2014), was carried out from 2015 to 2019. The aim of the project was to make the farmers and winegrowers aware of the consequences of their intensive viticultural practices on the surrounding ecosystems (for example the quality of water, soil, and biodiversity).

Multicriteria Decision Aiding for a Shift Towards Best. . .

121

This is essential in order to propose and discuss new production practices in terms of some scenarios to be used as a basis for decision-making when making a shift to sustainable viticulture. Within the framework of the integrated “PhytoCOTE” project, the use of pesticides on vine in vineyards in south-west France (Gironde) and their transport to the stream and impact on the surrounding ecosystems were studied along with the effects of change in production methods and practices. Such changes can be considered as issues with an important role for stakeholders in participatory analyses. These changes may be analysed by using multicriteria decision aiding methods that facilitate the analysis of environmental, economic, and societal aspects of a decision problem and, when required, can be linked to Geographic Information Systems (GIS) (Sobrie and Pirlot 2012). It was therefore decided to apply a multicriteria approach to decision aiding, whether for the geographical studies (of vineyard parcels in a catchment area1) or in the case of a group of winegrowers. Such approaches had also been applied in some of the other projects, for example for the evaluation of agri-environmental risks in areas with large arable farms (Macary et al. 2014a) and in very erosive areas containing livestock (Macary et al. 2010). The methodology can be adapted to different conditions and specific regional characteristics. This project was chosen to illustrate in this chapter a sequence of multicriteria analyses at different levels of spatial organisation: the watershed for the evaluation of the agri-environmental risks of surface water contamination by pesticides (Aouadi et al. 2018); the viticultural systems within agricultural areas, which sometimes comprise other types of production (e.g. asparagus in our study area and even livestock) (Aouadi et al. 2019); and, in areas where agriculture is more diversified, large-scale arable crops, in particular oilseed and protein crops. The first section proposes a synthetic description of the decision aid project, aim and content of each step and some details on the involved stakeholders. The second section describes the multicriteria approaches adopted in the three steps of the project, whilst the third section presents multicriteria models, used methods and results. The last section analyses some difficulties of these complex projects and proposes some useful approaches. Some results are synthesised in the Conclusions.

2 Project Description Habits and traditions have an important role in agriculture and viticulture, each region has its own peculiarities (Pelsy and Merdinoglu 2021). It is, therefore, very difficult to make professionals aware of the real dangers of pesticides to ecosystems and human health (including their own), because that knowledge-based decisions are

1 Watershed: An area of land, demarcated by a drainage divide that drains all the streams and rainfall to a common outlet.

122

F. Macary

difficult to manage even is those people are “professionals”. To remedy this, awareness first needs to be raised regarding the risk of ecosystem degradation as a result of the heavy use of pesticides2 (Aouadi et al. 2018). To analyse these multiple processes, we built and developed the PhytoCOTE project. Human health is not dealt with in this project; this would require an epidemiological study to be carried out, as well as measurements of pesticide contamination in winegrowers and their staff at the time of phytosanitary treatment and during the different post-treatment vineyard practices, thus calling upon other skills—especially in epidemiology—and very different approaches (Baldi et al. 2012). Moreover, this is still a very sensitive subject in the professional world overall. The project, therefore, focused actually its research on environmental and socioeconomic issues; it is too earlier in the process, to transfer knowledge to future generations. The PhytoCOTE project comprised two main work packages: WP1—Agri-environment and agricultural economics • Understanding the choices made by winegrowers and their practices, including vineyard health protection. • Evaluation of the environmental and socioeconomic performance of winegrowing systems and of prospective scenarios. WP2—Environmental chemistry and ecotoxicology • Understanding transfer processes, bioaccumulation, and impacts of plant protection products on ecosystems linked to vineyard parcels (e.g. soil and water). To illustrate the applications of the multicriteria methods for decision aiding, this chapter only deals with WP1, in which they were coupled with a GIS. The project took place in the winegrowing region of Blaye to the north of Bordeaux in south-west France (Fig. 1). This winegrowing region comprises 6700 ha of vines. Detailed research was carried out on the experimental watershed of Marcillac, alongside hydrologists, hydrobiologists, and ecotoxicologists, who evaluated the water quality of the upstream river (“La Livenne”) and associated tributaries. This small watershed has an area of 830 ha and the agricultural soil is almost exclusively used for winegrowing (75%), the rest being meadows along the river and streams (Fig. 2). It should be noted that this zone is relatively isolated from other vineyards, being in the middle of a largely forested area; this was an important aspect in the choice of study area to ensure that the analysed pesticides in the river and its streams came only from this winegrowing area and not another. Winegrowing practices can thus be linked with the pesticides measured in the water and soil and a multicriteria analysis can be used to identify best-existing practices and production methods in the field. This first analysis facilitated the elaboration of production system scenarios (Aouadi et al. 2021).

2

Pesticides or plant protection products: in agriculture, a chemical preparation to control selfpropagating plants (weeds) or to protect crops from insect pests or disease-causing organisms (e.g. fungi and bacteria).

Multicriteria Decision Aiding for a Shift Towards Best. . .

Fig. 1 Location of the study area

Fig. 2 Land use of the experimental watershed of Marcillac

123

124

F. Macary

Organisaon of an inial public informaon meeng for wine growers and their advisers on the project and its objecves in the municipality of the study area

Idenficaon of the geographical locaon of the watershed and producon of the list of winegrowers who manage at least one parcel in the watershed

Step 1 - Environmental risk assessment at plot level in the experimental watershed

Presentaon of results and discussion

Step 2 - Assessment of environmental and socio-economic performance of wine growers mainly in the watershed and surroundings

Presentaon of results and discussion

Step 3 – Design of scenarios for the development of pracces and evaluaon of their environmental and socio-economic performances

Presentaon of results and discussion

Technical and scienfic presentaon of results: scienfic conferences, final conference, meengs in the field with wine growers, intervenons at the Instute of Agricultural Sciences (Bordeaux), technical and scienfic publicaons

Fig. 3 The different steps of an agroecological approach

The project involved the close collaboration of the winegrowers and their advisers, and winemaking and marketing cooperatives. Initially, we asked whether it would be better to model some best production practices or to develop completely new systems to reconcile production with nature environment and consumers. The winegrowers said that, as a public research institute, the aim should be to produce some innovative scenarios, then to assess them to obtain sound references for decision aiding when choosing the best system of production. The project therefore considered environmental and socioeconomic aspects. Forty vineyards were studied, and new systems were modelled in an agroecological way to obtain the best economic and environmental performances. Before the project even started, the whole process needed to be preceded by a presentation of the project background, objectives, and expected results during a public information meeting with all participating parties. This was necessary before the first step (identification of the agri-environmental risks to ecosystems) and then for the second step (Assessment of environmental and socioeconomic performance of winegrowers) (Aouadi et al. 2021). The third step describes the adopted approach (Fig. 3) that evolved towards achieving an agroecological production process, that is necessary for breaking with conventional production that is the systematic use of chemical pesticides to the detriment of biological processes and microbial activity in the soil. The

Multicriteria Decision Aiding for a Shift Towards Best. . .

125

objectives and expected results of each step are here synthesised, the multicriteria analyses that were carried out in the steps to facilitate decisions are described in Sects. 3 and 4. Figure 3 summarises the three steps, which are described below. Step 1. Evaluation of the risk of pesticides reaching a winegrowing watershed Our first aim was to determine the risk of pesticides from vineyards in the Bordeaux area (France) reaching the watercourses of the experimental watershed (8, 5 km2). A survey was carried out on professional winegrowers active in at least one vineyard parcel in the watershed, whether on a full- or part-time basis or retired, were surveyed. From the information gathered from the winegrower survey, it was possible to propose a multicriteria model to evaluate the 690 agricultural/winegrowing parcels in the experimental watershed. A method of multicriteria analysis, coupled with a Geographic Information System (GIS), was used to assign each parcel to a level of risk of pesticide transfer for each vineyard (Aouadi et al. 2018). Fieldworks results were then presented in the form of a risk map of the study area to the winegrowers and their advisers. Recommendations were given, during the presentation of the results, for possible improvements in terms of pesticide pressure, cultivation practices, and buffer zones. Step 2. Evaluation of the environmental and socioeconomic performances of the viticultural systems within each vineyard Step 2 aimed at understanding winegrowing enterprises and, above all, the decisionmaking rules behind choices linked to the existing production systems and practices. Professional winegrowers based in the experimental watershed were surveyed; others outside of this watershed (in the winegrowing region of Blaye, north of Bordeaux) were also contacted, to better represent the different viticultural systems and practices. It was also necessary to take into account their concerns about the risk of reduced harvest and crop quality. The survey showed that the winegrowers were careful with respect to the maximum number of treatments in a crop year, and they tried to consider any potential impacts on the environment and for the society, as well as the cost of pesticides. In terms of plant protection measures within the vineyard, the most important aspects to be dealt with resulted the climate, vineyard observations, agricultural warnings, and agricultural advice. The agricultural and winegrowing practices of each of the studied vineyards were recorded, in particular, those linked to the protection of vineyard health. We also recorded the reasons for choices made by the winegrowers in terms of the pesticide treatments applied to the vines or the alternative protection measures. The environmental and socioeconomic performances of each of these real production systems were evaluated and methods of multicriteria analysis were used to determine their environmental and social values (Aouadi et al. 2020). Step 3. Design and evaluation of new scenarios of viticultural systems Using these results, new realistic scenarios to change the existing practices and thus strive for best environmental and socioeconomic performances were developed. This

126

F. Macary

was done by taking into account the best practices which had been identified over the whole study area, and theoretically combining them into one virtual vineyard (Aouadi et al. 2019). Three scenarios were built in relation to three different viticultural systems: – Application of more or less sustainable practices, which can be improved without being completely organic (Conventional, or intensive, system). – Application of specific agroecological concepts and methods, but continued use of chemical fertilisers for the protection of plant health (Agroecological system3). – Organic and agroecological system. In all three cases, crop management techniques were developed and adapted to each scenario, but it was decided to keep only the best practices already identified in this study area in different viticultural systems, but never, all in the same, exception for the only agroecological way. In order for the different production systems to be well represented, the involved winegrowers were conventional and organic winegrowers from the study area. It was important to convince the sector stakeholders of the credibility and reliability of the proposed scenarios. Both the alternative systems (the three scenarios) and the aforementioned real systems were assessed, and their value was determined. The results were then discussed with experts to determine their applicability in the shift to sustainable farming.

2.1

The Research Institutes and Different Stakeholders Mobilised for the Project

The research group in the WP1 of PhytoCOTE included three French public research institutes in environmental and life sciences, INRAE, CNRS, and IFREMER. Ten teams from public research institutes took part. INRAE is the National Research Institute for Agriculture, Food, Environment and the coordinator of the project and tasks in agronomy and agroecology. CNRS is the National Centre of Scientific Research and was involved through the Laboratory of environmental chemistry and ecotoxicology, in mixed structure with Bordeaux University. IFREMER is the French Institute for Marine Research (laboratory of ecotoxicology in Nantes). The

3

Agroecological system: Production system linking agriculture and ecology and relying on the functions of ecosystems, which it amplifies while aiming to reduce pressure on the environment (e.g. greenhouse gas emissions and plant protection products) and preserve natural resources. Agroecological production was promoted by the French Ministry of Agriculture from 2012 (Maaf 2012a) (Guillou et al. 2013) (Maaf 2014) based on scientific work (Altieri 1995) (Gliessman 1998). It was also integrated within the international work of FAO (2018). The National Institute of Appellations of Origin (INAO) and the French Institute of vineyard and wine (IFV) produced a book on agroecological practices in the vineyard, from which we were inspired to build the scenarios (INAO, IFV 2017).

Multicriteria Decision Aiding for a Shift Towards Best. . .

127

High School of agronomy “Bordeaux Sciences Agro” was also associated with the task in the economics of viticultural systems. The project was funded for 50% by the National Research Agency, via the University of Bordeaux; the regional council of New-Aquitaine for 30%; the University of Nantes for a PhD thesis (10%); High school of Bordeaux agronomy (5%) and laboratories involved in the project (5%). The Tutiac Winegrower Cooperative, the most important cooperative for protected designation of origin wines in France, was an important partner in the field as the winegrowers on our study site are members and the research group had a permanent relationship with its field technicians. The headquarter of Tutiac is in the village of Marcillac, which is the site of the experimental watershed, and the town hall provided the use of a village community centre for all the meetings with the winegrowers and external advisors. Different institutions and professional organisations were asked to participate in the WP1 of the PhytoCOTE project with different roles. The Nouvelle Aquitaine Regional Council is the regional institution that fulfils its economic, environmental, and social policies by providing funding for research projects and economical stakeholders. The regional councillor responsible for viticulture in our study area followed our progress. Furthermore, the Council’s research department was represented in the project’s steering committee and this link contributed significantly to the regional visibility of the project and to establishing the necessary institutional contacts. The Winegrowers Union of the wines of Bordeaux, Côtes de Blaye (our study area), represents all producers, whether in cooperatives or independent. It helped communicate the project to all the regional winegrowers via its regular internal newsletter. The union’s agricultural engineer provided us with the winegrowers’ contact details, useful because these winegrowers do not necessary apply the same practices as those in the experimental watershed. The Technical Institute of Vine and Wine in Bordeaux is the regional branch of the National Institute, which develops missions for applied research on vine and wine. It was consulted in Steps 2 and 3 of the WP1, when the criteria for evaluating the performance of the winegrowing systems were modelled and when new scenarios for more efficient production methods in terms of environmental and economic objectives were developed. The federation of municipalities of the Gironde Estuary was involved as an information source. We worked with a federation project manager who is in charge of monitoring the water quality of the river downstream of the outlet of the experimental watershed. The Gironde Chamber of Agriculture has been given the mission of public service in agricultural development by the French ministry of agriculture. No matter the agricultural sector, all farmers are members by right. Technical advisors disseminate the latest innovations from research institutes to farmers and winegrowers. The advisors could indicate if the information from our surveys were representative or not of the whole vineyard in this region. The Regional Directorate for Agriculture, food, and forests is the ministerial department that is responsible for implementing both national and European Union

128

F. Macary

policies in the field. A national pesticide expert represented it in the steering committee. The project’s steering committee also included as a member the director of the technical service department of the Interprofessional Council for Bordeaux Wines, who gave some advice to clarify presentations and when we built scenarios of new practices. The Council represents the whole of the Bordeaux wine sector, promotes Bordeaux wines worldwide, and provides producers with important technical support and represents the sector legally. The working group also included advisors from the Tutiac Winegrower Cooperative, and a group of multidisciplinary researchers from INRAE (in the fields of agronomics, economy, ecophysiology, and plant protection). These experts worked on different stages of our study, namely the identification of the study area, conducting surveys, criteria selection, and weighting and design of the new systems. During Step 2, conventional and organic winegrowers from the study area, who wished to develop and apply more sustainable practices in the vineyard, had been surveyed for their winegrowing practices. They were either members of the winery or independent and actively participated in several discussions with the research group and the other participants. To achieve the aim of proposing environmentally and socioeconomically viable sustainable practices, all stages of the project were discussed with the stakeholders previously mentioned. Their continual involvement from the very first design stage of the project was essential to ensure that the outcome of our research fulfilled their expectations, i.e. the creation of relevant decision aiding tools. It was important to build and maintain mutual trust, and to ensure that the analysts committed to refraining from divulging any personal results, especially in public meetings. All the individual results were of course given anonymously, and some were combined.

3 A Multicriteria Analysis of Decision Aiding This decision aiding study comprised some main stages (and several activities in each stage): analysis of the problem, with the involvement of all the stakeholders; model structuring with the definition of the potential alternative actions and evaluation criteria; assessment of the alternatives, use of multicriteria methods, presentation of the results. Figure 4 shows the general structure of the agri-environmental approach used in the project. This approach was used in the three steps that were followed by the communication of the results, paying particular attention to the challenges that had to be overcome and the solutions that were chosen to carry out the project successfully. The aims and main activities of each stage are here described. Then, the multicriteria models and procedures, and the results of the methods applications are synthetically proposed.

Multicriteria Decision Aiding for a Shift Towards Best. . .

129

Mulcriteria decision aiding project with spaal reference

Potenal alternaves

Stakeholders

Criteria for judging alternaves

Assessment of alternaves

Mulcriteria aggregaon

Mapping of results Discussion with stakeholders

Fig. 4 Structural diagram of the agri-environmental approach of decision aid

3.1

Step 1. Assessing of the Risk of Pesticides Reaching a Viticultural Watershed

The public project information meeting took place in the multi-purpose hall of Marcillac, Gironde, where our study is based. The winegrowers were then asked to participate in a survey to gather information on the following: • The vineyard manager (e.g. age, qualifications, role in the vineyard, availability, and decision-making responsibilities), cooperative member, or independent winegrower. • The characteristics of the vineyard, what it produces and its spatial distribution. • The vineyard workforce. • The agri-environmental approaches applied: type of methods (i.e. conventional, organic, or agroecological), any quality labels (e.g. organic, Terra Vitis, and Agriconfiance). • Soil maintenance techniques. • Perceived extent of pests and diseases in the vineyard. • Plant health protection: level of observations, rules for the application of chemical or biological treatments, role, and importance of external advisors. • Implementation of prophylactic measures (e.g. destruction or removal of vine stock diseases).

130

F. Macary

Table 1 Survey carried out on professional winegrowers and others in 2016 for the mapping of agri-environmental risks linked to pesticides

Owning parcels within the watershed Owning parcels outside the watershed (and not within it)

2016 Survey Winegrowers in a cooperative 32 1

Independent winegrowers 4 5

• Spraying: characteristics of the equipment used, satisfaction regarding the material used, table of treatments applied over a whole year (date of application, commercial products and active substances, doses applied per hectare), satisfaction in terms of protection, choice of priority parcels in a vineyard for spraying. • Harvested crops in the year of survey: quantity and quality, and marketing. The survey, in the form of both questionnaires and interviews, was carried out on the winegrowers of the Tutiac winery, which had informed them of the project and requested that they take part. This was much appreciated, given the general reluctance of conventional winegrowers to discuss pesticides. Independent winegrowers were also asked to participate in the survey (Benkirane 2015; Pupier 2015). All the 36 winegrowers who had a vineyard in the watershed took part, whether the vineyard was their main or secondary activity (pensioners included). We also targeted already known six other professional winegrowers with vineyards outside of the watershed in order to complement other production methods (organic) and to understand their choices as regards practices for protecting the vineyards (Table 1). The survey aimed to gather information on all the 690 parcels and to establish the level of agrienvironmental risk in each parcel and represent it on a map (Aouadi et al. 2018). It was possible to produce a summary of winegrowing management, practices, and decision-making procedures for plant health protection, from the information gathered from the winegrower survey. Viticulture was either their main or secondary activity (the latter case pertaining to those with multiple jobs, or even retired winegrowers, who have most often inherited just one vineyard, or the remainders of a farm). Using the treatment calendars provided by the winegrowers, their Indicators of Treatment Frequency Index (IFT4) (MASA 2022) were calculated and applied to their vineyards. The acquisition of environmental characteristics or on-site observations of vineyard operations did not require direct contact with the winegrowers; it was just necessary to inform them of our presence, as notes were being taken on their land and thus on private property. The choice of criteria used in the multicriteria analysis was initially discussed in the field with the winegrowing advisors from the cooperative, the Winegrowing union, and the Chamber of agriculture, as well as with the conventional and organic winegrowers. The main factors involved in pesticide transfer were considered. The

4

TFI: Number of pesticide reference doses per hectare.

Multicriteria Decision Aiding for a Shift Towards Best. . .

131

criteria were then built according to the literature and based on experience we had gained in similar previous projects, and of course taking into account the availability of quantitative or qualitative data for these criteria. It is important to note here that the analyst must have thorough field knowledge and have visited each parcel to be able to carry out a rigorous evaluation for all parcels. This work should preferably be carried out by two people to have a second opinion and thus reach a unanimous decision, as was done in the present study by the agronomist responsible for the project and his intern, who had also conducted a field visit to the whole of the watershed. Such field knowledge is essential in the development of the multicriteria model. However, it is clear that this entails limitations associated with the size of the geographic study area. When aiming to assign each parcel to a risk category—as in the present study—this method is perfectly applicable to a watershed of a small area (8, 5 km2 here), but it is hard to imagine the same operation in a watershed of several tens of square kilometres! Once the results have been obtained, they are exported on the GIS and presented in the form of maps showing the level of risk of each parcel, as shown later in § 4.2. For large areas, the GIS software, IDRISI© (Eastman 1988), can be used; this software integrates a multicriteria module for this purpose, based on the weighted sum. However, due to the simplicity of this module, the analysis and processing of the results are not as precise as that required in this project, also we did not use it here. In collaboration with decision makers, the analyst must make a choice between a large area for which a general information processing method is used or a smaller experimental area for which a more detailed analysis can be carried out. It is sometimes possible to link both of the above approaches. This was done here when zoning the risks of pesticide transfer into a large river basin with an area of 1150 km2, using Landsat satellite images (pixels: 30 m x 30 m) to determine land use and GIS to integrate the risk criteria (Macary et al. 2014a). In parallel, we coupled a multicriteria method with a GIS for a small watershed of 3 km2 (690 parcels) which was representative of a zone of high pesticide pressure from arable crops (Macary et al. 2013a). From this detailed analysis, we were able to develop a spatial model for the whole of the large river basin (Macary et al. 2013b).

3.2

Step 2. Evaluation of the Environmental and Socioeconomic Performances of Viticultural Systems

Knowledge regarding the current production systems of winegrowing enterprises was acquired to propose new systems that reduce pesticide use in vineyards. Agricultural enterprises were analysed by applying a systemic approach aimed at capturing both the environmental and socioeconomic aspects. The latter is important, because most farmers or winegrowers would be willing to change their practices if they can be reassured that they will not lose out financially and that their production system will not become too complicated to manage.

132

F. Macary

Some professional winegrowing enterprises were selected to be part of the study, namely those whose revenue from agricultural activity is higher than that of any other potential source of revenue. We surveyed 38 winegrowers—all volunteers— who apply different winegrowing practices (Table 4). The majority of them (29) belong to the Tutiac cooperative, of which 50% are located in the experimental watershed and 50% in the study winegrowing area outside of the watershed. Winegrowers members of a cooperative, winegrowers must apply the instruction of the direction and advisers, unlike the independents. The other nine surveyed enterprises are independent winegrowers, of which six were located outside the experimental watershed, to determine the diversity of their practices. The main aim of the survey was to characterise the winegrowing practices adopted by the surveyed winegrowers in relation to pest management and plant protection (e.g. treatments carried out, doses applied, and spraying equipment used), soil management (grassing and tillage), other vineyard operations like de-budding and pruning. It was also important to verify the presence of agroecological practices, such as biological protection of the vineyard, and the plantation of trees or hedges. Also, to understand how the work was generally organised in the vineyard parcels, the distances from the vineyard buildings, any management difficulties encountered because of production mode (e.g. organic farming requires more observations and work). Any information that would contribute to calculating expenses and help identify the winegrowing products was also recorded. Moreover, it was important for us to understand the reasons for choices made by the winegrowers, especially in terms of vineyard plant protection. We also identified the equipment used for the different operations—especially spraying equipment—and collected information on labour and wine production (yield and type of wine produced) to assess economic performance. All these information elements were used to describe and evaluate the 38 real production systems. First, they were assigned to the four pre-established performance categories (4.2.2 Table 12) by means of a multicriteria outranking method, ELECTRE TRI-C sorting algorithm (Almeida-Dias et al. 2010). Then the systems of each category were ranked using another outranking method, ELECTRE III ranking algorithm (Roy 1978; Martin and Legret 2005). Then the analysed systems were anonymously linked to their real modes of production: conventional, more or less sustainable in connection with a type of environmental certification (e.g. High Environmental Value and Terra Vitis), or organic. It was necessary to show the modes of production only at the end of the presentation of the ranking results, in order not to influence the analysts. Some details of the adopted procedure are described in Sect. 4.2 (Table 2).

Multicriteria Decision Aiding for a Shift Towards Best. . .

133

Table 2 Surveys were carried out amongst professional winegrowers and others in 2017 and 2018 to assess environmental and socioeconomic performances

Vineyards inside the watershed Vineyards outside the watershed

3.3

2017–2018 Surveys Winegrowers in cooperative 15 14

Independent winegrowers 3 6

Step 3: Development of Realistic Scenarios of Production Systems

Once the Assessment of environmental and socioeconomic performance of existing wine grower’s systems had been evaluated in Step 2, we developed three “realistic” scenarios that can enhance the existing winegrowing practices and improve their environmental and socioeconomic performances by strongly reducing the use and impact of pesticides. A group of experts in viticulture and involved stakeholders collaborated with us to design these new winegrowing systems. The development of organic viticulture is currently on the rise due to public pressure to stop or significantly reduce the use of synthetic pesticides, and due to consumer demand for organic wines. Some agroecological practices had been already applied in the surveyed winegrowing system, but not in the same vineyard; for example the winery encourages its winegrowers to adopt agroecological practices, like planting hedges and using biocontrol agents. Some winegrowers had also invested in confined sprayers to reduce pesticide drift. Therefore, we tried to combine these practices in new realistic scenarios. One of the eight organic vineyards involved in the study had developed many agroecological practices and its progressive shift towards an agroecological approach was considered a success story. This production model was the main source of inspiration for the design of the new scenarios. The identification of a set of constraints, specific to the production situation of the study area, the development of new systems, based on existing practices identified in the field, and the evaluation of their performances were the main activities of this participatory approach. In Scenario 1 (Maximised-conv-sys), we improved the strategies applied in the conventional systems. We assumed that these conventional winegrowers would never integrate concepts of organic farming and agroecology, but that they could improve their practices. We, therefore, modelled the best practices that they could achieve. In Scenario 2 (Agroecological system) and Scenario 3 (Agroecological organic system), a set of agroecological practices were introduced in the conventional model and the organic model respectively (Figs. 5 and 6). In scenarios 1 and 2, the pest control strategy consisted in using chemicals to remove CMR products (Carcinogenic, Mutagenic, or Reprotoxic). Copper was indicated for the initial treatments—essentially against downy mildew—and sulphur

134

F. Macary

3 Scenarios - Phytosanitary strategy

SC1- Maximised-conv-sys

SC3- Organic Agroecology

SC2- Agroecology

•

•

•

Synthec fungicides, no CMR; 3 treatments using copper, and treatment with sulphur against powdery mildew No an-botrys Synthec inseccides against bunch worms

•

Natural fungicides (copper & sulphur) Give preference to copper sulphate; Max. of 3 treatments with Cuprous oxide: No anbotrys

No inseccides against bunch worms Dosage depends on the vegetave stage

Dosage - First 2 treatments: < 35 % - Full vegetaon treatments = 80 % - End of season treatments: < 30-40 %

Cu metal dose = 200 g Cu max/ha/treatment

Zero Herbicide: Mechanical weeding under the row & in the inter-row

Fig. 5 Three scenarios for the sanitary strategy

3 Scenarios - Management of soil & green pracces

SC1- Maximised-conv-sys

SC2- Agroecology

SC3- Organic Agroecology

Natural grass cover ½ interrow Green manure ½ inter- row Cover management: mowing

Permanent green manure - mixture of species (legumes, grasses) Cover management: rolling

Grassy headlands

Grassy headlands; planng of hedges, trees, shrubs, bushes; nesng boxes; insect hotels, etc

Pruning; manual de-budding; mechanical leaf stripping and shoot trimming

Fig. 6 Three scenarios for soil and green operations management

for powdery mildew treatment. In the agroecological organic system, copper and sulphur are used to treat fungal diseases. The use of copper sulphate is recommended (lower phytotoxicity) and copper oxide (higher phytotoxicity according to advisers) is allowed depending on the weather conditions. Anti-botrytis treatments are

Multicriteria Decision Aiding for a Shift Towards Best. . .

135

replaced in the three systems by operations like de-budding, de-suckering, and pruning. Pesticide dose reduction was included in the three systems with the following decision rules: in the initial treatments, the applied dose must be 35% lower than the reference dose; in the vegetative growing season, the dose must be 80% lower than the reference dose; and in the final treatments, the applied doses must be lower than 30–40%. In terms of soil management, a permanent vegetative cover in all vine rows was included in the agroecological systems. This involves sowing a mixture of grass and leguminous seeds and rolling to renew the cover and enhance the supply of organic matter in the soil. The maximised conventional system included natural grass cover in every other row, which is mechanically cut. Within-row mechanical weeding was included in all three scenarios. The photos in Fig. 7 show some examples of good practices in an agroecological organic system located in the study area, which inspired us in the development of the scenarios.

4 Multicriteria Models, Methods, and Results A multicriteria approach was used in all the steps and as unifying languagefacilitated communication between analysts and stakeholders during the project. In the first step, a model above all from the field literature, was used to deeply analyse 690 agricultural/winegrowing parcels (only a part, 8 km2, of the watershed), which were representative of a zone of high pesticide pressure from arable crops. A method of multicriteria analysis was coupled with a Geographic Information System (GIS) and a spatial model for the whole of the large river basin was developed. Each parcel was assigned to a level of risk of pesticide transfer and these results were then shown in the form of two risk maps of the study area to the winegrowers and their advisers. Therefore, the results and the use of the maps are described in detail in this Section (4.1). Only a quick description is proposed of the multicriteria model and used method. A new multicriteria model was developed and used in the second and third steps and the integrated use of two multicriteria methods produced interesting results. Model and procedure are described in Sects. 4.2 and 4.3. From the very beginning of our involvement in this type of project, we have chosen to use the ELECTRE (Elimination and Choice Expressing Reality) family of methods. They were created in the modelling Laboratory of Decision Aid Systems (LAMSADE) of the University of Paris-Dauphine by Professor Bernard Roy and his team (Roy 1968, 1990; Vincke 1989; Maystre et al. 1994; Rui-Figueira et al. 2010). These methods have already been used by our team in many studies to model the risks of pesticide contamination of surface water (Macary et al. 2010, 2013a, b, 2014b).

136

F. Macary

Planng trees in the vineyard

Field beans between rows

Insect hotel for biodiversity

Permanent grass in the vineyard

Mulching the soil between rows

Nestbox for birds

Fig. 7 Ecological practices in an agroecological organic system (© Francis Macary)

The methods are well suited to the use of quantitative and qualitative criteria. The model is adaptable to each type of project by building the criteria according to the problem studied and by choosing a very precise setting.

Multicriteria Decision Aiding for a Shift Towards Best. . .

4.1

137

Step 1: Model and Results

The criteria of a multicriteria model must be understandable, independent (not related to each other), and discriminatory (sufficiently indicative of preference between two actions, and eventually decision-making, when all the evaluations on the other criteria are identical). The quality of a family of criteria should be “legibility” (to become a discussion basis allowing the analyst to assess inter-criteria information) and “operationality” (to be considered by all the involved actors as a sound basis for the decision aid study) (Bouyssou 1990). These conditions limit the criteria number. The criteria of this first model were six, in relation to two main aspects that characterise the risks associated with two types of pesticide molecules (Aouadi et al. 2018). Four criteria were included in the model, to express the conditions of the physical environmental characteristics: • Average slope of each parcel obtained from a digital field model. • Soil type determined from the local soil map. • How the vineyard and the nearest watercourse were hydraulically connected (ditch, path) and the distance between them. Two criteria indicated the agricultural practices: • The established pesticide pressure based on the combination of a treatment frequency index (number of approved doses applied to a surface) and the vineyard area calculated via the GIS. • Cultivation practices (grass cover, tillage, row orientation). One criterion concerns the environmental protection of the stream: • The buffer zone (quality, width) of the parcel for reducing pesticide drift which is a combination between riparian zones, wetlands, and grass along the stream. Figure 8 shows the relationship between the sources of information on the indicators obtained from the GIS and the winegrower survey. It shows how this information was used to create the criteria, which were implemented, in the multicriteria model. This representation was very useful for explaining the source of the retained criteria to the stakeholders (Macary et al. 2010). The weights for criteria were decided with the stakeholders, two field agricultural advisors, and four scientists (following the method of Simos, Roy, and Figueira (Figueira and Roy 2002). They took into account the two types of pesticide molecules: (i) hydrophilic; i.e. molecules dissolved and transported in water, and (ii) hydrophobic; i.e. not dissolved in water and thus adsorbed on suspended and organic matter and transported via erosive processes. These two situations highlighted for the winegrowers the importance of controlling water flow in the vineyard (the role of the buffer zones and inter-row grass cover) and of avoiding soil erosion, via, for example a significant amount of tillage on slopes. The weights are indicated in Table 3. It was easy to reach a consensus on those weights, because there

138

F. Macary

General structure of GIS

MCDA Method (Electre Tri-C)

(ArcGIS)

Types of Protecon

Survey informaon and ground observaon data

Wetlands & Grass along streams

6 Criteria

Cartographics data

Buffer zones

Riparian zones

Agricultural Pressure

Agroecological pracces Pescide pressure

Vulnerability of water

Land use Soils Soils

Nature of soils DEM 10 m

Slopes Slopes

slopes

Hydrographic network

AEPs Pescide pressure

Connecvity

Road network and ways

Fig. 8 Relationship between the GIS and MCDA method

Table 3 Step 1. Results of the weighting process Criteria Average slope Soil type Hydrological connection Buffer zone Pesticide pressure Vineyard operations

Hydrophilic molecules 18 7 22 10 30 13

Hydrophobic molecules 14 22 7 10 28 19

were only environmental criteria, and this result appears very credible in the field. There is no bibliography about this. We just have an experience about this kind of study for many years and published papers about this subject. The previous experience acquired in creating and implementing risk models facilitated the definition of the other parameters. It is important to pay particular attention to the definition of the thresholds. We used the threshold of veto for the multicriteria analysis of environmental risks because there are natural meadows along the edge of rivers and streams, which do not receive any pesticides, but their natural characteristics, make the parcel vulnerable to pesticide transfer. Therefore, the veto threshold reduced the risk level of these parcels to a lower level (no risk) as in reality they are not subject to any pressure. To validate the assignment of each parcel into a risk category, we chose a degree of credibility, δ = 0.7, which was strong enough to obtain a high processing level (an assignment can only be validated if 70% of the weighted criteria accept it). A value above 0.7 was considered excessive.

Multicriteria Decision Aiding for a Shift Towards Best. . .

139

Table 4. Step 1. Preference threshold, Indifference threshold assigned to each criterion, and Veto threshold for pesticide pressure

Categories Weights Indifference threshold Preference threshold Veto threshold Criterion direction Nature of thresholds

Average slope 14 0.1

Soil type 22 0

Hydrological connection 7 0

Duffer zone 10 0

Pesticide pressure (hydrophobic molecules) 28 0

0.2

1.9

1,9

1,9

0.95

1.9

–

–

–

–

7

–

Max

Max

Max

Max

Max

Max

Constant

Constant

Constant

Constant

Constant

Constant

Vineyard operations 19 0

Table 5. Step 1. Reference value of each category for all criteria Categories

Performances

Average slope

Soil type

Hydrological connection

Buffer zone

Pesticide pressure

Vineyard operations

C5 C4 C3 C2 C1

Very high High Medium Low Very low

12 8,5 5,5 3,5 1,5

7 5 3 2 1

10 8 5 3 1

18 14 11 7 3

27 23 19 15 10

30 24 17 9 2

The evaluation matrix was created in the form of an Excel spreadsheet, which was directly imported by the MCDA ULaval software tool (https://mcda.fsa.ulaval.ca/) developed by the team of Irene Abi-Zeid, a professor at the University Laval (Québec). For each criterion, Table 4 shows the preference direction, the criteria weights, the thresholds of Indifference and Preference, which were retained considering the scale of the performance values, and the veto threshold used for pesticide pressure. The direction of each criterion is maximum because the aim was to identify the high level of risks for the contribution of the pesticide transfers from each parcel. Table 5 shows the characteristic values by category for each criterion, determined according to the scale of values applied to wine systems by criterion. Results of Step 1 The 690 agricultural/winegrowing parcels in the watershed, which were representative of a zone of high pesticide pressure from arable crops, were evaluated in relation to risk criteria and assigned to hierarchical predetermined categories of risk by means of a multicriteria method, ELECTRE TRI. In this method, a parcel is assessed for its absolute value independently of the other analysed parcels. Each parcel is

140

F. Macary

Fig. 9 Contribution of the parcels to the risk of contamination of streams by hydrophilic pesticides in the experimental watershed

compared to different reference values that have been created to define the categories. The method is based on sorting and a process of segmentation (Mousseau et al. 2001) (Almeida-Dias et al. 2010). In terms of the contribution of each agricultural parcel to the risk of contamination of a watercourse, we created four categories of risk levels from high to very low and the levels were defined for each criterion under consideration. The method was used to compare the value of a parcel for each criterion to those of the different reference values and assign each parcel to one of the predefined categories. An integration between a multicriteria method coupled with a GIS allowed the results to be represented in maps. Two types of maps were thus produced: one for the hydrophilic molecules (Fig. 9) and the other for the hydrophobic molecules (Fig. 10). In terms of current practices, the maps show that there are more areas of high risk for pollution contribution by the hydrophobic molecules, because they are carried away on soil particles during erosion given the texture of the soils of this watershed (Aouadi et al. 2018). The maps show each parcel according to its final risk category. It is important to be able to explain the condition of each parcel in terms of the criteria and their weight. For example, the parcel shown as high risk on the map is near a watercourse on a steep slope, and it does not have a buffer zone. It is thus under severe pressure and the recommendations for the winegrower would be to apply the same measures

Multicriteria Decision Aiding for a Shift Towards Best. . .

141

Fig. 10 Contribution of the parcels to risk of contamination of streams by hydrophobic pesticides in the Marcillac area

as his/her neighbours in the same situation and to protect the watercourse with a buffer zone. To facilitate dialogue with the stakeholders, we also simulated diverse scenarios incorporating cultivation practices and used the same type of maps to show the results. In this way, for the hydrophobic molecules in each case, the high-risk and very high-risk parcels represent the percentage in brackets that was associated to each specific scenario: a simulation of reduced pesticide use (6%); an inter-row area weeded mechanically and chemically (20%); grassy and mown inter-row area and under the row (2%). Therefore, from a minimal to a maximal situation, the percentage of high to very high-risk areas could increase from 2 to 20%, if winegrowers applied good agricultural and environmental practices. Maps were very efficient when presenting results to stakeholders. Moreover, without having to point out the different parcels at risk, the winegrowers were able to recognise which was theirs and to assess the room for improvement. However, during the presentation, it was important to take as examples some very different parcels and explain how they contributed to the estimated risk by going over each criterion related to habitat vulnerability and human pressure, and thus showing where there was room for improvement. There was a particular focus on criteria related to plant protection and agroecological practices. Professionals who still were applying very conventional practices recognised progress in organic farming.

142

4.2

F. Macary

Steps 2 and 3: Model, Methodology, and Results

Following Step 1, we first identified the vineyard parcels which were the most likely to contribute to the transport of pesticides to surface water and which had the highest pesticide pressures. Then, it was necessary to gain a better understanding of the functioning of these winegrowing systems to be able to make suggestions for the development of production methods aimed at promoting a sustainable approach through the adoption of agroecological practices. An evaluation of the environmental and socioeconomic performances of these systems was therefore carried out in Step 2. The geographical area of the vineyard parcels in the experimental watershed was widened to include the whole winegrowing region of Blaye to incorporate other types of systems, namely those in which organic production methods were applied. An integrated methodology was adopted for the implementation of a multicriteria analysis, the same in Steps 2 and 3. In the second step, 38 existing viticultural systems were studied: their environmental and economic performances were assessed in relation to a set of seven criteria. The ELECTRE Tri-C method (Almeida-Dias et al. 2010) was used to assign each system to one of the four pre-defined categories of socioeconomic and environmental performance levels. Subsequently, we compared all the systems within the same category of performance (comparable actions) and ranked them using the ELECTRE III method (Roy 1978) to identify the winegrowing systems that resulted in the most efficient pesticide reduction use whilst maintaining high profitability. The distinction between Modelling 1 and 2, in Fig. 11, indicates that the criteria may be different in the two procedures. They were the same in the study, to reduce complexity and work, for the stakeholders. In Step 3, the only change was that the three scenarios were included in the model, assessed once they had been designed and the methodology re-activated.

Modelling 1

Modelling 2

Sorng problem

Ranking problem

Electre TRI-C Model

Electre III model

Assigning each VS to one of the four predefined categories

In each category ÆClassificaon of VS

Idenficaon of producon models Characterizaon of producon models aer ranking VS

Study area

¤ ¤ ¤ 7 criteria Performance of VS ¤ ¤¤ ¤ ¤ ¤¤¤ C1: Very high y=4 ¤ ¤ ¤ ¤ ¤ categories C2: High ¤ ¤ ¤ ¤¤ ¤ N viticultural systems (VS)

C3: Medium C4: Low

Example Cn

Example Cn

Example Cn

VS1, VS2, VS3,VS4

VS3 VS2, VS4 VS1

VS3 VS2, VS4 VS1 Convenonal farming/ Organic farming/ Environmental cerficaon

Fig. 11 Methodology for MCDA modelling, using ELECTRE methods, for Steps 2 and 3

Multicriteria Decision Aiding for a Shift Towards Best. . .

4.2.1

143

Criteria

The aims of Step 2 were to determine the diversity of practices adopted by the winegrowers in the field, and to assess the environmental and socioeconomic performances of the systems, to assign each system to a performance level, and then to identify the best-performing ones. To do so, we met socioeconomic stakeholders in the field (winegrowers and professional advisers), and together we identified seven representative criteria to be used for evaluating the production systems. The same criteria were used in Step 3 to design and evaluate new production scenarios. Table 6 synthesises the structure of the multicriteria model, which includes three performance aspects and seven criteria. CR1 (REN) Profitability of the viticultural system—a quantitative criterion Our aim here was to compare all stages of the different winegrowing systems up to, but not including vinification and marketing in order to determine which ones would perform the best when reducing plant protection inputs. Agricultural revenue was not calculated, as we did not wish to ask the winegrowers for their financial documents in order to avoid creating problems in the collaboration, especially as the winegrowers were all participating on a voluntary basis and the issue of pesticides is already a sensitive area. Using data from the cooperative, we therefore established values per hectare of agricultural vine product depending on the environmental practices applied (Table 7). It was relatively easy to identify these practices from the survey, and to record the corresponding values of the produce, using X, or (X) when it is not possible, as shown in Table 7. CR2 (PPS) Pesticide pressure—A quantitative criterion This criterion was evaluated using the treatment frequency index (TFI) calculated for each pesticide using the following formula (Maaf 2012b): dose × treated surface area Pesticide TFI ðper hectareÞ = Applied Registered dose=total surface The overall TFI for each viticultural system was evaluated by calculating the sum of TFI for all the pesticides, which were weighted according to the fraction of treated surface. The preference is for decreasing TFI.

Table 6 The seven criteria were chosen to assess viticultural system performance Performance Economic performance Environmental performance

Social performance

Criteria CR1 (REN) Profitability of the winegrowing system (quantitative, increasing) CR2 (PPS) Pesticide pressure (quantitative, decreasing) CR3 (IRE) Risk of ecotoxicity (quantitative, decreasing) CR4 (PAE) Agroecological practices (qualitative, increasing) CR5 (PUL) Pesticide spray drift (qualitative, increasing) CR6 (TRA) Workload (quantitative, decreasing) CR7 (SYS) System complexity (qualitative, decreasing)

144

F. Macary

Table 7 Revenues determined according to winegrowing practices Ref 1 References of products according to environmental practices Full disbudding or unwanted shoots removal Good distribution of grapes Good vigour At least one treatment against Botrytis per year Grass cover on one row or soil tillage Grass cover at all the rows Gross total revenue: € / ha

Ref 4

Ref 5

Ref 6

Organic wine X

Ref 2 Ref 3 Conventional wines Red Red wine wine Ruby Garnet X X

Red wine Ocher X

Red wine Brick X

White wine

X X

X X

X

X

X

5400

7000

X X

(X) (X) 8000

(X) 7800

(X) 6300

(X) 5700

CR3 (IRE) Risk of ecotoxicity—A quantitative criterion This criterion was developed and calculated by researchers of the Mediterranean Agronomic Institute of Montpellier (CIHEAM IAM), France (Mghirbi et al. 2014). IRTE evaluates the toxicity of pesticides for non-target living organisms (i.e. terrestrial invertebrates, birds, and aquatic organisms) and considers the physicochemical proprieties of molecules (i.e. mobility, persistence in the soil, and bioaccumulation). We sent CIHEAM IAM a list of pesticides used in each of our viticultural systems, and they calculated the IRTE indicators, using the chemical properties of pesticides and applied doses. CR4 (PAE) Agroecological practices—A qualitative criterion This criterion contributes to preserving biodiversity and reducing the use of chemicals (pesticides and fertilisers). The present study focused on the following practices already adopted by some winegrowers: • Grass cover in the inter-row: natural/sowed, total/partial • Vine row management: chemical/mechanical weeding • Agroecological features (AS): grass strips, flowering strips, hedges, insect hotels, nest boxes • Use of biocontrol agents comprising natural (plant, animal, and mineral) substances used for plant protection. We used the official list of biocontrol agents published by the French Ministry of Agriculture, Food and Forestry. CR5 (PUL) Pesticide spray drift—A qualitative criterion Pesticide spray drift was evaluated by classifying the spray equipment used in the vineyard according to its capacity for reducing pesticide losses to the environment. This was based on a study carried out by the French Institute of Vine and Wine, which assessed the performance of different types of spray equipment (Table 8).

Multicriteria Decision Aiding for a Shift Towards Best. . .

145

Table 8 Classification of the spray equipment regarding drift control and notation of performances Spray quality for drift limitation Very high High Medium Low Very low

Spraying devices Confined sprayer (mostly) Face-to-face spraying Air blower (mostly/ other better equipment) Air blower Air blast sprayer

Score 8 6 4 2 1

Table 9 Assessment of the complexity of the implementation of the wine system and notation of performances Number of operations [20–25]

[25–30]

[30–35]

[35–40]

[40–45]

Distance Around the farm d < 5 km d > 5 km Around the farm d < 5 km d > 5 km Around the farm d < 5 km d > 5 km Around the farm d < 5 km d > 5 km Around the farm d < 5 km d > 5 km

Score 1 4 7 9 12 15 17 20 23 25 28 31 33 36 39

CR6 (TRA) Workload—A quantitative criterion Workload was calculated for each technical operation (mechanical and manual) that we registered during the survey as below: TRA = ½number of hoursðmechanical operationsÞ × 6:4 number of hourðmanual operationsÞ *Ratio between the number of hours for manual operations and mechanical operations CR7 (SYS) System complexity—A qualitative criterion The evaluation of System complexity is the result of a combinatorial analysis that considers the number of mechanical and manual operations, as well as the distance from the parcels to the main vineyard buildings (Table 9). Once each of the 38 systems had been evaluated, for each of the seven criteria the values were entered into the performance matrix for the implementation of the multicriteria model (Table 10, extract).

146

F. Macary

Table 10 Extract of the 38 viticultural systems (VS) performance matrix Alternatives VS04 VS05 VS07 VS08 VS09 VS10

1

REN 2196 2048 2013 2329 3180 2157

PPS 16 16 13 17 9 16

IRE 4553 3424 3151 4690 6980 4072

PAE 3 1 1 1 46 1

PUL 1 6 8 8 1 1

TRA 243 294 272 251 320 256

SYS 23 31 31 20 20 15

Associang a ‘playing card’ to each criterion TRA

PAE

REN

PPS

SYS

IRE

PUL

Ranking the criteria from the most important one to the less important

2

REN

PPS

IRE

PAE

PUL

TRA

SYS

Inserng white cards ( ): Degree of importance of the criterion regarding to the following one

3

CR 1. REN

4

CR 2. PPS

CR 3. IRE

CR 4. PAE

CR 5. PUL

CR 6. TRA

CR 7. SYS

Determinaon of the rao given between the dominant criterion and the weakest: Z= 3 Æ CR1 / CR7 =3

Fig. 12 Implementation of SRF Software for weighting criteria of evaluation

4.2.2

Parameters of the Model

A criteria weighting process was conducted collaboratively by four winegrowers (two conventional and two organic), three advisors in winegrowing practices, and six researchers (agronomists and economists from INRAE and Bordeaux Science Agro), using the SRF (Simos, Roy, and Figueira) method (Figueira and Roy 2002) involving a set of cards. The procedure according to Simos is summarised in Fig. 12. The SRF software gives the weight of each criterion expressed as a percentage of the total weight normalised to one (Table 11). For each criterion, Table 9 shows the preference direction, the criteria weights, the thresholds of Indifference and Preference, which were retained considering the scale of the performance values, and the thresholds of Veto (Table 9). This model, with criteria and parameters (weights and thresholds) was used two times, the first to apply the ELECTRE TRI-C method to the 38 real systems and 3 scenarios and the second to apply the ELECTRE III method into each category of performance (Fig. 11). For ELECTRE TRI-C, the model also included the categories

Multicriteria Decision Aiding for a Shift Towards Best. . .

147

Table 11. Preference Threshold and Indifference Threshold are assigned to each criterion Weights Threshold indifference Threshold preference Criterion direction Nature of the threshold

REN 22 0.05 0.1 MAX Variable

PPS 20 0.03 0.07 MIN Variable

IRE 15 0.05 0.1 MIN Variable

PAE 13 1 1.9 MAX Fixed

PUL 13 1.9 1.9 MAX Fixed

TRA 10 0.025 0.05 MIN Variable

SYS 7 1 2.9 MIN Fixed

PUL 8 6 4 1

TRA 230 250 270 290

SYS 12 20 28 36

Table 12 Reference values of each category for all the criteria Performance categories C4 very high C3 high C2 medium C1 low

REN 3300 2700 2100 1500

PPS 10,5 13 15,5 17

IRE 3500 4500 5500 6500

PAE 53 38 23 8

Table 13. Assignment of the viticultural systems to the four categories Categories C4 C3

Performances Very high High

C2

Medium

C1

Low

Viticultural systems VS59° VS07, VS09°, VS38°, VS39°, VS40°, VS42°, VS54°, VS55°, VS56, VS61, VS62, VS67* VS04, VS05, VS08, VS10, VS11, VS18, VS23, VS31, VS32, VS33*, VS36, VS50, VS51, VS52, VS53, VS57, VS58, VS60, VS63, VS64, VS65, VS66, VS68 VS22, VS34

Numbers and % 1 (2, 6 %) 12 (31, 6 %) 23 (60, 6%) 2 (5, 2 %)

Conventional VS/Environmental certification VS° / Organic VS*

to which the method assigned the systems. The reference values associated with the four categories by criterion are indicated in Table 12.

4.2.3

Applications of the Methods and Results of Step 2

The ELECTRE TRI–C method assigned each viticultural system to one of the four predefined performance categories. Table 13 shows the assignments of the 38 viticultural systems. It can be seen in this table that the organic winegrowing systems, obtained the best results, in particular the agroecological system VS59. VS67 and VS33 were the only VS with an environmental certification, all the other systems were conventional. It was thus possible for us to retain the best practices in the field and, in collaboration with the winegrowers and their advisors, to develop scenarios, which evolve towards better sustainable performances (Step 3). The production models (conventional, organic, or agroecological systems) were only highlighted after obtaining results to avoid any bias in the modelling process.

148

4.2.4

F. Macary

Assessment of the New Realistic Scenarios of Production Systems and Results of Step 3

The three realistic scenarios developed in Step 3 (see Sect. 3.3) could be inserted into the model for the application of the ELECTRE TRI-C method without changing the results of assignment of the 38 systems already evaluated. The method ELECTRE III was applied to rank the elements assigned to the same category in Step 2 without scenarios and in Step 3 with scenarios, the last step. Table 14 shows the results obtained once the winegrowing systems and the scenarios have been assigned to one of the four categories, and then after they have been ranked within each category according to performance (high to low). The three simulated systems (SC1, SC2, and SC3) were assigned to the category “very high performance” (SC3 and SC2) and to the category “high performance” (SC1), in modelling. The scenario of agroecological organic system SC3 resulted the most efficient. This result can be explained by the difference in the economic performance between SC3, which had the highest winegrowing margin per hectare, and the conventional systems (current systems, SC1 and SC2). Compared to the organic systems—whose margin was calculated using the same total revenue reference—SC3 slightly reduces costs. Furthermore, SC3 improves general environmental performance by reducing pesticide pressure (PPS represents 20% of the

Table 14. Results of the sorting and ranking of the 38 existing viticultural systems and three scenarios C4

Performances Very high

C3

High

C2

Medium

C1

Low

Viticultural systems SC3 SC2 VS59° [SC1, VS42°] [VS09°, VS38°, VS62] [VS40°, VS56, VS61, VS67*] [VS07, VS55°] [VS54°, VS39°] VS36 VS63 [VS32, VS58, VS65] [VS33*, VS50, VS53] [VS10, VS60, VS66, VS68] [VS05, VS52] VS04 VS11 [VS08, VS31] [VS18, VS57] VS51 [VS23, VS64] VS22 VS34

SC scenarios /Conventional VS/ Environmental certification VS* / Organic VS°

Numbers and % 3 (7, 3%)

13(31,7%)

23 (56,0%)

2 (5,0%)

Multicriteria Decision Aiding for a Shift Towards Best. . .

149

weight). In this system, the pesticide drift was also reduced due to the use of an efficient sprayer, and agroecological practices are maximised. The agroecological system (SC2) was ranked second in the category “very high performance”. In particular, this system reduces costs, especially those related to pesticide use. Out of the three simulated systems (SC1, SC2, and SC3), the lowest economical margin was obtained by the maximised conventional system (SC1), because of the total revenue difference and high total costs, such as pesticides and energy. In the study area, the vineyard “Domaine Emile Grelier” (VS59) is a good example of a successful winegrowing system in which a holistic agroecological approach is applied, and which was assigned by the MCDA to the “very high performance” category.

5 Main Difficulties and Some Recommendations for this Kind of Complex Project This type of complex multicriteria project involves certain difficulties linked to the subject of the study itself and the used methodological approach. The experience that arose from this project can suggest some recommendations.

5.1

The Subject of Pesticides as a Very Sensitive Area

The media constantly relays increasingly negative consumer opinions about the use of chemical pesticides, and there is now proof of its negative effects on human health, as well as on ecosystems. Organic farmers can respond more easily to the current demands of society regarding pesticides; however, even plant-based insecticides, like pyrethrums, can be very toxic. Conventional farmers use chemical pesticides, which can be carcinogenic, mutagenic and reprotoxic (CMR) or even endocrine disrupters (ED); they generally do not wish to discuss this, nor their practices. It is, therefore, clear that starting a research programme that deals with the subject of pesticide use are very delicate, in viticulture (and, even more so, in arboriculture) that needs to be handled carefully when interacting with all participating parties. Over the months, a relationship of trust was progressively built with the involved winegrowers and their advisors, as well as with the different winegrowing bodies. Such relationships are always fragile and can be affected by the slightest problem in the field. The producers feel like they are being scrutinised and criticised by the public. Whilst some of them—still a minority—have decided to change their production methods and adopt more sustainable practices, conventional farmers still take refuge behind an arsenal of chemicals. This has an impact in terms of

150

F. Macary

governance when implementing any project on this topic. We must be careful not to offend professionals such as conventional winegrowers who were shown as an example in the past. Now they are being fingered for their practices that are not good for the quality of the environment and human well-being. We must not oppose each other in the constructive phase of the project. On the other hand, in the results, as here, after presenting the methodology in a neutral way with respect to the various stakeholders, it is then necessary to explain the results related to the different practices of each, and a different environmental approach. The success of this project heavily depended on the quality of relations between the project partners and the socioeconomic partners on the one hand, and on the mutual respect established between all project participants (60 spread over 10 research units in this case) on the other. It was important to foster and maintain good relations and trust throughout a project and to respond to all queries and requests by the stakeholders on the ground. By respecting these rules—which were not always easy to comply with for certain members of the team—it was possible to fulfil the conditions necessary for the execution of the project. When coordinating the project, we therefore ensured that the professionals concerned were notified before any of the 10 project teams and 60 scientists, engineers, and technicians carried out an intervention on the site. A good relationship of mutual trust between all participants was essential for the success of the project. In light of this, meetings were held regularly over the 5 years in order to give an update on the project results and developments, and to hold discussions between all the participants (winegrowers, advisors, and managers from regional minister of agriculture and environmental managers of water quality). The choice of venue for such a meeting also plays a role in the establishment of good relations—it should be located on public and neutral ground; in this case, the community centre located in the experimental watershed was the ideal venue for holding meetings with winegrowers from cooperatives or independent vineyards. At the start of such a project, a coordinator needs to clearly explain the nature of the project to the professional stakeholders, as well as the importance of preserving the anonymity of the results of the field surveys. An initial public project information meeting must precede all team intervention in the field. It is also important to explain the project not only to the farmers themselves, but also to their advisors and related organisations, such as cooperatives, unions, and retailers. An article published in a local and relevant newsletter can reach those who never attend public meetings. Before any field surveys are carried out, the participants must be provided with an explanation of the project objectives and steps and be given a written summary of the project. All such stakeholders must have a sense of ownership of the project from the start in order for them, to take ownership of the results. In this project, the surveys were conducted on a voluntary basis, but essential support was provided by the local cooperative of winegrowers, that went as far as to make appointments with its members for the surveys! With this local support, it was therefore possible to contact all the winegrowers in the area. This was also the case for the independent winegrowers thanks to the connections with the winegrowing union in the project. It is worth noting that the general interaction with the main

Multicriteria Decision Aiding for a Shift Towards Best. . .

151

socio-professional structures to which the surveyed stakeholders belonged was key to the success of the project. The criteria were constructed according to the requirements of the issue at hand and not defined in advance, regardless of what these requirements were. All the criteria described in Sect. 4.2.1 were created for this study, apart from criterion Pesticide pressure, for which we only calculated a treatment frequency index at winegrowing system level, and criterion Pesticide ecotoxicity, calculated by colleagues at the Mediterranean Agronomic Institute of Montpellier from data we sent them. This model was the result of collaboration with the stakeholders, and before using the criteria, we explained our scoring grids to them, and discussed and even modified the grids with them. The close collaboration with the professional winegrowers over the 5 years of the project was also beneficial for weighting the performance criteria of the winegrowing systems, as well as for explaining the choices made when developing the new scenarios for a shift to sustainable viticulture. The scenarios were inspired by practices, which had been observed in this region, but never in the same vineyard all at once. At the local meeting for the winegrowers and the final conference, the agroecological winegrower (there was only one) was first asked to present his practices, which were the source of inspiration for the scenarios. This was a way of validating our methodology, which—along with the results—was particularly appreciated by the professionals who were present.

5.2

Presentation of Results and Modelling Assumptions

The presentation of the results is always important but in these delicate contexts, it may be problematic. The use of maps may facilitate any presentation and some details become essential. At the end of Step 1, in which the agri-environmental risks in the study area were evaluated and mapped, a first meeting was necessary to present the results (Aouadi et al. 2018). When presenting maps, it is important to use a colour gradient that is appropriate for the stakeholders concerned. In this case, shades of brown were used, and red was avoided, so that the winegrowers with highrisk vineyards did not feel as if they are under attack, especially during feedback sessions. Above all, it was important to highlight the levels of risk in each category and show that it could be possible to reduce the risk of high to very high-risk areas by adjusting especially plant protection measures and soil management practices. This was done by using maps to simulate the reduction of plant protection inputs and the modification of soil management by mechanical weeding. Other elements that must be well presented, this time before carrying out the multicriteria analysis, are the hypotheses and fundamental factors underlying the study, given their impact on decision aiding (Aouadi et al. 2021). Some assumptions that were presented and discussed during the study were that all the winegrowers were considered to obtain maximal yield (grapes/ha) as laid down by the registered designation of origin (AOC) for a normal climate year. Moreover, the economic

152

F. Macary

calculations were standardised for all the systems; i.e. we used economic references based on the cultivation techniques that influence grape quality and therefore the wine products. Type of wine marketing was not considered (retail, direct sales, bottle, bulk sales, etc.). We were interested in establishing a differential margin to be able to compare the winegrowing systems. During our presentations, they followed the application of both methods ELECTRE well for Steps 2 and 3. They were confident with the methodology approach because some of them were associated, and the related results. They did not raise any specific questions. The organic winegrower, from which our agroecological model proposed was developed, took active part in the conference. This added credibility to the work that was carried out. One may think that the conventional winegrowers were not very comfortable with their intensive practices. They did not take part in the discussion and their agroecology colleague showed that our scenarios are credible. The practices of organic winegrowers are already quite close to the scenarios. Therefore, these producers found this quite obvious. We think that it is essential to build the methodology and the model step by step with the stakeholders; because they can thus better appropriate the bottom-up approach, versus a top-down method.

6 Conclusion The realistic scenarios that we developed—with practices altered for each cultivation method (conventional or organic) showed that any production systems, when influenced by these scenarios, can be economically viable and can significantly reduce the use of pesticides and their theoretical ecotoxicity. The analysis showed that it is possible to improve each method of cultivation to improve the results by i) adopting the right combination of winegrowing practices, ii) taking into account the economic and technical constraints, and iii) choosing the adequate molecules and treatments to be used, as well as the appropriate equipment. The proposed scenarios were based on existing practices and could be successfully adopted by winegrowers in different winegrowing areas. To shift to more sustainable agriculture with less impact on the environment and human health, it is important to develop and assess new systems, which integrate the principles of agroecology. However, to change same practices, regular expert advice and monitoring would be necessary to guide vineyards in their transition towards agroecological methods. Furthermore, winegrowers would need to carry out regular observations in the vineyard to assess, for example pest pressure and biological regulation. The winegrowers and their advisors were included in the modelling process regarding the choice of criteria and the weighting of each one. They have been also associated with the practices in the three proposed and evaluated scenarios. This close collaboration was developed during the 5 years of the project and was a prerequisite for the credibility of the work and its results. The only winegrower to

Multicriteria Decision Aiding for a Shift Towards Best. . .

153

apply holistic and agroecological methods in his vineyard in the field provided information. He explained that was important for understanding how changing attitudes towards current conventional practices in terms of vineyard operations and, especially, plant health protection. This work and its results were presented to the winegrowers and advisors in the community centre in our experimental area, as well as to a wider public during the final conference of the project at the high agronomic school of Bordeaux. The audience could clearly perceive the nature of the suggested changes in winegrowing practices and the feasibility of a shift to agroecological systems. They were also reassured by the changes already made in the field by pioneers of agroecology. The multicriteria decision aid models and methods were revealed to be very suitable for this type of study, particularly the ELECTRE methods, which can be used to combine the quantitative and qualitative criteria. By using these methods in our project, we were able to obtain very useful results and the winegrowers found them to be reliable and able to provide a sound basis for discussion about the shift to agroecological systems. Furthermore, the methods contributed to alleviating any apprehension and doubt with respect to this new way of thinking and to winegrowing practices that will be more respectful of ecosystems and human health. We have been using the ELECTRE methods for evaluating agri-environmental risks for about 15 years, and we are therefore very familiar with their subtleties. Coming to grips with ELECTRE methods requires a little experience, especially for modelling the model parameters. It is therefore recommended for a new user to be accompanied by an experienced user at the beginning.

Appendix The Principle of Multicriteria Methods by Outranking These methods fall into the category of discrete and outranking methods. Discrete methods are used for decisions represented by a finite number of potential alternatives. Their main aim is to provide a basis for comparison of these potential alternatives according to several criteria. Outranking methods are used to help resolve a decision problem, considering any indecisiveness and imprecision, as well as the fact that a dominating relationship between two alternatives does not always exist; for example in Step 1 of our study, when the comparison of two parcels is not possible, because the level of risk is too similar. It should of course be acknowledged that two alternatives can be considered as equal and may even not be differentiated by the model; in that case, the operator would need to give the final result. The particularity of these methods is that they are based on the notion of outranking associated with the family of European methods comprising, for example ELECTRE (Roy 1968, 1978, 1985, 1990; Maystre et al. 1994; Schärlig 1996). These contrast with US models that do not consider

154

F. Macary

Fig. 13 The general principle of the outranking methods ELECTRE

indecision, incomparability, and a low preference for one alternative over another. These conditions do not reflect the reality in the field, whether in terms of purely physical approaches, like in Step 1, or of those which also integrate socio-economic criteria, as in Steps 2 and 3. In an outranking relationship, the criteria are aggregated according to a partially binary relationship; i.e. between two alternatives (in this case, agricultural parcels and then winegrowing parcels) taken two at a time per criterion and then across all criteria. The basic rule is that an action outranks another if it is at least as good for a majority of criteria, without being clearly worse than the other alternative for all criteria (Roy 1990). Outranking models exploit the available information, with any imprecision and shortcomings; this is particularly useful for our agri-environmental approach. The outranking hypothesis is always associated with the notions of concordance and discordance. The outranking relationship integrates the fuzzy relationship into the ELECTRE models, because whilst the relationship for some pairs of alternatives may seem indisputable, it can seem unconvincing for others (Roy 1978). Depending on the pair, this plausibility is expressed by an index that represents the level of credibility of the outranking relationship. It is no longer useful to classify the alternatives in one of the three outranking groups: indifference, low preference, or strict preference (Fig. 13); all intermediary positions between the two extremes are possible. Whether to accept or reject the whole outranking hypothesis is no longer at play; instead the level of its credibility needs to be determined, based on a score of 0–1 (Maystre et al. 1994). By using these methods, real quantitative criteria (e.g. parcel slope and pesticide pressure), as well as qualitative pseudo-criteria (e.g. link between the parcel and watercourse, soil characteristics, quality of the buffer strips and the riverine forests, and agri-ecological practices) can all be simultaneously integrated into the model. Each of the seven criteria is given an indifference threshold, q, and a strict preference threshold, p. These thresholds have been defined so as to take direct account of the uncertainty of the values of the evaluation matrix and are based on the fact that these values are either badly defined or known to have a margin of error (Vallée, Zielniewicz 1994). The thresholds allow the notion of low preference to be developed. Thus, the number of possible situations at the end of a comparison of two

Multicriteria Decision Aiding for a Shift Towards Best. . .

155

alternatives, a and b, according to a given criterion, goes from 3 (indifference, a preferred over b, b preferred over a) to 5 (adding the notion of low preference) (Fig. 13). To better understand these notions, which are fundamental to the implementation of the model, they are described below in more detail: • The indifference threshold represents the difference in performance from which point two alternatives are no longer indifferent. • The strict preference threshold expresses the difference in performance from which point one alternative is clearly preferred over the other. • A third threshold, the veto threshold, is used for the realisation of the notion of discordance; i.e. it represents the difference when alternative b is considered much better than alternative a for a criterion j, and in no case can a be generally considered better than b, no matter the performance of a and b for all other criteria. This useful notion is illustrated in Step 1, in which the aim is to group the different parcels in the watershed according to level of risk of pesticides contaminating the surface water. All the different criteria are considered, but the major one here is the dose applied to the parcel. This is useful for meadows, for example which are located at the edge of watercourse because they consist of very wet soils, they are unsuitable for growing vines on. They are in the high vulnerability category with zero pesticide pressure. The slightest vulnerability would put them in a high-risk category; however, in reality, without any pesticide addition, the risk is zero. To take account of this, a veto threshold was defined for the pesticide pressure criterion, which prevents the meadow from being raised from a zero-risk category to a higher category. The thresholds are considered as affine functions of performances g j (a) and are calculated as follows: Threshold g j (a) = α x g j (a) + β. The analyst must define the value of both the α and β coefficients per criterion and for each threshold. These can be calculated according to the worst or the best performances of a and of b; for the former the method for calculating the threshold is direct and for the latter it is indirect. These thresholds must be based on the performance value of each alternative per criterion (in the performance matrix); they are thus specific to the project in question. The method thus relies somewhat on the experience of an expert operator to help a beginner implement the model. Figure 13 illustrates the principle of outranking based on indifference, preference, and veto thresholds

References Almeida-Dias J, Rui-Figueira J, Roy B (2010) ELECTRE TRI-C: a multiple criteria sorting method based on characteristic reference actions. Eur J Opl Res 204:565–580. https://doi.org/10.1016/j. ejor.2009.10.018

156

F. Macary

Altieri MA (1995) Agroecology: the science of sustainable agriculture, 2nd edn. Westview Press, Oxford Angevin F, Fortino G, Bockstaller C, Pelzer E, Messean A (2017) Assessing the sustainability of crop production systems: toward a common framework? Crop Prot 97:18–27. https://doi.org/10. 1016/j.cropro.2016.11.018 Aouadi N, Macary F, Alonso Ugaglia A (2019) Evaluation des performances socio-économiques et environnementales de systèmes viticoles en bordelais par une méthode multicritère d’aide à la décision, In : 13e Journées de recherches en sciences sociales (JRSS), 12–13 décembre 2019, Bordeaux, Bordeaux, Société Française d'Economie Rurale, 13. https://www.sfer.asso.fr/source/ jrss2019/articles/B43_Macary.pdf. Accessed 21 Dec 2022 Aouadi N, Macary F, Alonso Ugaglia A (2020) Évaluation multicritère des performances socioéconomiques et environnementales de systèmes viticoles et de scénarios de transition agroécologique. Cah Agric 29(19):19. https://doi.org/10.1051/cagri/2020016 Aouadi N, Macary F, Deliere L, Roby JP (2021) New scenarios for a shift towards agroecology in viticulture. Agric Sci 12:1003–1033. https://doi.org/10.4236/as.2021.1210065 Aouadi N, Macary F, Gaillard J, Devier MH, Budzinski H (2018) Evaluation des risques de contamination des eaux de surface par les produits phytosanitaires. Application à un bassin versant viticole (projet PhytoCOTE), In: Contaminations, Environnement, Santé et Société. De l'évaluation des risques à l'action publique. Colloque international, 4-6 juillet 2018, Toulouse, CNRS, 12. https://hal.archives-ouvertes.fr/hal-01890946/document. Accessed 21 Dec 2022 Baldi I, Lebailly P, Rondeau V, Bouchard V, Blanc-Lapierre A, Bouvier G, Canal-Raffin M, Garrigou A (2012) Levels and determinants of pesticide exposure in operators involved in treatment of vineyards: results of the PESTEXPO Study. J Expo Sci Env Epid 22(6):593–600. https://doi.org/10.1038/jes.2012.82 Barriuso E (2004) Estimation des risques environnementaux des pesticides, INRA éditions, Paris. https://hal.archives-ouvertes.fr/hal-01192112/document. Accessed 21 Dec 2022 Bengtsson J, Ahnström J, Xweibull AC (2005) The effects of organic agriculture on biodiversity and abundance: a meta-analysis. J Appl Ecol 42(2):261–269. https://doi.org/10.1111/j. 1365-2664.2005.01005.x Benkirane I (2015) Contribution à la connaissance et au raisonnement des pratiques de protection sanitaire du vignoble dans le Bordelais (Bassin versant expérimental de Marcillac), Mémoire 2è année études, Agronomy, Bordeaux Sciences Agro, Bordeaux, 41 Berendse F, Chamberlain D, Kleijn D, Schekkerman H (2004) Declining biodiversity in agricultural landscapes and the effectiveness of Agri-environment schemes. AMBIO: A J of the human environment 33(8):499–502. https://doi.org/10.1579/0044-7447-33.8.499 Bockstaller C, Girardin P, Van Der Werf HMG (1997) Use of agro-ecological indicators for the evaluation of farming systems. Eur J Agron 7(1–3):261–270. https://doi.org/10.1016/S11610301(97)00041-5 Bohnen NI, Kurland LT (1995) Brain tumor and exposure to pesticides in humans: a review of the epidemiologic data. J Neurol Sci 132(2):110–121. https://doi.org/10.1016/0022-510X(95) 00151-Q Boithias L, Sauvage S, Srinivasan R, Macary F, Sanchez-Perez JM (2012) Assessment of pesticide application date shifts on surface water contamination during floods using SWAT model in the Save catchment (South-Western France), In : 2012 International SWAT Conference. New Delhi, India, 51 Chaumet B, Morin S, Boutry S, Mazzella N (2019) Diuron sorption isotherms in freshwater biofilms. Sci Total Environ 651(1):1219–1225. https://doi.org/10.1016/j.scitotenv.2018.09.286 Eastman JR (1988) IDRISI: a grid-based geographic analysis system, clark university graduate school of geography, worcester. http://books.google.fr/books?id=GsSbtwAACAAJ. Accessed 21 Dec 2022 European Commission (2002) Building agro Environnemental indicators. Focussing on the european area frame survey LUCAS, European Commission, ISPRA. https://publications.jrc.ec. europa.eu/repository/handle/JRC24247. Accessed 21 Dec 2022

Multicriteria Decision Aiding for a Shift Towards Best. . .

157

FAO (2018) FAO'S work on agroecology. A pathway to achieving the SDGs. Food and agriculture organization of the United nations, Rome, 28. https://www.fao.org/3/i9021en/i9021en.pdf. Accessed 21 Dec 2022 Figueira J, Roy B (2002) Determining the weights of criteria in the ELECTRE type methods with a revised Simos procedure. Eur J Opl Res 139:317–326. https://doi.org/10.1016/S0377-2217(01) 00370-8 Girardin P, Bockstaller C, Van Der Werf H (2000) Assessment of potential impacts of agricultural practices on the environment: the AGRO*ECO method. Environ Impact Asses 20:227–239. https://doi.org/10.1016/S0195-9255(99)00036-0 Gliessman SR (1998) Agroecology: ecological processes in sustainable agriculture. Chelsea, Ann Arbor Gliessman SR (2015) Agroecology: the ecology of sustainable food systems. Boca Raton, London Granoulis J (2009) Risk analysis of water pollution, 2nd edn WILEY-VCH Weinheim Guillou M, Guyomard H, Huyghe C, Peyraud JL (2013) Vers des agricultures doublement performantes pour concilier compétitivité et respect de l'environnement. Agreenium, INRA, Paris, 163. https://www.inrae.fr/sites/default/files/pdf/rapport-inra-pour-cgsp-volume-2sommaireweb07102013.pdf. Accesssed 21 Dec 2022 Hatt S, Artu S, Brédart D, Lassois L, Francis F, Haubruge E, Garré S, Stassart PM, Dufrêne M, Monty A, Boeraeve F (2016) Towards sustainable food systems: the concept of agroecology and how it questions current research practices. A review. Biotechnologie, Agronomie, Société et Environnement BASE [En ligne] 20(S1):215–224. https://doi.org/10.25518/1780-4507.12997 Hildebrandt A, Guillamón M, Lacorte S, Tauler R, Barceló D (2008) Impact of pesticides used in agriculture and vineyards to surface and groundwater quality (North Spain). Water Res 42(13): 3315–3326. https://doi.org/10.1016/j.watres.2008.04.009 INAO, IFV (2017) Guide de l'agroécologie en viticulture. Développement des mesures agroenvironnementales, Paris, 52. https://www.inao.gouv.fr/Archives-des-actualites-de-lINAO/Presentation-du-Guide-de-l-agroecologie-en-viticulture. Accessed 21 Dec 202 Inserm (2021) Pesticides et effets sur la santé. Nouvelles données. Synthèse de l'expertise collective. INSERM, Montrouge, 164. https://www.inserm.fr/sites/default/files/2021-06/Inserm_ ExpertiseCollective_Pesticides 2021_Synthese.pdf. Accessed 21 Dec 2022 Leonard RA (1990) Movement of pesticides into surface waters. In: Cheng HH (ed) Pesticides in the soil environment: processes, impacts, and modeling. Soil Science Society of America, Madison, pp 303–349 Maaf (2012a) Agricultures, produisons autrement : projet agro-écologique pour la France. In: Ministère en charge de l'agriculture. https://agriculture.gouv.fr/le-projet-agro-ecologique-enfrance. Accessed 21 Dec 2022 Maaf (2012b) Bilan du plan ECOPHYTO:2 Maaf (2014) 12 clés pour comprendre l'agroécologie. In: Ministère de l'agriculture, de l'agroalimentaire et de la forêt, Paris. https://agriculture.gouv.fr/le-projet-agro-ecologique-en12-cles. Accessed 21 Dec 2022 Macary F, Devier MH (2014) PhytoCOTE: using pesticides in agrosystems, transfers, bioaccumulation, and impacts on ecosystems Modeling of changes in farming practices. Université de Bordeaux 1. Ecole doctorale des Sciences et Environnements, Bordeaux, 33. https://cote.labex.u-bordeaux.fr/Projet/Volet-Recherche/Projets-bottom-up/PhytoCOTE-i5898. html. Accessed 21 Dec 2022 Macary F, Ombredane D, Uny D (2010) A multicriteria spatial analysis of erosion risk into small watersheds in the low Normandy Bocage (France) by ELECTRE III method coupled with a GIS. Int J Multicriteria Decision Making 1(1):25–48. https://doi.org/10.1504/IJMCDM.2010.033685 Macary F, Guerendel F, Alonso Ugaglia A (2020) What contributions from the literature to understand and build the agroecological transition in viticulture? Cah Agric 29:38. https://doi. org/10.1051/cagri/2020035 Macary F, Almeida-Dias J, Uny D, Probst A (2013a) Assessment of the effects of best environmental practices on reducing pesticide pollution in surface water, using multi-criteria modelling

158

F. Macary

combined with a GIS. Int J Multicriteria Decision Making 3(2–3):178–211. https://doi.org/10. 1504/IJMCDM.2013.053725 Macary F, Morin S, Probst JL, Saudubray F (2014a) A multi-scale method to assess pesticide contamination risks in agricultural watersheds. Ecol Indic 36(0):624–639. https://doi.org/10. 1016/j.ecolind.2013.09.001 Macary F, Almeida Dias J, Rui-Figueira J, Roy B (2014b) A multiple criteria decision analysis model based on ELECTRE TRI-C for erosion risk assessment in agricultural areas. Environ Model Assess 19:221–242. https://doi.org/10.1007/s10666-013-9387-x Macary F, Leccia O, Almeida Dias J, Morin S, Sanchez-Pérez JM (2013b) Agro-environmental risk evaluation by a spatialised multi-criteria modelling combined with the PIXAL method. Rev Int Géomatique 23(1):39–70. https://doi.org/10.3166/rig.23.39-70 Madelrieux S, Alavoine-Mornas F (2012) Withdrawal from organic farming in France. Agron Sustain Dev 33(3):457–468. https://doi.org/10.1007/s13593-012-0123-8 Martin C, Legret M (2005) La méthode multicritère ELECTRE III. Définitions, principe et exemple d'application à la gestion des eaux pluviales en milieu urbain. Bulletin des Laboratoires des Ponts et Chaussées 258-259: 29-46. https://www.ifsttar.fr/collections/BLPCpdfs/blpc__258-2 59_29-46.pdf. Accessed 21 Dec 2022 MASA (2022) Indicateur de Fréquence de Traitements phytosanitaires (IFT). https://agriculture. gouv.fr/indicateur-de-frequence-de-traitements-phytosanitaires-ift. Accessed 21 Dec 2022 Maystre LY, Pictet J, Simos J (1994) Méthodes multicritères ELECTRE :"description,conseils pratiques et cas d'application à la gestion environnementale". Lavoisier,TEC § DOC, Paris, Gérer l'environnement Méndez VE, Bacon CM, Cohen R, Gliessman SR (2016) Agroecology: a transdisciplinary, participatory and action-oriented approach. Boca Raton, London Merot A, Wery J (2017) Converting to organic viticulture increases cropping system structure and management. Agron Sustain Dev 37:19. https://doi.org/10.1007/s13593-017-0427-9 Métral R, Rapidel B, Delière L, Petitgenet M, Lafond D, Chevrier C, Bernard FM, Serrano E, Thiolet-Scholtus M, Wery J (2015) A prototyping method for the re-design of intensive perennial systems: the case of vineyards in France. In: 5th International Symposium for Farming Systems Design, Montpellier, 7-10 septembre 2015, p 2 Mghirbi O, Ellefi K, Le Grusse P, Mandart E, Fabre J, Ayadi H, Bord JP (2014) Assessing plant protection practices using pressure indicator and toxicity risk indicators: analysis of therelationship between these indicators for improved risk management, application in viticulture. Environ Sci Pollut R 22:8058–8074. https://doi.org/10.1007/s11356-014-3736-4 Mousseau V, Figueira J, Naux JP (2001) Using assignment examples to infer weights for ELECTRE TRI method: some experimental results. Eur J Opl Res 130(2):263–275. https:// doi.org/10.1016/S0377-2217(00)00041-2 Payraudeau S, Van Der Werf HMG (2005) Environmental impact assessment for a farming region : a review of methods. Agric Ecosyst Environ 107:1–19. https://doi.org/10.1016/j.agee.2004. 12.012 Pelsy F, Merdinoglu D (2021) La vigne, miracle de la nature ? 70 clés pour comprendre la viticulture, Quae, Versailles, Clés pour comprendre. https://www.quae.com/produit/1695/9782 759233328/la-vigne-miracle-de-la-nature?utm_source=dialoginsight&utm_medium=email& utm_campaign=2021.11%20-%20SPE%20-%20Catalogue%20cadeaux%20PR&oft_id=1194 747&oft_k=pBf9Ccva&oft_lk=bEU9hC&oft_d=637725603077600000. Accessed 21 Dec 2022 Pupier J (2015) Comment réduire l’emploi des produits phytosanitaires en viticulture sans pénaliser l’activité économique viticole dans le bassin versant de la Livenne à Marcillac? Agronomy, Bordeaux Sciences Agro, Bordeaux, p 19 Renaud-Gentié C, Burgos S, Benoît M (2014) Choosing the most representative technical management routes within diverse management practices: application to vineyards in the Loire Valley for environmental and quality assessment. Eur J Agron 56:19–36. https://doi.org/10.1016/j.eja. 2014.03.002

Multicriteria Decision Aiding for a Shift Towards Best. . .

159

Roubeix V, Mazzella N, Delmas F, Coste M (2010) In situ evaluation of herbicide effects on the composition of river periphytic diatom communities in a region of intensive agriculture. Atmusephere: Int Online Mus Jl 60(2):233–241. https://hal.archives-ouvertes.fr/hal-02167842. Accessed 21 Dec 2022 Roy B (1968) Classement et choix en présence de points de vue multiples (la méthode Electre). Rev Fr Automat Infor 2(8):57–75 Roy B (1978) Electre III: un algorithme de classements fondé sur une représentation floue en présence de critères multiples. Cahier du CERO 20:3–24 Roy B (1985) Méthodologie multicritère d'aide à la décision. Economica, Paris Roy B (1990) The outranking approach and the foundations of ELECTRE methods. In: Bana e Costa CA (ed) Readings in multiple criteria decision aid. Springer-Verlag, Heidelberg, pp 155–183 Rui-Figueira J, Greco S, Roy B, Slowinski R (2010) ELECTRE methods: Main features and recent developments. In: Zopounidis C, Pardalos P (eds) Handbook of multicriteria analysis. Springer, Berlin, pp 51–89 Schärlig A (1996) Pratiquer Electre et Prométhée. Un complément à Décider sur plusieurs critères. Presses Polytechniques et Universitaires Romandes, Lausanne Sobrie O, Pirlot M (2012) Implementation of ELECTRE TRI in an open source GIS. Newsletter of the EWG-MCDA 3(26):14–18 Surgan M, Condon M, Cox C (2010) Pesticide risk indicators: unidentified inert ingredients compromise their integrity and utility. Environ Manag 45(4):834–841. https://doi.org/10. 1007/s00267-009-9382-9 Van Der Werf HMG (1996) Assessing the impact of pesticides on the environment. Agric Ecosyst Environ 60(2–3):81–96. https://doi.org/10.1016/S0167-8809(96)01096-1 Vincke P (1989) L'Aide Multicritère à la Décision. Université de Bruxelles, Bruxelles

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions Irène Abi-Zeid, Francis Marleau Donais, and Jérôme Cerutti

1 Introduction Multicriteria decision aiding (MCDA) is a well-established field that has seen many methodological developments and numerous applications in various areas including transportation, healthcare, environmental science, nature conservation, land use planning, heritage conservation, and innovation management, to name a few (Adem Esmail and Geneletti 2018; Cegan et al. 2017; Cestari et al. 2018; Frazão et al. 2018, 2021; Langemeyer et al. 2016; Marleau Donais et al. 2019a; Norese and Carbone 2014). It proposes a variety of families of methods that aim at constructing and explicitly taking into account multiple criteria, to aid individuals or groups in their evaluation and decision-making process (Belton and Stewart 2002; Cinelli et al. 2020). It is particularly suited when quantitative and qualitative information, as well as stakeholders’ preferences, must be integrated (Roy 2016). MCDA allows for structured and rigorous evaluations in contexts characterized by conflicting objectives. Rather than being based on market preferences, as in benefit-cost analyses, MCDA models are often jointly constructed according to stakeholders’ values and priorities (Munda 2016). Socio-technical MCDA processes, where an MCDA model is co-constructed in a social setting with various participants, promote dialogue, transparency, and increase acceptance since they are conducted in the context of organizational and environmental considerations (Phillips and Bana e Costa 2007). I. Abi-Zeid (✉) Université Laval, Québec, Canada e-mail: [email protected] F. Marleau Donais Direction des projets structurants bus et tram de l’ouest, Ministère des Transports, Montréal, Canada J. Cerutti Département opérations et systèmes de décision, Pavillon Palasis-Prince, Université Laval, Québec, Canada © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_6

161

162

I. Abi-Zeid et al.

A considerable portion of the MCDA literature focuses on the development of methods rather than on problem structuring. This gives the impression that decision processes are already structured in sets of alternatives and criteria, and that the stakeholders have a common value system. However, the main challenge in applying MCDA is undoubtedly the problem structuring phase since decision problems are often ill-structured and surrounded by uncertainty. Alternatives and criteria are rarely readily available (Belton and Stewart 2010). Nonetheless, even in application papers, relatively few papers describe real MCDA interventions and the difficulties and hurdles encountered during socio-technical MCDA processes (Ormerod 2018; Bana e Costa et al. 2019). Our aim in this chapter is to remedy this situation by (1) providing concrete examples of problems encountered in real-life socio-technical MCDA interventions, and (2) offering advice and insights on what to expect and what to look out for in such a process. We concentrate on evaluation processes and present two examples. The partner organization in both projects was Quebec City, the capital of the province of Quebec in Canada, with approximately 531,000 inhabitants. Founded in 1608, it is one of the oldest cities in North America. Its historic district, old Quebec, was included in the world heritage list in 1985. The first project, of a larger scope and duration (2 years), is in the transportation field and is related to street redesign prioritization. The second project, of smaller scope and duration (a few months), is in the urban planning field regarding the role that should be played by a municipal governing body to ensure architectural conservation and protection. Rather than describing the interventions in detail, we highlight some difficulties encountered before, during, and after the completion of the interventions and share some insights. Through these examples, we wish to emphasize that, contrary to what some believe, MCDA is not a straightforward process where the most challenging part is the technical application of a method. For each intervention, we provide the decision problem description, its environment, its funding structure, and how the project emerged; the organizational context, the stakeholders involved and the nature of the facilitation team’s interaction with the participants; the main aim of the intervention and the various steps conducted in each of the structuring, modeling, and evaluation phases; and finally, a discussion including the post-project evaluation.

2 Intervention Process Description The interventions presented here were action-research projects conducted by our research team at Université Laval (UL team) where we acted as facilitators, in a co-constructive process with members of the partner organization, who were implicated at every stage of the project (Franco and Montibeller 2010). Action research is a research strategy that goes beyond the description and explanation of a phenomenon. It combines theory and practice in order to develop solutions that make it

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

163

possible to act on problems observed in the field (Roy and Prévost 2013). It is considered an appropriate method for MCDA interventions (Montibeller 2007). It “involves the researcher in working with members of an organization over a matter which is of genuine concern to them and in which there is an intent by the organization members to take action based on the intervention” (Eden and Huxham 1999). We adopted a constructivist view where the knowledge produced results from the interaction between subjects (partner organization and UL team) and a problem. We acted as group facilitators where group facilitation can be defined as a “goalorientated dynamic process, in which participants work together in an atmosphere of genuine mutual respect, in order to learn through critical reflection” (Burrows 1997, p. 401). We conducted the workshops according to decision conferencing principles (Phillips 2007) over a period of many months. Decision conferencing is a social process where key actors are engaged in the modeling process, thereby ensuring their ownership of the developed artifacts and their subsequent implementation (Phillips and Bana e Costa 2007). Such a process is assisted by specialist(s) in decision aiding who act as an impartial facilitator(s) “whose role is to enhance communication between the participants and to get them to constructively deal with the conflicting issues at hand” (Marttunen et al. 2015). In order to structure the problem and develop a set of objectives and criteria, we followed a value-focused thinking approach, which allows to articulate and use the core values of a person or a group to guide their decisions (Keeney 1996, 2007). Our interventions were multimethodology based on “soft” and “hard” Operations Research methods (Mingers and Brocklesby 1997; Henao and Franco 2016) consisting, of a mix of conceptual and cognitive maps (Eden 1988), MCDA methods and software, and where applicable, geographic information systems (GIS). We used different MCDA methods in the interventions presented here: An additive value function-based method, MACBETH (Bana e Costa et al. 2012) in the first intervention and an outranking method for sorting, ELECTRE Tri-nC, (Almeida-Dias et al. 2012) in the second intervention. Our intervention process, illustrated in Fig. 1, was similar in both projects. It is the one we follow in all our interventions: An intervention is usually activated by the partner organization (PO) that, following the observation of its environment, has identified, and retained a problem, as defined in Landry (1995), as well as the need for knowledge production to support action. Subsequently, contact is established between the PO and our UL team such that, following preparatory meetings, the two parties come to an agreement on a project definition, a timeline, and a financing structure. Subsequently, the intervention is initiated and consists of the following phases of a socio-technical MCDA evaluation process: a structuring phase, a modeling phase, and a model-based evaluation phase. Following model and results validation, these are communicated through a decision support tool or through recommendations. Knowledge transfer is continuous during the intervention and is

164

I. Abi-Zeid et al.

Fig. 1 Illustration of our intervention process

completed through the delivery of a decision support tool or recommendations to the PO, that normally implements it in their decision-making process. Sometimes, it is necessary to have further subsequent interaction to provide the PO with technical support with the tool. Finally, the project is concluded with a follow-up in the form of a post-project evaluation, that is more or less extensive. We present in the following sections the two interventions.

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

165

3 First Intervention: Ranking Streets According to their Potential to Become Complete Streets 3.1

Decision Problem Description

This project emerged as an initiative of the newly founded Unité Mixte de Recherche en sciences urbaines (UMR-su). The UMR-su was a research unit at Université Laval active between 2015 and 2020 that aimed at bringing together players from business, government, and the academic community. The objective was to exchange and transfer knowledge and develop innovative solutions to problems inherent to Quebec City through the creation of an urban laboratory. Our project was one of the first to be carried out by the UMR-su and was presented as a flagship of this new venture. It aimed at showing the potential of MCDA in the field of transportation and involved Quebec City administration, later referred to as the City or the PO, as well as an international private company as the business partner. Initially, the City had chosen a street to be redesigned as a complete street, an approach that aims at ensuring that streets are safe, accessible, and comfortable for all users, no matter their mode of travel or capacities (McCann 2013). The City’s professionals needed help in choosing the best design for that street. However, the situation had changed internally, a situation that sometimes occurs during the conduct of a project where priorities shift or personnel changes. Therefore, the redesign of the street had been postponed. In the meantime, the municipal professionals involved had realized that what they really needed was rather to develop a new process to identify and choose the streets that could be designed as complete streets. In fact, the City had started in 2014 to change its approach to prioritizing and selecting street rehabilitation and redesign projects. Its previous street selection process was tedious, time consuming, and sometimes generated mistakes, which created frustration among the professionals. The City wished to move away from decisions taken solely by the engineering department, based on technical obsolescence criteria, and to become more inclusive by involving the transportation, urbanism, urban design, and environmental departments. The City was ready for a structured, rigorous, and transparent decision process, conducted in collaboration with stakeholders from different departments, which considers various viewpoints (engineering, urban planning, environment, etc.), preferences, and objectives, in order to ensure that the implemented choices are acceptable to all (Marleau Donais et al. 2022). In summary, it was ready for a socio-technical MCDA process. The project was funded in two phases over a span of 2 years: the first phase in 2016 was financed through the UMR-su by a joint research grant from the Canadian federal government (MITACS Accelerate1) and by a private international company; a second phase in 2017 was funded directly by Quebec City through a research contract with Université Laval. The first phase served to develop an MCDA model to

1

https://www.mitacs.ca/en/programs/accelerate

166

I. Abi-Zeid et al.

evaluate the potential of a street segment to be redesigned as a complete street, to apply this model to approximately 5000 street segments, and to implement it in a multicriteria spatial decision support system (MC-SDSS) (Malczewski and Rinner 2015). The second phase saw the extension of the MC-SDSS to all the street segments within the urban perimeter (approximately 20,000 street segments). It also led to the development of new software for automatic data treatment and transfer between the M-MACBETH software and ESRI’s ArcGIS2 (Marleau Donais et al. 2017). This new software was recently expanded into an application called Othello, available as open source on Github.3 Othello was funded by CANARIE,4 a Canadian non-profit networking corporation to connect researchers, educators, and innovators.

3.2

Organizational Context and the Participants

A total of 6 group workshops and 11 subgroup workshops were held to develop the model in the first phase. The participants included 11 representatives of the different departments involved in street selection and design process in Quebec City: a transportation engineer, an infrastructure engineer, three urban planners from different departments, a project manager, an urban designer, an environmental planner, a landscape architect, an advisor in public participation and the sustainable development project director. A professional from the private business partner was also present at the workshop as an observer. Two to three members of the UL team who acted as facilitators were present during the group workshops. The main facilitator (a MSc student in land planning at the time, second author) led the discussions, the second facilitator captured the information using software, and the second and the third facilitators (professors) advised the first facilitator and analyzed the workshops. As for the subgroup workshops, they involved smaller groups of city professionals with specific expertise in a field. Table 1 present the main differences between the characteristics of the group and subgroup workshops.

3.3

Main Aim of the Intervention

The aim of this intervention was to assist Quebec City in selecting streets to be redesigned as complete streets. To better scope the project, a first preparatory meeting was organized to understand the current decision environment, define the problem, and identify key players to invite to the workshops as participants. Another preparatory meeting was held with the project manager to further understand the

2

https://www.esri.com/en-us/home https://github.com/ulaval-rs/othello 4 https://www.canarie.ca/about/ 3

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

167

Table 1 Differences between the characteristics of group and subgroup workshops Length Group size Number of facilitators Expertise Data access and manipulation during the workshop Main objectives

Group workshop 2–3 h 8–11 professionals 2–3 facilitators Varied Hard

Subgroup workshop 30–60 min 1–4 professionals 1 facilitator Specific Easy

Different at each workshop

Define criteria references and preference scales

decision process in place. Following discussions, we framed the problem as a scoring and ranking problem where we needed a method that will provide each street with a score. The modeling phase, consisting of brainstorming sessions with the participants in order to identify their values and their objectives, led to the construction and validation of 11 criteria that reflected a street’s potential to be redesigned as a Complete Street (Marleau Donais et al. 2019b). We chose the MACBETH method to develop an MCDA model to evaluate streets and obtain a score that represented a street’s potential to be redesigned as a complete street. We constructed local “attractiveness” cardinal (interval level) scales for each of the 11 criteria, that reflected the values and preferences of the workshops participants. A street’s performance on a criterion was translated to its attractiveness score such that, the higher the attractiveness, the higher the potential of a street to be redesigned as a complete street from the perspective of that criterion. The scaling constants, commonly called weights, were also derived using MACBETH. There were several reasons for selecting this method. First, we wanted the simplest and most familiar model, namely the additive aggregation model (weighted sum) and MACBETH avoids the main traps associated with aggregating several criteria based on a weighted sum (Bana e Costa et al. 2012). In fact, the use of a weighted sum in practice often leads to methodological problems due to the incorrect use of non-cardinal scales and/or to the incorrect interpretation of weights as importance scales. To ask the question “what is the relative importance of criterion” is meaningless in a weighted sum. Weights cannot be evaluated by directly comparing criteria without considering the ranges of the performance measures on the criteria. This is a mistake encountered in several popular weighting procedures and is the most common and critical mistake in weighted sums (Kirkwood 1997; Keeney 2002). Second, the method has its own supporting computer software (M-MACBETH) that allows simple and effective recording of qualitative judgments formulated by the participants, thereby enabling the real-time construction of the model and visualization of the resulting scores. Furthermore, this software identifies possible inconsistencies between the judgments expressed during the process and proposes alternative solutions, when necessary. Finally, the fact that participants need only to provide qualitative information in MACBETH makes it easier to arrive at a consensus within a group.

168

I. Abi-Zeid et al.

In order to validate the model, participants were then asked to rank a set of 20 anonymized streets, based on their performances on the 11 criteria. The set of anonymous alternatives was representative of a wide variety of contexts, i.e., streets which potential to be redesigned as a complete street ranged from a very low priority to a very high priority with a variety of high or low performances on different criteria. When the ranking obtained by M-MACBETH differed from the ad hoc ranking, the participants were asked to explain their choices. We also looked at the difference in scores of anonymous streets. At the end of this process, 80% of segments (16 segments) were ranked in the same order as that obtained by the MCDA model and the scores’ differences reflected the participants’ vision. As for the four segments that had a different ranking from the one obtained by the model, it was not possible to modify the model while remaining coherent with the judgments elicited through the whole process and the validation. We could not change the scales further without ending up with inconsistencies with the qualitative judgments that the participants had provided. Following discussions, the participants came to the conclusion that the model was satisfactory as developed. Once the MCDA model was validated, the next phase was to combine a GIS with the MCDA model to process data, to incorporate spatial information, to assess the street segments, to obtain their final scores and provide their rankings. The aim was to produce maps presenting street rankings, rather than showing the scores, which do not mean much when we are dealing with thousands of evaluated streets. A higher ranking represents a higher potential and therefore higher priority to be redesigned as a complete street. Ten categories were then defined based on the scores’ deciles. Figure 2 provides an example of such a map. The model and the results, implemented in a spatial MC-SDSS were transferred to the City along with a final report (end of 2016). In March 2017, the mayor of Quebec City revealed the city’s Complete Streets strategy to the population and the media. The multicriteria model and the MC-SDSS were integrated as one of the key elements of their strategy, which strategy received good local media coverage (newspaper and television). The mayor qualified the project as the “quintessence of transdisciplinary” work. Following the favorable acceptance of the project results by the elected officials, the City wished to further develop the tool. It expressed its wish to extend the model to the whole street network in order to rank the 20,000+ street segments inside the urban perimeter and to update the data from 2016 to 2017. This led to Phase II of the project, completed in 2017.

3.4

Discussion

Following the MC-SDSS’ transfer, the professionals from the Complete Streets team of Quebec City started using it in an operational setting in their day-to-day planning activities. They tested it in 2018 and 100% of the streets planned in 2017 which had already been identified (before our project) as having a strong Complete Street potential were also identified as high potential streets by the model. The test thus

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

169

Fig. 2 Complete street priority index for the central neighborhoods

further validated the model and convinced the professionals of its value. A subsequent analysis of the data led them to estimate that the model is accurate in approximately 95% of all cases. A new decision process was subsequently implemented to prioritize and select the streets, integrating the new model and tools developed (translated from Ville de Québec 20175). This new decision process, according to the project manager, allows them to substantially save time, financial, material, and personal resources. What would have taken several days (or possibly several weeks) of analysis using the City’s previous methods was completed in a half-day meeting using the new model and tool. Furthermore, it was now possible to take Complete Street principles into account since the inception of a project, which was not the case before. In 2018, the City went public with the tool, and started using it and the maps produced during public consultation around specific Complete Street projects. Many awards followed the completion of our project. In fact, Quebec City, and the tool developed, were cited as one of the top 12 Complete Streets initiatives in 2017 in North America by Smart Growth America and National Complete Street Coalition6 and received a 2018 “best Practice” award from the Upstate New York 5

https://www.ville.quebec.qc.ca/apropos/planification-orientations/amenagement_urbain/ruesconviviales/docs/2017-03-02_RuesConviviales_pl%C3%A9nier.pdf 6 https://smartgrowthamerica.org/resources/best-complete-streets-initiatives-2017/

170

I. Abi-Zeid et al.

Chapter of the American Planning Association.7 On the academic front, the project won the Canadian Operational Research Society’s practice prize in 20198 and was a finalist in the INFORMS Decision Analysis Practice Award in the same year. During the intervention, we faced many challenges. The multiparty nature (University, City, private business partner) of the project created some management and governance problems in the project. Initially, before we got involved in the project, the UMR-su had informed Quebec City that a software developed by the private partner company would be used in this project. This was a clear example of a hammer looking for a nail instead of the other way around. However, this method did not properly address the problem since, it was in our opinion, too complicated to use and understand. It implemented a method that models criteria interaction. Criteria interaction occurs when there is a synergy, positive or negative, between criteria, meaning that the difference in preference between two performances on a criterion depends on the fixed values of the other criteria. Consider, for example, two criteria A and B with performances A1, A2 and B1, B2 where A1 is preferred to A2 and B1 is preferred to B2. There is interaction between A and B when the difference in preference between the performances (A1, B1) and (A1, B2) is not equal to the difference in preference between the performances (A2, B1) and (A2, B2), i.e., when the difference in preference between two levels of criterion B depends on the value of criterion A. It is, in our opinion, a better practice in interventions, such as the one described here, to always redefine criteria to ensure that there is no interaction. This allows us to use the simpler additive model, which is easier to understand and requires less parameters that are difficult to obtain and to interpret. However, our position regarding the private partner’s software created some management issues from the start of the project since the private partner was one of the project funders. As facilitators, we defended our point to use criteria reformulation and the simpler MACBETH method that, in our opinion, best suited the problem in practice, rather than forcing the use of a more complicated method. Nonetheless, this situation may have created some discontent for the business partner. As for the method’s application, many difficulties were encountered. The project required us to find an equilibrium between our research vision from Academia and the City’s operational vision (consumer of our research). For all the participants, it was their first experience with multicriteria workshops. Several participants were out of their comfort zone and expressed doubts about the project, and we were faced with two opposing views of the method’s validity. Some perceived the approach to be too “rational” (mainly the non-engineers) and others too “subjective” (mainly the engineers). Some also feared that the project would be too theoretical and its results not applicable in real life. It was not clear to them what the project could bring and what would be its added value. By adopting a facilitator approach as opposed to an

7

https://www.nyupstateplanning.org/2018-ny-upstate-apa-chapter-awards/f0gec64edj79ojqeezfx3 lmeq99n2p 8 https://www.cors.ca/?q=content/practice-prize-competition#:~:text=The%20Practice%20Prize% 20Competition%20recognizes,to%20the%20CORS%20Annual%20Conference

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

171

expert approach, we were able to quickly reach an equilibrium and gain their trust. In an expert mode, one goes away with a problem and comes back with a solution, whereas in a facilitated mode, the client learns about his/her problem and a solution is co-constructed. As facilitators, we encouraged full participation, promoted dialogue, and used a “sharing” procedure when facing conflicting views as defined in Belton and Pictet (1997). Our aim was therefore to reach a consensus through discussions and negotiation of agreements by addressing the differences of opinions and trying to reduce them. The involvement of professionals from different fields and departmental cultures also created challenges. More time was needed for discussion and negotiation to reach a consensus and formalize their experience into useful information for the model. The participants’ different knowledge backgrounds sometimes created a confusion due to a lack of a common vocabulary. Resistance to change is normal in an organization and breaking professional silos is never an easy task (de Gooyert et al. 2017). Furthermore, the cognitive burden required during some workshops were perceived as mentally exhausting by some of the participants. Moreover, since some heads of departments did not fully understand the project, they did not encourage their employees to participate in the workshops. In the end, the cumulative effects of these difficulties led to attendance problem; there were several workshops where some participants were missing, which negatively affected the group’s esprit de corps. It is worth mentioning that we had no control over the choice of participants in the process and no power to influence that choice.

3.5

Post-Project Evaluation

Two years after the end of the project, we conducted a post-project analysis to gain insights on how to improve our future research and development practices by interviewing the workshops’ participants, the professionals responsible for the MC-SDSS and the MC-SDSS users (Marleau Donais et al. 2021). Although very important for learning, a posteriori (ex post) analyses following MCDA processes are often neglected. We conducted individual interviews with the professionals responsible for the tool (2), the workshops’ participants (7 out of 11), and with the tool’s users (5). These interviews allowed us to identify several strengths of our approach, some weaknesses, and potential improvements, and to better understand the impacts of the tool’s development and of its use on the organization. Several aspects of the tool development process were identified as strengths by the professionals: (1) the group workshops created an exchange forum between the professionals where everyone could express his/her concerns in order to reach a consensus; (2) the tool was built with the professionals according to an interdisciplinary vision and facilitated by external neutral actors, namely the UL team; and (3) the organization of sub-workshops midway in the process made the project more effective and efficient. These were perceived very positively by the professionals. They allowed them to be more creative in their thinking, gain confidence, and

172

I. Abi-Zeid et al.

increase their trust in the process. Moreover, several aspects encourage the tool’s utilization in the professional practices and were also perceived as strengths: (1) the geomatical nature of the tool allows to update the tool; (2) the pooling of a large amount of data allows a global vision of the situation; (3) the visual quality of the tool supports communication between the professionals, the elected officials and the citizens; and (4) the tool allows the users to identify the higher priority streets, to rigorously justify the decisions, and to easily explain the rationale behind decisions for maximum transparency. Nonetheless, the professionals identified, during our post-project interviews, aspects that could limit the use of the tool in the future. These include: (1) Control issues—The tool has little flexibility to add or to remove criteria since it requires organizing new workshops in order to keep the tool coherent. The professionals perceived this as a limit while simultaneously appreciating that it safeguards against misuse; (2) Expectations management issues—The tool identifies the street segments to prioritize but does not suggest specific designs. Quebec City would have liked to add a decision tree that suggests street design depending on the urban context. This possibility was discussed at the start of the project but was completely out of our project’s scope. (3) Scale issues—A large amount of data makes it difficult to update the tool. This limitation was partially solved by developing the tool Othello referenced earlier. However, as long as Quebec City keeps on changing the database structure of some criteria from one year to another, it will be difficult to completely automate the process. Finally, the professionals responsible for the Complete Street project mentioned that some people were already trying to misinterpret the result in order to conduct non-priority projects. This is obviously outside of our control since, with the right (or wrong) mind-setting, everything is subject to being manipulated.

4 Second Intervention: Rethinking the Role and Jurisdiction of a Municipal Planning and Conservation Commission 4.1

Decision Problem Description

This second intervention was also conducted with Quebec City as the partner organization. Like in many parts of the world, when a private citizen or a commercial entity wishes to undertake construction or renovation work on a property, a permit from the municipal authorities can be required. However, Quebec has a particular decision-making process when it comes to issuing building and renovation permits. As a matter of fact, depending on the types of construction or renovation work and on its location, an initial authorization from the Commission d’Urbanisme et de Conservation Québec (CUCQ) may be required before the permit is issued. The CUCQ, a unique instance in the province of Quebec, has the mandate to control the implementation and architecture of constructions, land development, and related

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

173

works when they are carried out in specific sectors or on certain properties. A Quebec City regulation, R.V.Q. 13249, of more than 1400 pages, describes the sectors and properties as well as the type of work under the commission’s jurisdiction, for which municipal permits and certificates of authorization can only be issued with its prior approval. Created in 1928 in response to social conflicts between citizens and the provincial and municipal levels of government, its aim is to ensure the city’s architectural coherence and preserve its landscape. It was established as a response to the partial destruction of the city center to make way for the modernity of the 1930s, namely roads widening or skyscrapers building (Therrien 2011). Today, the commission consists of experts from within and outside the City’s administration (architects, urban designers, landscape architects) and elected officials. Together, they analyze and decide on permit applications according to an extensive set of conditions related to architectural criteria (quality of materials used, size of windows, color of materials, etc.) or landscaping criteria (volume of the building, green area, etc.). Needless to say, the City has changed considerably between the commission’s creation almost a 100 years ago and today. It has grown substantially following its fusion with neighboring cities in 2002. In addition, major development projects (e.g., public transit system) and major neighborhood revitalization are underway. These numerous transformations have and will have an impact on the face of the city. In order to accompany these changes, the city of Quebec wished to review the role and scope of the CUCQ’s jurisdiction in order to best meet its control and conservation needs. In addition, in keeping with the Orientation 6 “Efficient Capital” of the revised land use and development plan for the Quebec City metropolitan area,10 the City wished to change the decision-making process for the issuance of construction/renovation permits to make it simpler and more transparent for citizens. However, the question was: how to rethink such a symbolic and powerful institution? To examine this question, the City had set up an advisory committee made up of internal professionals with different backgrounds. Aware of multiple MCDA interventions that we had conducted with the City, the CUCQ project manager contacted our research team in early 2020. The mandate was to assist the advisory committee in their task. Following a preparatory meeting with the project manager, an agreement was reached, and the intervention was planned to start in September 2020. Prior to beginning the intervention, the UL team was invited to the kick-off meeting of the advisory committee. We were there as observers, but this allowed us to understand the issues and the power dynamics within the committee. The project was funded through a contract with the University. It led to the development of an MCDA model to sort combinations of work types and sectors in one of five categories as a function of their relevance level to be under the jurisdiction of the CUCQ.

9

https://reglements.ville.quebec.qc.ca/fr/showdoc/cr/R.V.Q.1324/ Schéma d’Aménagement et de Développement Révisé de l’agglomération de Québec, Vol 1, Ch. 4 p.30, entered into force February 7, 2020. 10

174

4.2

I. Abi-Zeid et al.

Organizational Context and the Participants

A total of five 3-h group workshops were held, between September and December 2020, with eight participants from four different administrative services, who were also members of the advisory committee. They consisted, from the Planning and Environmental Service, of its director, a team leader, an advisor in architecture and urban design, an advisor in architecture, and the intervention’s project manager. A lawyer from the Legal Affairs service, an advisor for the implementation of development and heritage projects from the Heritage and International Relations service, and a division director from the Territory Management service completed the participants’ list. The Université Laval (UL) team consisted of the first author of a PhD student in land planning (third author) and a research professional for data processing support. Two members of the UL team led the intervention, the first author acting as the lead facilitator, assisted by the PhD student during the meetings. Due to COVID-19, the first workshop was in hybrid mode, online and in person, while the other four were online. Again, we had no control in this intervention over the choice of the participants.

4.3

Main Aim of the Intervention

The first workshop was mainly used to understand the problem and structure it. The question of rethinking the role of the CUCQ was vague and we needed to give the project a direction. In preparation for this workshop, the UL team had analyzed the R.V.Q. 1324 regulation document and l’assistant permis, an online tool11 available to the citizens to guide them in the various steps involved in obtaining a building or renovation permit. The main concepts and relationships were extracted in a conceptual map and served as a basis for discussion. This map was extremely valuable as it summarized 1400 pages in a figure, albeit a dense one. . .The main elements of the map pointed toward a set of work types and a set of sectors or specific building types. We suggested that a possible way to rethink the role of the CUCQ would be by evaluating all the elements under its jurisdiction and assessing how relevant it is for the CUCQ to control these elements. It was not a priori evident what were he elements to be evaluated, and together with the group, a consensus was reached to develop a model to evaluate combinations consisting of work types and their locations (in a sector or a building). These work type/location combinations, 926 of them, would be our “alternatives” in the MCDA vocabulary. An example of a combination is: “Accessory construction attached to a main building (carport, garage, heat pump, etc.) in Saint-Roch district.” The discussions that led to the definition of these combinations during the first workshop helped the participants to subsequently eliminate a large number of 11

https://www.ville.quebec.qc.ca/services/assistant-permis/index.aspx

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

175

combinations reducing them to 259. During and after the exchanges, it had become obvious to them which ones could remain under the CUCQ’s jurisdiction, and which ones had to be excluded. The question was then what to do with the remaining 259 combinations or in MCDA terminology, what is the decision problematic? Since it was clear that this was not a ranking problem, we formulated it as a sorting problem where the combinations had to be classified into five ordered categories representing a combination’s relevance level to be under the commission’s jurisdiction: very low, low, medium, high, very high. The number of categories was validated with the group and arrived at following numerous discussions. In order to construct the criteria, we conducted a brainstorming session to identify the values using a value-focused approach (Keeney 2007). Prior to the meeting, we shared with the project manager the type of questions that we would be asking during the workshop. These included: What are the City’s planning objectives? In what ways do the work/sector combinations differ from each other? What bothers you about the current scope of the CUCQ mandate? In relation to the CUCQ mandate, what do you want? What is important to you? What is your ideal? What would be a perfect combination (work/sector) to keep under CUCQ’s jurisdiction? To be excluded from CUCQ’s jurisdiction? What would be a bad combination to keep under CUCQ’s jurisdiction? To be excluded from the CUCQ? What would be a reasonable combination? What is good or bad about a reasonable combination? What has gone right or wrong with the current situation in relation to CUCQ’s mandate? What could go right or wrong in the future? What might concern elected officials, or citizens, or commercial companies about the scope of the CUCQ’s mandate? How about other stakeholders? What could concern you in the future? Again, we used a conceptual map to organize the results of the brainstorming session and presented it at the following meeting with a suggestion of criteria based on the regulation and on the discussions. We finally retained eight criteria, all measured on ordinal scales with six performance levels: null, very low, low, medium, high, and very high reflecting the relevance that a combination be under the CUCQ’S jurisdiction from the perspective of that criterion. Since the aim was to sort the combinations into categories, we chose ELECTRE Tri-nC (Almeida-Dias et al. 2012) as an appropriate method to apply in this context, especially with ordinal criteria. In addition, we had a software, MCDA-ULaval, developed by our team,12 that implements different methods of the ELECTRE family (Figueira et al. 2016) and that is available for free online. Furthermore, the information needed from the participants is minimal with this method, namely some typical performance profiles representing each category (although this is not always easy to obtain), and importance weights of the criteria which we derived based on the revised Simos’s method (Figueira and Roy 2002) using a computer implementation developed by our team. There was no justification for using additional parameters such as thresholds. The procedure was therefore straightforward and led to weights that were accepted and validated during the workshop with minimal discussion. A value function method

12

https://mcda.fsa.ulaval.ca/

176

I. Abi-Zeid et al.

such as MACBETH where categories are defined based on score intervals could also have been applied. However, such a method would have required quite a high cognitive burden to develop the cardinal scales for the criteria and we felt that it was unnecessary in this context. The following step was to evaluate the 259 combinations based on the eight criteria. This work was conducted internally by the City without our help, since they are the domain experts. We received a first evaluation matrix in early December 2020. We were also provided with typical reference profiles, three per category. Two adjustments to the profiles were needed to ensure that the profiles were separable (Almeida-Dias et al. 2012), a condition for the application of the ELECTRE Tri-nC method, namely that a profile from a better category must always have, on each criterion, a better or equal performance than a profile from a worse category. We then used the MCDA-ULaval software to process the data and presented the results in the following week. No combination was classified in the very low category and over 40% were in the very high category. The proportions in the high and the medium category were around 25% with less than 10% in the low category. Eighty five percent of the combinations were classified in a single category with the remaining being classified in adjacent categories (for example, between very low and low). As a matter of fact, the method ELECTRE Tri-nC provides a lower and a higher category for each element under evaluation. The model is more adequate and more stable when this “hesitancy” between the categories is kept to a minimum. As a rule of thumb, in our practice, when a model classifies over 80% of the alternatives in the same minimum and maximum category, we consider it valid, as long as the minimum and maximum categories, when different, are adjacent. In this project, we had achieved over 85%. The results were presented and validated in the last workshop. It was not surprising that none of the combinations were in the very low categories, since these were probably eliminated when the list of combination was reduced from over 900 to under 300. The very low category probably represented a null in the participants’ minds. Following the presentation of these results, the participants took some offline time to reanalyze the combinations and chose to reduce their number again, this time from 259 to 132. They also changed some of the previous performance evaluations on the criteria. The final evaluation matrix of the 132 combinations was received in March 2021 and analyzed, 89% of which were assigned to single categories. Again, as could be expected, no combination was in the very low category, under 10% were in the low category, around 25% in each of the very high and medium categories and around 32% in the high category.

4.4

Discussion

Contrary to the first intervention described above, no decision support tool per se was delivered. Rather, the report containing the model and the sorting results, along with a frequency analysis, were the final deliverables. The workshops were completed in December 2020, but the project was still active in the spring of 2021 and the

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

177

final report was delivered in June 2021. In the meantime, the project manager had changed positions in March 2021 and was replaced by another person who reached out to us a month after we had delivered the report. She wrote the following (translated from French) “As I did not participate in last year’s work, your report has allowed me to better understand and appreciate the approach taken. The results, as well as the methodology, will be an excellent basis for further reflection. The anchoring with the City’s orientations, the exercise around the weighting of the criteria and the results in the tables which are found in the opposites (very (sic) low or very high), are revealing. Also, since last May, we have modified the CUCQ’s jurisdiction; our decision is supported by the findings of your report.” It seems that the project’s results have been useful and have helped the City in its process to rethink the role of the CUCQ. From that point of view, the project can be considered a success. Nonetheless, we encountered some difficulties during the intervention. First, it was clear during the workshops that the different participants had very different opinions regarding the role that the CUCQ should play. Three different positions in the group were quite clear. Some participants thought that the CUCQ should have less power, other that it should have more power, while the remaining would have opted for the status quo. This is understandable since many political, economic, and social issues were at stake. However, in the end, the divergence of opinions did not have an impact on the model since the whole group ultimately agreed on the criteria and the weights, as constructed during the workshops. The facilitation process was successful in bringing the group to a common and shared vision through exchanges. So where did these differences of opinion play out? It is hard for us to say. We do not know who exactly provided the profiles set characterizing the categories as needed in ELECTRE Tri nC, whether it was a group’s effort or not. Another point is related to the final list of combinations retained. It is not surprising that the list had evolved during the intervention, a consequence of the learning during the workshops. So, concerning the reduction from over 900 to under 300 combinations, we can safely hypothesize that the group was in agreement, since no opposition was expressed when we presented the evolving list of combinations. A question can still be raised regarding the performance evaluations, where we received the first performance matrix (grid) in December 2020 that seemed to be a sort of an “average” of individual matrices. Again, since we were not involved in this process and since no opposing opinion was voiced, we can safely assume that the participants agreed with the combinations’ performances on the criteria that we analyzed. We did offer to evaluate, in addition to the “average” grid, the individual grids but our offer was declined. A final question remains regarding the final set of combinations, that was reduced to 132 and which evaluations were sometimes different from the those in the first average grid (received in March 2020). Here, we cannot speculate on why the evaluations were different; we were asked to evaluate the final grid which we did. Should we have asked more questions? We felt that it was outside our mandate as facilitators since we were not present anymore during this phase, the group meetings having come to an end in December 2020. In hindsight, we should have maybe pressed this issue a bit further.

178

I. Abi-Zeid et al.

During the intervention, we were also confronted with the challenge of adapting to events external to the project. In fact, due to the health measures related to COVID-19, we experienced two major difficulties. The first was in switching the facilitation process to a virtual mode. This meant that, for example, we did not have our usual tools such as a flip chart for taking notes that were visible to all participants. At the time, we had no online tool and little experience that could allow us to properly conduct a structuring phase in a virtual mode. We had already switched to virtual workshops during two ongoing interventions with two other cities at the time of the first restrictive health measures. However, in these two cases, we had already structured the problem in presence, before the health restrictions, whereas we needed to start from scratch with the intervention described here. We have, since then, well adapted and use online tools that allow for better virtual facilitation such as the collaborative whiteboard Miro13 and have conducted at least two interventions completely online. The other difficulty was related to the uncertainty in the schedule which depended on the state of the health emergency, an uncontrollable hazard during the facilitation process. It led to a longer project duration since many of the participants were deployed elsewhere and assigned other tasks. What should have taken 3 months, took almost 9 months. Furthermore, moving from face-to-face facilitation to virtual facilitation impacted the dynamics of the group with the following consequences, some of which are not unique to online facilitation but exacerbated by it: – Disengagement of the participants who sometimes turned off their camera and probably multi-tasked. – Exacerbation of the power relations which led to some participants retreating and not fully taking an active part in the conversation. – Repeated absences of some group members due to the state of health emergency and the changing priorities of other internal projects. – Difficulties for novices unfamiliar with the technical aspects of MCDA methods. As a matter of fact, only group member had already participated in such a sociotechnical process, in two previous interventions that the first author had conducted with the City. Before moving to fully virtual workshops, we had attempted to conduct a mixed mode workshop, in presence and online simultaneously. It did not go very well. We had difficulty hearing the interventions of the online participants. They gradually became disengaged from the process. It also required a reorganization of the facilitators’ tasks, with one facilitator having to juggle between face-to-face and remote members while the other was taking notes. Things may have been better if we had third facilitator since we could have shared the tasks. As a matter of fact, at one point, a group member took on the task of managing the interventions of the participants who were attending virtually. Because of this experience, we now conduct workshops that are either completely online or completely in presence.

13

https://miro.com/

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

179

We are not comfortable with mixed mode. We do, however, believe that a mixed mode could still be conducted successfully given enough experience with the process, an adequate number of facilitators and the right tools.

4.5

Post-Project Evaluation

As we usually do after each intervention, we sent a series of questions14 to the project manager in the form of an ex post evaluation (in French). The answers to the questions were sent to us by the new project manager. Contrary to the first intervention described above, we did not interview individually the group members. In response to the question of whether the project met their expectations and needs, the answer was yes in terms of the methodology and the rigor that supports it, but less so in terms of the qualitative analysis of the results. They would have liked to see, in addition to the statistical analysis of the results namely, the frequency distributions of the evaluation levels per criterion and the number of alternatives assigned to each category, a qualitative analysis and interpretation of the results. This is in our opinion an issue of how we were perceived by them. We had initially made it clear that we were not domain experts, but rather methodological experts acting as facilitators. However, it seems that they still expected us to provide a qualitative opinion regarding the “why” of the results, which was clearly outside the scope of our mandate. In fact, at the first presentation of the results in December 2020, they had indicated that they would like to see an analysis. For us, this meant a statistical analysis, probably due to our professional deformation (job conditioning), the first author being trained and active in quantitative methods. We realize now that we should have explicitly asked, “what type of analysis do you mean?” In response to the question of whether the project had a real impact on their reflection regarding the revision of the CUCQ’s mandate, the answer was that the project allowed them to better frame their reflection by linking it to the various planning tools and orientations of the City. It also helped to reduce the “subjectivity” of the discussions. It allowed them to remove some specific types of work from the CUCQ’s jurisdiction based on the final results. In terms of how they experienced the intervention process, they, on the positive side, appreciated that the process reduced the “subjectivity” in the discussions, allowed them to structure the method for assessing the scope of the CUCQ and gave them new perspectives. It is interesting that they perceived that the process reduced “subjectivity.” They were probably referring to the arbitrariness of an ad hoc evaluation since they mentioned that they deemed the co-constructive evaluation process to be structured and organized. On the negative side, they mentioned the difficulties, due the context of COVID-19, in synchronizing and balancing efforts and timelines on both sides. They felt that we required a lot of input from them and that we were not aligned with their capacity to

14

The questions and answers are translated from French.

180

I. Abi-Zeid et al.

provide us with answers. This is probably a reference to the time pressure that they were under. They also suggested that the facilitation mode could be improved, possible by having a third person in the team. As for the final report, the project manager indicated that “(translated from French) It represents very well the work done and demonstrates the depth of our thinking process.” When we look back at the process from our perspective, it occurs to us that we might have not always properly managed the expectations. For example, after the first meeting, we were asked to provide the criteria, although we had clearly presented ourselves as facilitators and methodological experts. We replied that the criteria are the results of a joint constructive effort and that we would get to them together. Also, after having delivered the final results, we were asked what the impact of the criteria weights were on the results. We had however already expressed, that given the agreement in the group regarding the weights’ distribution, we would not conduct further analyses. In hindsight, we should have clearly indicated why we believed a sensitivity analysis of the weights was not necessary or maybe even conducted one. There was also the issue of “qualitatively” analyzing the results; we should and could have done better there.

5 Conclusion We have presented two interventions of different scopes conducted with the City of Quebec. These were in support of problem solving in strategic decisions contexts as opposed to routine decision-making (Simon 1960) and were at the top of the decision type hierarchy (Howard and Abbas 2014, Fig. 1.4). They, therefore, required a formal rigorous decision process. Both dealt with unstructured, non-programmed decisions with high importance and low frequency (French et al. 2009, Fig. 1.2) and can be considered complex as defined by the Cynefin model (French et al. 2009). However, the definition of the elements to be evaluated were not as challenging as in interventions where the alternatives have to be defined through a complex and creative process such as in Norese (2020). The interventions were characterized by the search and construction of solutions, learning, knowledge sharing and conflict reconciliation in the modeling process (Luoma 2016). We described some of the difficulties encountered and humbly took responsibility where it was due. In this conclusion, we would like to provide some insights, based on our global experience, that may be helpful in conducting socio-technical interventions. For a more complete list of practical issues and insights, we refer the reader to Chapter “GIS Based/ MCDA Modelling for Strategic Environmental and Social Assessment of Land-Use Planning Scenarios in Conflictual Socioecosystems” of Belton and Stewart (2002). Before the start of an intervention, one should be mentally prepared to encounter various difficulties, some familiar and others new. We have conducted multiple interventions over the years, and with each new one, there is always some new challenge. First, it is important to manage expectations on both sides and make sure, that these are understood by all involved, at the beginning of each workshop. The

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

181

role of facilitators as, ideally, neutral methodological experts should be emphasized. At the beginning of the intervention, the methodology should be presented, as well as the expected number and duration of the workshops. Participants should be informed of what is expected of them in each workshop and between the workshops. We always insist that all participants must be present at each workshop. If for any reason they cannot, then they must be prepared to accept the decisions made by the group without them. When possible, we suggest recording the meetings, easier done in a virtual mode, as this allows the absentees to hear what happened during the workshop. A workshop should always begin with a summary of what has been accomplished to date and what is to come. A special attention should be given to ensure that all participants play a role and that they are able to express themselves comfortably. Those who are less vocal should be gently prompted for their opinions before a modeling decision is made. This requires good facilitation skills. A good reference on facilitation is by Kaner (2014). Another crucial point is that criteria are constructed with the group. Beware of project managers who say, before the intervention, that they have the criteria and that they only need a method to aggregate the evaluations. This was never true in our experience, since people often confuse criteria, constraints, performance levels, and indicators. They are not always aware that MCDA implies integrating the preferences and the values of the participants. People tend to think that quantitative criteria are more “objective” than qualitative criteria. This is a trap since, in both cases, performances need to be interpreted as a function of the values and preferences. Some also believe that a quantitative measurement automatically defines a criterion. This is again misleading. For example, in one other intervention where all the criteria were quantitative, it was very difficult to agree on the type of measurement to use: absolute numbers (frequencies) or relative numbers (percentages). It took two workshops to make a choice and arrive to an agreement. It was not a question of disagreement within the group; the participants could not decide individually. We chose on purpose two interventions where we used two different methods that represent two classes of MCDA methods: measurable value function methods (MACBETH for scoring and ranking) and ordinal outranking methods (ELECTRE Tri-nC for sorting in ordered categories). We wished to emphasize that a different problem may call for a different solution and thereby to encourage readers to avoid an attitude where one looks for a problem that fits their favorite or most familiar method. The MCDA field is very rich and provides many methods (Greco et al. 2016); the final choice of a method in an intervention should be carefully thought out. There is no perfect method, and it is crucial to be aware of the weaknesses of the method chosen. An additive value function model, such as the one obtained with MACBETH, for example, is totally compensatory. This may or may not be desirable in the intervention at hand. AHP, a very popular method, has many shortcomings described in Munier and Hontoria (2021). Ordinal ranking methods such as the ELECTRE outranking methods can suffer from rank reversal. For an overview of rank reversal in different methods, see de Aires and Ferreira (2018). Also, when applying ELECTRE III for ranking, for example, one should examine the outranking graph between the alternatives rather than the ranks alone, since these may be difficult to interpret when there is incomparability.

182

I. Abi-Zeid et al.

There are some theoretical results that normally constrain the choice of a method. When a score that can measure the strength of preference and compare the differences in strength of preference is needed, a (simple) additive measurable value function model can be built in the absence of uncertainty. However, this assumes a difference independence between the criteria (Dyer 2016; Smith and Dyer 2021). Difference independence is achieved when the preference difference between two multicriteria alternatives that differ only on one criterion does not depend on the common values of the other criteria (Dyer 2016). This is difficult to verify before constructing the model and is made as a working hypothesis. In practice, we do it through validation, using a sample of alternatives, where the participants confirm whether the final scores and the differences between them are consistent with their vision and experience. We believe that using a method such as MACBETH, to correctly construct cardinal scales by defining criteria units as well as the scaling constants, followed by validation with the users is sufficient for us to consider that we have constructed a measurable value function in practice. A measurable value model requiring a weaker assumption, namely weak difference independence, where the ordering of preference differences on a criterion does not depend on the values of other criteria, is the multilinear model. This allows to take to into account interaction between criteria (Keeney and Raiffa 1993). In the absence of the weaker difference independence assumption, one can also use methods such as the Choquet integral (Labreuche and Grabisch 2003) to take into account interaction. There are also versions of the ELECTRE outranking methods that consider interaction (Figueira et al. 2009). In our opinion, models with interaction are more opaque for the participants since it is difficult to interpret the extra parameters needed. Our strategy, in the presence of interaction between criteria, is to redefine the criteria in order to avoid interactions so we can use a simpler model (Keeney 1981). Our aim is always to correctly develop practical models that are requisite, that is, good enough for the situation at hand (Phillips 1984; Keeney and von Winterfeldt 2007). When the problem at hand is that of sorting into ordered categories, we have often used ELECTRE Tri-nC. There is no issue with rank reversal with this method. However, it might be difficult to apply when one cannot obtain profiles that are separable, a situation we faced in one of our interventions (not described here), which forced us to change methods and switch to MACBETH instead, where we computed scores and defined intervals based on the scores. When uncertainty represented by probabilities is present, one should construct utility functions rather than value functions (Keeney and Raiffa 1993). In outranking methods such as ELECTRE, thresholds are used to account for uncertainty, ambiguity, or imprecision (Roy et al. 2014). Finally, the availability of a software is crucial for an intervention. The methods we used in our interventions here were supported, as mentioned earlier, by software available either commercially such as M-MACBETH, or for free such as MCDA-ULaval and Othello. It should be said that we have in the past tentatively chosen the type of method, following the preparatory meetings and prior to the intervention, and in some cases, we had to change methods during the process. An a priori idea of the method type is, in our opinion, necessary in order to estimate the project’s duration and prepare the financial terms in the project proposal, especially

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

183

when in a context of a contract. In our experience, more workshops are needed when we use the MACBETH method compared to when we use ELECTRE methods. One further element that we would like to address relates to the definition of stakeholders and their participation level in the process. This is a crucial step in an intervention and methods are available to help identify them and their level of participation (Marais and Abi-Zeid 2021). In both interventions presented here, we did not conduct this step since the interventions were less complex than what can be found in larger scope projects. Our interventions addressed internal strategic decision situations and we had no control over the choice of the participants, who were chosen by the partner organization, albeit from different departments. Nonetheless, we always raise this point in the preparatory meetings and ask questions to ensure that no relevant participants have been overlooked. Finally, we would like to mention that in the interventions presented here, the clients came to us because of our expertise in MCDA, and because they had heard of other MCDA interventions that we conducted. In that sense, it was easier in these projects than possibly in other projects where the facilitators often have the burden of convincing their clients that MCDA is pertinent and helpful. In our experience, our clients perceive our interventions as more rigorous and the results more legitimate because of our positions as academics. This might not be the case in other cultural or scientific contexts. In conclusion, interventions require both hard and soft skills. It is our opinion that modeling is also an art, an intuitive process that is shaped by experience (Morris 1967). In quoting George Box, “all models are wrong, but some are useful” (Box 1979), we can say that in the interventions presented above, the models constructed proved useful. The takeaway message here is that we conduct decision aiding to help decision makers and not decision-making where we replace decision makers. We construct models that reflect the values and concerns of the participants and are accepted by them. The models are naturally subjective, as they should be in decision aiding processes. Subjective does not however mean arbitrary; in our interventions it means consistent with the information, knowledge, and best expert advice available. Acknowledgments We wish to sincerely thank Professors Owen Waygood and Roxane Lavoie who were co-researchers in the first intervention. We also express our gratitude to all the participants from the City of Quebec who took part in the workshops. Finally, we thank Oscar Nilo Mellado, research professional at Université Laval, for his technical support during the second intervention.

References Adem Esmail B, Geneletti D (2018) Multi-criteria decision analysis for nature conservation: a review of 20 years of applications. Methods Ecol Evol 9:42–53. https://doi.org/10.1111/ 2041-210X.12899 de Aires RF, Ferreira L (2018) The rank reversal problem in multi-criteria decision making: a literature review. Pesqui Oper 38:331–362. https://doi.org/10.1590/0101-7438.2018.038.02. 0331

184

I. Abi-Zeid et al.

Almeida-Dias J, Figueira JR, Roy B (2012) A multiple criteria sorting method where each category is characterized by several reference actions: the Electre tri-nC method. Eur J Opl Res 217:567– 579. https://doi.org/10.1016/j.ejor.2011.09.047 Bana e Costa CA, De Corte J-M, Vansnick J-C (2012) MACBETH. Int J Info Tech Dec Mak 11: 359–387. https://doi.org/10.1142/S0219622012400068 Bana e Costa CA, Vieira ACL, Nóbrega M et al (2019) Collaborative value modelling in corporate contexts with MACBETH. Procedia Computer Science 162:786–794. https://doi.org/10.1016/j. procs.2019.12.051 Belton V, Pictet J (1997) A framework for group decision using a MCDA model: sharing, aggregating or comparing individual information? J Decis Syst 6:283–303. https://doi.org/10. 1080/12460125.1997.10511726 Belton V, Stewart TJ (2002) Multiple criteria decision analysis: an integrated approach, softcover reprint of the hardcover, 1st edn. Springer-Science+Business Media, B.V, Dordrecht, Netherlands Belton V, Stewart T (2010) Problem structuring and multiple criteria decision analysis. In: Ehrgott M, Figueira JR, Greco S (eds) Trends in multiple criteria decision analysis. Springer, US, Boston, MA, USA, pp 209–239 Box GEP (1979) Robustness in the strategy of scientific model building. In: Robustness in statistics. ACADEMIC PRESS, Cambridge, MA, USA, pp 201–236 Burrows DE (1997) Facilitation: a concept analysis. J Adv Nurs 25:396–404. https://doi.org/10. 1046/j.1365-2648.1997.1997025396.x Cegan JC, Filion AM, Keisler JM, Linkov I (2017) Trends and applications of multi-criteria decision analysis in environmental sciences: literature review. Environ Syst Decis 37:123– 133. https://doi.org/10.1007/s10669-017-9642-9 Cestari JMAP, de Loures FR, EAP S (2018) A method to diagnose public administration interoperability capability levels based on multi-criteria decision-making. Int J Info Tech Dec Mak 17: 209–245. https://doi.org/10.1142/S0219622017500365 Cinelli M, Kadziński M, Gonzalez M, Słowiński R (2020) How to support the application of multiple criteria decision analysis? Let us start with a comprehensive taxonomy. Omega 96: 102261. https://doi.org/10.1016/j.omega.2020.102261 de Gooyert V, Rouwette E, van Kranenburg H, Freeman E (2017) Reviewing the role of stakeholders in operational research: a stakeholder theory perspective. Eur J Opl Res 262:402–410. https://doi.org/10.1016/j.ejor.2017.03.079 Dyer JS (2016) Maut—multiattribute utility theory. In: Multiple criteria decision analysis: state of the art surveys. Springer-Verlag, New York, USA, pp 265–292 Eden C (1988) Cognitive mapping. Eur J Opl Res 36:1–13. https://doi.org/10.1016/0377-2217(88) 90002-1 Eden C, Huxham C (1999) Action research for the study of organizations. In: Clegg S, Hardy C (eds) Studying organization: theory & method. Sage Publications, London, pp 526–542 Falcão AP, Machete R, Castilho Gomes M, Gonçalves AB (2021) Spatial multi-criteria decision analysis for rehabilitation priority ranking: a collaborative application to heritage workforce housing sites. Int J Archit Herit 15:790–806. https://doi.org/10.1080/15583058.2019.1650132 Figueira JR, Roy B (2002) Determining the weights of criteria in the ELECTRE type methods with a revised Simos’ procedure. Eur J Opl Res 139:317–326. https://doi.org/10.1016/S0377-2217 (01)00370-8 Figueira JR, Greco S, Roy B (2009) ELECTRE methods with interaction between criteria: an extension of the concordance index. Eur J Opl Res 199:478–495. https://doi.org/10.1016/j.ejor. 2008.11.025 Figueira JR, Mousseau V, Roy B (2016) ELECTRE methods. In: Greco S, Ehrgott M, Figueira JR (eds) Multiple criteria decision analysis. Springer, New York, USA, pp 155–185 Franco LA, Montibeller G (2010) Facilitated modelling in operational research. Eur J Opl Res 205: 489–500. https://doi.org/10.1016/j.ejor.2009.09.030

Mutlicriteria Decision Aiding: Challenges in Real-Life Interventions

185

Frazão TDC, Camilo DGG, Cabral ELS, Souza RP (2018) Multicriteria decision analysis (MCDA) in health care: a systematic review of the main characteristics and methodological steps. BMC Med Inform Decis Mak 18:90. https://doi.org/10.1186/s12911-018-0663-1 French S, Maule J, Papamichail N (2009) Decision behaviour, analysis and support, 1st edn. Cambridge University Press, New York, USA Greco S, Ehrgott M, Figueira JR (eds) (2016) Multiple criteria decision analysis: state of the art surveys, 2nd edn. Springer, New York,USA Henao F, Franco LA (2016) Unpacking multimethodology: impacts of a community development intervention. Eur J Opl Res 253:681–696. https://doi.org/10.1016/j.ejor.2016.02.044 Howard RA, Abbas AE (2014) Foundations of decision analysis. Pearson, Boston, USA Kaner S (2014) Facilitator’s guide to participatory decision-making, 3rd edn. Jossey-Bass, San Francisco, USA Keeney RL (1981) Analysis of preference dependencies among objectives. Oper Res 29:1105– 1120. https://doi.org/10.1287/opre.29.6.1105 Keeney RL (1996) Value-focused thinking: identifying decision opportunities and creating alternatives. Eur J Opl Res 92:537–549. https://doi.org/10.1016/0377-2217(96)00004-5 Keeney RL (2002) Common mistakes in making value trade-offs. Oper Res 50:935–945. https:// doi.org/10.1287/opre.50.6.935.357 Keeney RL (2007) Developing objectives and attributes. In: Edwards W, Miles RF, Von Winterfeldt D (eds) Advances in decision analysis: from foundations to applications. Cambridge University Press, New York, USA, pp 104–128 Keeney RL, Raiffa H (1993) Decisions with multiple objectives: preferences and value trade-offs, 1st edn. Cambridge University Press, New York, USA Keeney RL, Von Winterfeldt D (2007) M13 practical value models. In: Edwards W, Miles RF, Von Winterfeldt D (eds) Advances in decision analysis: from foundations to applications. Cambridge University Press, New York, USA, pp 232–252 Kirkwood CW (1997) Strategic decision making: multiobjective decision analysis with spreadsheets. Duxbury Press, Belmont, CA, USA Labreuche C, Grabisch M (2003) The Choquet integral for the aggregation of interval scales in multicriteria decision making. Fuzzy Sets Syst 137:11–26. https://doi.org/10.1016/S0165-0114 (02)00429-3 Landry M (1995) A note on the concept of “problem.”. Organ Stud 16:315–343. https://doi.org/10. 1177/017084069501600206 Langemeyer J, Gómez-Baggethun E, Haase D et al (2016) Bridging the gap between ecosystem service assessments and land-use planning through multi-criteria decision analysis (MCDA). Environ Sci Pol 62:45–56. https://doi.org/10.1016/j.envsci.2016.02.013 Luoma J (2016) Model-based organizational decision making: a behavioral lens. Eur J Opl Res 249: 816–826. https://doi.org/10.1016/j.ejor.2015.08.039 Malczewski J, Rinner C (2015) Multicriteria decision analysis in geographic information science, 1st edn. Springer, Berlin, Germany Marais A, Abi-Zeid I (2021) A method to identify, characterize and engage relevant stakeholders in decision processes, report FSA 2021–001. Université Laval, Québec, Canada, p 57. https:// www.researchgate.net/publication/351366080_A_Method_to_Identify_Characterize_and_ Engage_Relevant_Stakeholders_in_Decision_Processes Marleau Donais F, Abi-Zeid I, Lavoie R (2017) A loose-coupling integration of the MACBETH approach in ArcGIS. Proceedings of the EWG-DSS international conference on decision support system technology data. Information and Knowledge Visualisation in Decision Support Systems, Namur, Belgium Marleau Donais F, Abi-Zeid I, Waygood EOD, Lavoie R (2019a) A review of cost–benefit analysis and multicriteria decision analysis from the perspective of sustainable transport in project evaluation. EURO J Decis Process 7:327–358. https://doi.org/10.1007/s40070-019-00098-1 Marleau Donais F, Abi-Zeid I, Waygood EOD, Lavoie R (2019b) Assessing and ranking the potential of a street to be redesigned as a complete street: a multi-criteria decision aiding approach. Transport Res A- Pol 124:1–19. https://doi.org/10.1016/j.tra.2019.02.006

186

I. Abi-Zeid et al.

Marleau Donais F, Abi-Zeid I, Waygood EOD, Lavoie R (2021) A framework for post-project evaluation of multicriteria decision aiding processes from the stakeholders’ perspective: design and application. Group Decis Negot 30:1161–1191. https://doi.org/10.1007/s10726-02109753-y Marleau Donais F, Abi-Zeid I, Waygood EOD, Lavoie R (2022) Municipal decision-making for sustainable transportation: towards improving current practices for street rejuvenation in Canada. Transport Res A- Pol 156:152–170. https://doi.org/10.1016/j.tra.2021.12.009 Marttunen M, Mustajoki J, Dufva M, Karjalainen TP (2015) How to design and realize participation of stakeholders in MCDA processes? A framework for selecting an appropriate approach. EURO J Decis Process 3:187–214. https://doi.org/10.1007/s40070-013-0016-3 McCann B (2013) Completing our streets: the transition to safe and inclusive transportation networks. Island Press, Washington, DC, USA Mingers J, Brocklesby J (1997) Multimethodology: towards a framework for mixing methodologies. Omega 25:489–509. https://doi.org/10.1016/S0305-0483(97)00018-2 Montibeller G (2007) Action-researching MCDA interventions. In: Shaw D (ed) Key-note papers, 49th British operational research conference (OR 49). Univ. of Edinburgh, The OR society, UK Morris WT (1967) On the art of modeling. Manag Sci 13:B-707–B-717. https://doi.org/10.1287/ mnsc.13.12.B707 Munda G (2016) Multiple criteria decision analysis and sustainable development. In: Greco S, Ehrgott M, Figueira JR (eds) Multiple criteria decision analysis. Springer, New York, NY, USA, pp 1235–1267 Munier N, Hontoria E (2021) Shortcomings of the AHP method. In: Uses and limitations of the AHP method. Springer International Publishing, Cham, Switzerland, pp 41–90 Norese MF (2020) Profiling analysts and actors in interaction: how behavioural aspects can positively affect the decision aid process. EURO J Decis Process 8:125–150. https://doi.org/ 10.1007/s40070-020-00113-w Norese MF, Carbone V (2014) An application of ELECTRE tri to support innovation: an application of electre tri to support innovation. J Multicrit Decis Anal 21:77–93. https://doi.org/10. 1002/mcda.1508 Ormerod RJ (2018) The logic and methods of OR consulting practice: towards a foundational view. J Oper Res Soc 69:1357–1378. https://doi.org/10.1080/01605682.2017.1392407 Phillips LD (1984) A theory of requisite decision models. Acta Psychol 56:29–48. https://doi.org/ 10.1016/0001-6918(84)90005-2 Phillips LD, Bana e Costa CA (2007) Transparent prioritisation, budgeting and resource allocation with multi-criteria decision analysis and decision conferencing. Ann Oper Res 154:51–68. https://doi.org/10.1007/s10479-007-0183-3 Phillips LD (2007) Decision conferencing. In: Edwards W, Miles RF, Von Winterfeldt D (eds) Advances in decision analysis: from foundations to applications. Cambridge University Press, Cambridge New York, USA, pp 375–399 Roy B (2016) Paradigms and challenges. In: Greco S, Ehrgott M, Figueira JR (eds) Multiple criteria decision analysis. Springer, New York, New York, NY, USA, pp 19–39 Roy B, Figueira JR, Almeida-Dias J (2014) Discriminating thresholds as a tool to cope with imperfect knowledge in multiple criteria decision aiding: theoretical results and practical issues. Omega 43:9–20. https://doi.org/10.1016/j.omega.2013.05.003 Roy M, Prévost P (2013) La recherche-action : origines, caractéristiques et implications de son utilisation dans les sciences de la gestion. Recherches qualitatives 32:129–151. https://doi.org/ 10.7202/1084625ar Simon HA (1960) The new science of management decision. Harper, New York, USA, p 56 Smith JE, Dyer JS (2021) On (measurable) multiattribute value functions: an expository argument. Decis Anal 18:247–256. https://doi.org/10.1287/deca.2021.0435 Therrien Y (2011) La CUCQ, véritable chien de garde de l’urbanisme. In Le Soleil on 27 August 2011. https://www.lesoleil.com/2011/08/27/la-cucq-veritable-chien-de-garde-de-lurbanisme-f4 908f1b85a86e268803c74a1a91cc2f

Contrasting Applications in Environmental Planning and Public Procurement Jacques Pictet and Dominique Bollinger

1 Introduction Real-world applications of Multiple criteria decision aid (MCDA) present similarities that correspond to the basic tenets of the field, paraphrasing Roy (1985): aiding people to formally evaluate discrete alternatives on several criteria using their characteristics and performances, to elicit the criteria relative importance and to usually compare them using an aggregation method to possibly recommend further action. Beyond that, every aspect of an application differs greatly. Who are the actors? What is the history of the project? What is the time horizon? What are the legal constraints? How much time is available to reach a decision? Who will face the consequences if things turn sour? Is it possible to obtain exploitable results even with the uncertainty involved? This chapter presents two types of applications—in Environmental Planning and Public Procurement—that are quite different. A specific methodology was developed for each one and the chapter proposes a common framework, which includes four key elements, to describe the different aspects of these methodological approaches: • Context: Types of problems at hand, political and legal situation at the time, state of the art that influenced how the specific methodology developed.

J. Pictet (✉) Conseils en aide à la décision, Lausanne, Switzerland e-mail: [email protected] D. Bollinger Haute école d’ingénierie et de gestion du canton de Vaud, Yverdon, Switzerland © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_7

187

188

J. Pictet and D. Bollinger

• Content management: MCDA model core (alternatives, criteria, evaluations, and weights), MCDA aggregation method and optional tools to interact with the actors (e.g., Problem structuring and Group decision), other aspects of the project. • Process management: Insertion in the general process, actors, roles distribution, steps, “negotiation.” • Outcomes: Recommendation, decision, implementation.

2 Applications in Environmental Planning A methodology was developed at the Institute for environmental engineering of the Swiss school of technology (EPFL) in Lausanne, Switzerland, mainly by a small team (Lucien Yves Maystre {, Jean Simos, and the two authors). This institute was founded in the early 1970s and adopted upfront a broad understanding of engineering. The integration of the economic, social,1 and environmental dimensions was a central issue (Maystre 1985)—long before the concept of sustainable development was coined—and a formal way to do this was needed. The constructivist MCDA approach and the ELECTRE methods (Roy 1985; Roy and Bouyssou 1993) were chosen for tests.

2.1

Context

Environmental issues were quite hot in the mid-1980s in Switzerland, like in many countries. A new law (1983) and specific regulations, including one on Environmental impact assessment (EIE) for large projects (1988), brought environmental evaluation to the forefront.2 The period was also favorable in the country for broader participation of stakeholders in public projects of this type. Several approaches were proposed and some were tested (Ruegg et al. 1992). The first application of the methodology was proposed within a broader project on waste management for the County of Geneva—centered on the handling of urban wastes and their possible treatments—and many aspects tested there were then adopted for the emerging methodology, as the actors were quite pleased with the handling of the project and the outcome (Simos 1991). It was then used for other projects in the broader field—for instance, in hydrology, transportation, urban, and regional planning—and some cases were published along a detailed presentation of the ELECTRE methods for non-mathematicians (Maystre et al. 1994).

1

The social dimension and criteria often remained the poor relative in such studies, due to the limited interest of the clients, methodological issues, and the lack of expertise of the analysts. 2 Even though, unlike other countries, Switzerland decided not to include alternatives assessment as a compulsory part of the process. The similarities and differences between EIE and other environmental tools, such as eco-balance and ecolabel, were analyzed in (Pictet 1996).

Contrasting Applications in Environmental Planning and Public Procurement

189

The cases dealt with different aspects of the environment—e.g., air, water, soil, and nature—in connection with the activities that disturb it—e.g., transportation, agriculture, urban planning, energy, and wastes—at different scales—from a single plant to a whole region—for different purposes—e.g., planning, plant location, and choice of technology. Pictet and Simos (1996) also proposed MCDA as a more formal way to operationalize the concept of “balancing the interests,” as defined in Article 3 of the Swiss ordinance on urban planning (Confédération helvétique 2000): (a) to describe the parties impacted by the decision and their legitimate interests, (b) to assess and prioritize these interests in relation to the object of the planning procedure, (c) to propose a decision based on this analysis. Similar concepts exist under different names in many countries for various fields (e.g., “reasonable accommodations” in Canada). But many experts thought and still think that this process should be handled in an ad hoc approach and that an excess of formalism could be counterproductive (see, e.g., DTAP (2017)).

2.2 2.2.1

Content Management Alternatives

These applications were often associated with technical enquiries. New technologies had to be analyzed, legal plans (e.g., for water and air protection) drawn, and new plants built. These studies provided the bases for the elaboration of alternatives and criteria. To overcome the “technical gap” between experts and non-experts, visual displays were often used to communicate about the procedure adopted to elaborate alternatives, and the place and function of proposed partial solutions as a whole (Fig. 1). Such systemic representations of environmental and technological components were used in these applications, as they offered a middle ground, by focusing on the broad issues and avoiding many of the technicalities. Decision-makers and experts were compelled to make their input understandable by the “other side,” as part of the building of a “common vocabulary,” necessary in any case. The representation in Fig. 1b was also helpful to summarize fluxes analysis—an input-output analysis of each component using transfer functions allows to know fluxes state, mass, volume and chemical elements concentration—that were sometimes hard to follow.3 The alternatives were intended to reflect the breadth of partial solutions discussed and some of the possible combinations. Rather large sets of alternatives were common, and part of the job was to maintain them at a manageable size, on average 3 For instance, it is possible to show if a component allows to isolate a toxic substance (using the breadth of the arrows to visualize it in a “fluxosaur” as we labeled them). Putting all components analyses together constitutes an input-output analysis of each alternative, that will be used in the alternatives evaluation phase.

a

J. Pictet and D. Bollinger

Main processes 1

2

3

Detailed processes 4 5 6 7 8

9

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

b

Sludge

Ashes

Water

Alternatives

190

Back solution

Inertization Extraction

Sludge Water Electrolyze

Solidification

Solidified residues

Water treatment

Heavy metals

Water

Transport

Sewer sludge treatment plant

Recycling

Residues land-filling

Fig. 1 Different way to explain how alternatives are generated, defined by the presence or absence of a component in an alternative: (a), for the treatment of urban wastes, as a tree to show how they differ on main components (left) and then as a binary table with more details (right), (adapted from Simos (1991)); (b) for the treatment of waste incineration residues as a network of components, to detail how they interact for a given alternative, notably the main resulting fluxes (Bollinger and Pictet 2008). Note that (b) is the equivalent of one alternative in (a)

Contrasting Applications in Environmental Planning and Public Procurement

191

Table 1 Example of a set of criteria and groups in an urban/transportation planning case Effects on transportation Access to the area Compatibility with an attractive public transportation offer Attractivity for transit traffic Road users’ safety Effects on the environment Noise Distances made by cars Impacts on natural spaces Political acceptability Solidarity among municipalities Political support

Effects on urban planning Total surface needed Number of buildings and flats impacted Subdivision of the rural area Impact on heritage Financial consequences Investments Maintenance costs

10–15 items in each set. Some cases involved many more alternatives, such as road segments to be prioritized as needing intervention to reduce air pollution or noise.

2.2.2

Criteria

Criteria were organized in groups (Table 1), to ease communication, to control the natural tendency to multiply criteria, and to reduce the “mechanical” bias in their weighting. This bias occurs as usually some groups contain more criteria than others. Weighting all criteria in one step would mechanically influence the importance of some of them. Weighting the groups and then the criteria within each group allows to clarify this issue. In a few cases, these procedures were performed in parallel to exemplify the impact on the results. The specific criteria varied from one case to the other, but the groups were often dealing with the effects on human activities and on the environment, the implementation phase (e.g., duration and easiness), costs, and acceptability.

192

2.2.3

J. Pictet and D. Bollinger

Evaluations

Evaluating future performances is based on the present state and the future state at the planning horizon in present settings (status quo ante) and required the use of some of the following tools: • Forecasting: Estimating a state of a parameter or criterion, using trends and discounts. • Simulation: Analyzing the behavior of each alternative (e.g., transportation or pollutants fluxes in anthropic and/or natural environments, like in Fig. 1), to check its consistency with the objectives of the project and the regulations, and defining its future state. • Scenario analysis: Analyzing the elements on which the actors have no influence, e.g., climate, commodities prices, and political decisions. To avoid the explosion of information, we sometimes used simplified “Best case” and “Worst case” scenarios, leading to two evaluation matrices. Or, in a case, an upcoming political decision at the national level about border crossing was seen as the main factor impacting some alternatives at the county level (Table 2). The use of these techniques generates uncertainties that have to be handled in the robustness analysis. Even with the best efforts, the possibility always remains of an “unknown unknown” —overlooked by definition—that would render the exercise pointless. But we tried to handle “known unknown” and avoid “unexpected consequences” that were foreseeable.

2.2.4

Weights

Treating the DMs as peers (§ 2.3) led to the proposal of using individual weights, leading to individual results, that would serve as a basis for the recommendation (see below). To elicit them, a simple analogic procedure with cards was developed and used extensively. Later, it was adapted to elicit evaluations for criteria without natural scale, as the first steps are quite similar (Box 1).

(a) Direct Inter Sud-Ouest Sud Troinex (b) Direct Inter Sud-Ouest Sud Troinex

82 82 66 66 50

77 77 56 56 40

47 61 39 25 29

-3 11 -11 -25 -21

219 188 199 177 177

219 188 199 177 177

Transportations 1.1 1.2 1.3

-26 -28 -27 -30 -29

-22 -24 -22 -25 -24

1.4.1

2 -5 0 -4 -2

1 1 7 3 1

1.4.2

-191 -165 -151 -141 -144

-126 -119 -87 -88 -94 100 95 80 60 0

100 95 80 60 30

Environment 2.1 2.2

90 160 70 20 0

90 160 70 20 0

2.3

-6 -2 -11 -19 -7

-6 -2 -11 -19 -7 -10 -3 0 0 0

-10 -3 0 0 0 140 140 70 20 0

140 140 70 20 0

Urban planning 3.1 3.2 3.3

-5 -1 -3 -6 -5

-5 -1 -3 -6 -5

3.4

-21 -7 -6 -10 -10

-21 -7 -6 -10 -10

-169 -47.8 -88.7 -16.4 -65.5

-169 -47.8 -88.7 -16.4 -65.5

Finances 4.1 4.2

0 1 2 0 0

0 1 2 0 0

4.3

-15% -15% -15% -14% -14%

-17% -17% -17% -16% -15%

-109% -103% -92% -96% -101%

-106% -102% -93% -97% -102%

Acceptability 5.1 5.2

Table 2 Matrices based on scenarios with (a) and without (b) the “border crossing effect.” Colors in the original report (that cannot be reproduced here) divide each criterion scale (in columns, see Table 1) into three segments of equal length, which provides an overall view and helps get a sense of (i) the “profile” of each alternative (in rows), (ii) the impact of the scenarios on each criterion (some were not affected), and (iii) the impact of the scenarios on each alternative (some were more impacted than others)

Contrasting Applications in Environmental Planning and Public Procurement 193

194

J. Pictet and D. Bollinger

Box 1 Simos’ card method This practical method for the elicitation of weights is rather simple to implement as it requires only pieces of paper (cards) or stickers (Simos 1991). Its use by the DM is also rather simple as no computation is needed (at first). The basic steps are: (1) order the criteria by decreasing order of importance (from A to Z); (2) insert blank cards at will to amplify some differences; (3) characterize the ratio f = importance of A / importance of Z. This third step was missing from the original method, and Figueira and Roy added it later (2002). To overcome the slightly higher complexity of this step, dual questions allow to elicit an initial value of the ratio f, of the kind: “If you give 40% to A, how much would you give to Z? 10%, 5%, 1%, other?” and “If you give 3% to Z, how much would you give to A? 15%, 25%, 50%, other?” A simple affine transformation of the non-blank inverse “ranks”—normalized to 100%—provides the DM with initial weights (Table 3). A simple spreadsheet allows them to “play around” and possibly revise their initial answers. The original cards method was later adapted to elicit a notation of the alternatives on criteria without a natural scale (Pictet and Bollinger 2008). Cards, the items themselves (e.g., cups of soup in one case) or a symbol (e.g., a photograph or a drawing in another one) can be used. The two first steps are identical: ordering the items and adding blank cards. In the third step, at least two reference cards are inserted to prop the scale into more absolute terms. A similar affine transformation is used to obtain an interval scale—somewhere between an ordinal and a cardinal scale—providing numerical grades. It can advantageously be combined with a method often used in practice: the verbal scale, easy to use, but with known threshold effects.4 The proposed combination improves the overall consistency, within a framework that practitioners know and use (Fig. 2). The mathematical soundness of this method is dubious, but it allowed every DM we encountered to provide weights with a good level of understanding and confidence, sometimes with added support.

2.2.5

Aggregation Method and Results Presentation

The choice of ELECTRE methods to aggregate the evaluations was generally well accepted, perhaps thanks to the academic status of the analysts, even if people had problems understanding its intricacies. The acceptability could also be explained by

4

There are two main effects: Exaggerating minor differences and minimizing larger ones. Examples can be found in the jurisprudence about public procurement (see below) when cardinal criteria were transformed into notes, following a similar logic. Judges in some cases could not exclude that it was intentional.

Contrasting Applications in Environmental Planning and Public Procurement

195

Table 3 Example of the calculation of the weights elicited with the cards method. Note that (i) joint alternatives are ranked at the average of their “ranks,” (ii) the absolute weights respect the choice of the ratio f (here 8/1) and the interval between “ranks” (here 0.78 by “rank”), (iii) their sum excludes the “ranks” of the blank cards (stricken out), (iv) relative weights are rounded. A graph would allow to visualize all this instantly (not presented here) Direct rank 1 2 3,5 3.5 5 7 7 7 9 10 f=

Inverse rank 10 9 7.5 7.5 6 4 4 4 2 1 8

(a)

Criteria C3 – C6 C2 – C4 C7 C1 – C5 Sum

Absolute weight 8,00 7,22 6,06 6,06 4,89 3.33 3.33 3.33 1,78 1,00 31,11

(b)

(c) Alternative C

Alternative C Very good Alternative B

Very good

Good

Good Alternative A

Relative weight 26% – 19% 19% – 11% 11% 11% – 3% 100%

Alternative B

Alternative F Alternative A

Alternative F

Medium Medium Alternative G Alternative D

Alternative E

Alternative G Bad

Alternative D

Alternative E

Bad Very bad

Very bad

Fig. 2 Building the interval scale of a criterion using (a) an “augmented preorder” of the alternatives, (b) an order with the verbal labels positioned relative to the alternatives; (c) after some adjustments, the “equidistance” among successive verbal labels—that practitioners expect— is respected, allowing to generate an interval scale that provides the grades for the alternatives (not detailed here)

the way they were implemented, as the decision-makers liked the idea of the visual presentation of the individual and collective results (Box 2). Individual results allow to visualize the robustness analysis—how the uncertainties influence the results for

196

J. Pictet and D. Bollinger

each DM—and collective ones (by overlay) show how individual weights impact the results overall, besides the uncertainties. Box 2 Synthesizing multiple ordinal results Multiple ordinal results can be synthesized by providing for each alternative a value (e.g., the median of the ranks) or a range. A dominance analysis is also possible: alternative A dominates B if it is always as good as B and at least once better. In practice, a lesser proportion (e.g., 75%) can be used to implicate most alternatives. With dual results—e.g., two complete preorders like in ELECTRE methods— the dominance analysis should be completed with an incomparability analysis. Alternatives A and B are incomparable if A dominates B in one result and B dominates A in the dual result, in a significant majority of results. A table or a synthetic partial preorder can summarize these relations. A presentation of such graphs—e.g., with decreasing levels of majority of results but an increasing number of alternatives involved (Fig. 3, left)—can help the decision-maker understand the “strength” of each relationship and its impact on the synthesis. Comparing the two shows a rather similar pattern: Alternative 3 occupying the top-right corner (dominating many alternatives), Alternative 9 the top-left corner (incomparable with most alternatives), Alternatives 2, 7, and 8 the bottom and left ranges, and the others scattered in-between. Simos (1991) proposed a different approach, named Surmesure (Pictet et al. 1994; Rogers et al. 2000), that plots, for each DM, the dual ranks of each alternative on a plan and then overlay the plans (Fig. 4). The main interest of SURMESURE is that a visual display is more intuitive than calculation, and the DMs can quickly check that their dots are present.

2.3

Process Management

These applications were usually managed by a working group, about 10 people on average. Representatives of public agencies (the client and related ones) and other stakeholders (e.g., associations, and lobbies) formed the group of decision-makers (DMs). They were considered as peers, for several reasons: • Psychologically, it eases the discussion, making all suggestions a priori legitimate and debatable. This brainstorming principle was completed with a “constructive demolition” rule stating that any item adopted by the group could not be removed unless another item was proposed and accepted to replace it. • Mathematically, it allows a simple visualization of all the individual results by overlay (see Fig. 4) that drastically diminishes the amount of calculation needed and appeals to the visual capacity to discern patterns; it also avoids the delicate question of putting a weight on people to operationalize their alleged differences of status, if their input were to be aggregated (see below).

Contrasting Applications in Environmental Planning and Public Procurement

(a)

197

(b) A9 A3

A9

A5

A3 A6

A4

A5 A1

A8

A6 A8

A1 A7

A4 A2

A7 A2

Fig. 3 Representing multiple dual ordinal results (e.g., ELECTRE): (a), A synthetic partial preorder, mentioning dominance relations by decreasing percentage of presence in the results (e.g., >90%, >80%, >70%, and > 60%); incomparability is implicit (absence of arrows); (b) a plane on which alternatives are located according to their ranks in the dual results (SURMESURE, see Fig. 4); relations are implicit; note that alternatives are represented here as a single point for clarity

• Politically, it reduces the importance of the working group composition, as there is no vote. The diversity of opinions was as large as possible and a particular attention was given to a potential “Wicked fairy godmother,” i.e., somebody who could use its nuisance power if not included in the process. Technical experts often had a leading role in the propositions, and the working group had the final say. Discussions allowed the development of a common vocabulary and the sharing of a general understanding. If needed, further enquiries clarified specific points. Genard and Pirlot (2002) argue that the role distribution has to be consistent with the main realms of legitimacy and the type of contribution they can provide: • True-false axis: The truth statement belongs to the knowledge sphere, and should be provided by experts. • Good-bad axis: The value judgments on this axis belong to the political sphere, and should be provided by people representing an authority or an interest group. Besides, Belton and Pictet (1997) proposed a typology with three ways to manage the evaluations and the weights in Group MCDA, based on examples found in the literature5: • Sharing: Discussion goes on until the DMs reach a consensus. • Aggregating: Individual inputs are recorded and somehow aggregated. • Comparing: Individual inputs are recorded and used to produce individual results that are then compared. In our cases, every aspect was managed in a different way, depending on its need for legitimacy: The silent negotiation (see Box 3), does not fit into this typology, being between sharing and aggregating.

5

198

J. Pictet and D. Bollinger

Alternative 2 descendant

descendant

Alternative 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

ascendant

ascendant

Alternative 4 descendant

descendant

Alternative 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

ascendant

ascendant

Alternative 6

descendant

descendant

Alternative 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

ascendant

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

ascendant

Fig. 4 Example of overlay of ELECTRE individual results presented by alternative, showing relative positions (best on top-right corner), compacity (size of the “cloud”), and incomparability (distance to diagonal) (Bollinger and Pictet 2008)

Contrasting Applications in Environmental Planning and Public Procurement 0,0%

10,0%

20,0%

30,0%

40,0%

50,0%

199

60,0%

Transportation

Environment

Urban planning

Minimum Average Maximum

Finances

Acceptability

Fig. 5 Basic statistics about individual weights awarded by the DMs to the five groups of criteria of Table 1, used as collective feedback to protect the anonymity, but not used in the calculations

• Alternatives and criteria were often proposed by the experts, then discussed, completed, and validated by the DMs. • Shared set(s) of evaluations were often established by the experts and validated by the DMs.6 • Weights were defined individually by the DMs (Fig. 5). • Aggregation was performed by the analysts, providing DMs with individual results and a presentation of the overlayed (anonymized) results (Fig. 4). • Recommendation: Proposed by the analyst, based on the results, discussed, amended, and approved by the DMs.

2.4

Outcomes

Adopting a participatory approach was quite risky for the client: they temporally abandoned their leading role to leave space for third parties’ inputs and they were supposed to accept and implement a recommendation not necessarily aligned with 6

Usually, the work was divided among the experts, based on expertise and workforce, and discussed internally before presentation to the DMs. Ensuring the overall consistency of the evaluations can be an issue when they are the result on an aggregation. In one case about public procurement (see below), evaluations, on some criteria, were calculated as the average of several experts’ evaluations; a tribunal found that the evaluations were more statistically correlated with each expert than with each alternative, and saw it as a problem.

200

J. Pictet and D. Bollinger

their own priorities. That is why they sometimes used the recommendation in the most favorable way: they usually did not contradict it, but exploited every uncertainty in the recommendation, e.g., favoring an alternative among the good ones or proposing a new one mixing some of their features. Similar phenomena occur at the upper echelon, as officials have to cope with larger issues and do not necessarily feel constrained by the recommendation. For instance, some seemed to practice a form of “territorial democracy,” i.e., locating plants with negative impacts in areas exempted so far. Moreover, if something concrete emerges from the executive branch, it usually passes in the hands of the legislative one and sometimes the judiciary one that can modify it in substantial ways, not to mention the popular vote, in some cases. Judging the success overall needs shades of gray. Some projects were implemented as such or amended, and others served as a marker for a policy change or opened the door for new solutions. In one case, the implemented project was deemed impossible during the whole process, until an agency reversed its policy just before the results presentation, making it possible to choose an option favored by many actors (Maystre et al. 1994). In another case, the results did not impress much the villa owners impacted by the proposed changes in a city’s planning legislation. Their skepticism about the entire process was very palpable during the final meeting. But later, city officers stated that the long-time broken dialogue restarted almost immediately after that. This general framework was applied in many cases, except the ones described below, that differed on one or several aspects. While discussing the fate of a city’s main administrative building, authorities invested a large group of public servants—as the main group of regular users—with the evaluations. A minority political party was quite vocal about an alternative that nobody else favored. Unsurprisingly, the evaluations highlighted its many drawbacks and its few positive aspects, and the city chose an alternative among those favored by almost everybody. The analysts accepted this participatory exercise of a special kind but warned that it would be counterproductive to bias the evaluation. In another case, county officers had prepared a transportation project up to the final authorization documents, with every “t” crossed and every “i” dotted. At the last minute, a concerned municipality produced an alternative proposal. Due to the lack of time, both teams evaluated their own alternative against the other one. The intention was to see the light beyond the biases. The presentation of both sets of evaluations was almost comical: each team found their project to be better than the others on all criteria. Surprisingly (for the analysts), the county side (8 senior officers) started losing its composure while the municipality side (the mayor and two sidekicks) had a growing smile on their faces. The county officers proposed a compromise solution during the meeting, but it was dismissed on the spot and 3 days later the county authorities accepted the municipality’s proposal. It is still unclear what really happened. Clearly, the county project failed to consider a major change in the municipality planning policy, undermining the rationale for their chosen alternative. Possibly, the mayor’s political clout also played a role, as he became a county minister in the following years.

Contrasting Applications in Environmental Planning and Public Procurement

201

A consultative body requested a counter expertise in a large remediation case driven by private companies. It showed the lack of seriousness of the technology comparison, without discussing the evaluations themselves. The MCDA analysts were publicly criticized by the industry (for not being experts in the technology field), but their pet technology—untested at such a large scale, but much cheaper than existing ones—was put aside in favor of a more expensive and more reliable one. The dynamic can be quite different when decision-makers have a professional stake in the decision. It is tempting to use MCDA as a negotiation tool, i.e., by putting together representatives of divergent interests and trying to resolve their differences by identifying a good alternative. Examples can be found in (Maystre and Bollinger 1999) and one was translated by Rogers et al. (2000). For instance, in two different regions of Switzerland, representatives of regional waste management authorities—who had their own projects—participated in a process, alongside some representatives of supra-regional authorities. The outcome was not straightforward but lead to useful adaptations in the original regional projects.

3 Public Procurement This methodology was developed in the small company the authors founded after they left EPFL. Public procurement is the acquisition of goods and services by public authorities, from color pencils to insurance to tunnels. The contract is awarded, usually to a private company, following a procedure that ensures increasing concurrence (e.g., number of participants and opening to foreign companies) as the size (price) of the contract goes up.

3.1

Context

Contracts have been awarded for ages, often based on loose methodologies and national legislations. In 1995, a major change occurred in the 1947 General Agreement on Trades and Taxes (GATT) with the creation of the World Trade Organization (WTO). This occurred in a period of globalization, in which barriers to foreign companies were shunned upon. GATT new rules (1995) clarified the way to adjudicate contracts, mainly by formalizing how calls for tender had to be publicized and who could participate. The explicit indications of elements to be used to adjudicate the contract were minimal—“criteria,” “weights,” “order of preference”—without explicit mention of aggregation methods or specific procedures. Signatories had to adapt their legislation at the multilateral (e.g., European Union), national and subnational (e.g., Swiss counties) levels, and did so in rather

202

J. Pictet and D. Bollinger

similar minimal terms. The analysis here is limited to the European Union and Switzerland. Early Swiss jurisprudence on the new law made it clear that previous decisions about school grading7—i.e., applying a form of weighted average and how to do it properly—would serve as a basis for public procurement ones. Entities elsewhere may have been less clear about this link, but past practice led to similar choices. This false sense of knowledge at least partially explains why the GATT agreement, national laws and jurisprudence, and adjudication decisions, did not involve mathematical expertise. Proposed support to develop a minimal methodology was turned down in several countries by various authorities and professional associations, who often produced half-baked guidelines. Overall, the adaptations were made mostly by lawyers who had many issues to address, and it is not completely surprising they tended to leave aside those they understood the less. Practitioners had to fend for themselves, and many points were clarified—sometimes wrongly— by jurisprudence. Three major points were overlooked. From the “lowest bidder” to the “best offer” Past practice, centered on the “lowest bidder,” i.e., the less expensive offer respecting the minimal requirements, became history. Voluntarily or not, the GATT rules hindered this practice, without excluding it, by treating all criteria almost in the same way. This was the interpretation in Europe, while some other countries (e.g., USA) more routinely manage the price aside from the other criteria. Some proposals on the use of the price and the handling of “abnormally low tenders” can be found in Ballesteros-Pérez et al. (2015). Identifying the “best offer” requires a parallel approach while the previous practice was sequential: conformity check to requirements one by one, then selection of the lowest price. From performance to score For the sake of simplicity, everybody adopted simple transformations of the performances into scores within bounds. For increasing criteria, this could be achieved easily using a linear transformation; for decreasing criteria—including the price— the same principle was applied, based on the inverse of the performances, overlooking its non-linearity and asymptotic nature. This lack of uniformity did not bother people much, even though a marginally more complicated linear formula transformation is possible (see Fig. 6). Global versus local references A major feature of school grading got lost in the process, at least as a principle: the transformation of scales based on “global” reference points instead of “local” ones. Although imperfect, it tries to ensure that a student would receive a similar grade in a

7

Switzerland follows the general framework dominant at least in Western Europe and for decades, with variations, using numerical grades and importance coefficients.

Contrasting Applications in Environmental Planning and Public Procurement

203

Max

Note

Lowest offered price

Min Low reference price

High reference price

Price

Fig. 6 Transformation of prices into notes using a global scale—based on two reference points independent from the bids (white circle)—and a local scale—based on the lowest offered price (black circle). Note the asymptote visible on the two examples of the latter, based on different lowest price

different class or with a different teacher.8 If her classmates’ grades were to be used as references, the result might vary dramatically: for instance, in a class full of students with a good command of German, she would receive low grades in that branch, while, with the same performance, she would receive very good grades in a class with poor command of this language. Proponents of the local reference points put forward its apparent neutrality but overlook its drawbacks. The intention seems to have been to “let the market talk”—mixing it with a sort of benchmarking exercise—and led to serious mistakes. Pictet and Bollinger (2003) addressed these and other issues in a book by analyzing many real examples from the early jurisprudence (mainly from Switzerland) and proposing minimal changes to current practices to ensure a simple but sound approach.9 Swiss courts never mentioned it, but a preliminary article (Pictet and Bollinger 2000) was cited a few times. In France, a guide from the legal service of the Ministry of Economy summarized its vision about the transformation of prices into scores, including some mathematical errors (Direction des affaires juridiques 2013). Even the Conseil d’Etat (2012) the highest administrative court of this country, made a basic mistake about the mathematical inappropriateness of negative notes (for a counterexample, see Bana e Costa et al. (2002)).

8

Schools are trying to reduce this impact by having common examinations on a large scale or having teachers grading the same copies and discussing the main differences to better understand their origins. 9 An easy-to-use spreadsheet file (MP2) was freely available to allow the implementation of the proposed approach (evaluations, weights, aggregation, robustness analysis).

204

3.2

J. Pictet and D. Bollinger

Content Management

Content management was impacted by the inertia deriving from past practices. It was difficult to propose even minor changes that would greatly improve the quality of the decision from the point of view of MCDA. For instance, simple additive methods (i.e., weighted average) have been used all along and nobody seriously suggested to change that, but the way it is implemented has a strong effect on its meaningfulness. The legislation imposes a major constraint that stems from the obligation to publish the criteria and a single set of weights in the call for tenders. Mathematically, it means building and testing a complete multicriteria model, even before the initial publication of the call for tenders; this means: • Defining a set of fictive, but realistic, alternatives that cover the possible range of the real offers. • Defining the set of criteria that will be used. • Defining the set of weights.10 To make the weights meaningful, a uniform scale, common to all criteria, is necessary. Without it, scaling constants—i.e., information mixing the importance of the criteria and differences among criteria scales— would be necessary, that would ruin the information provided to the bidders about which aspects are important or less so. • Defining the transformation formula from performances to notes for the quantitative criteria and the ways to attribute notes for qualitative ones. The still popular transformation based on extrema performances (local scale) represents a risk, as the model cannot be tested properly. One way to open markets deals with ensuring a minimal level of transparency. Weights publication is an important means to this end. But the transformation of performances into notes might also have a major influence on the outcome. That is why it is important to precisely describe the formulas in the call publication, or at least in a document that can be proved to be dated before the opening of the envelopes. One reason to choose the second option deals with the possibility for tenders to tamper with the process (collusion) which is easier if the model is fully known. The cases stemmed from services for municipalities to major plants for counties, mainly about solid waste transportation and treatment, and energy production. The set of alternatives—the offers—varied a lot, from 2 to more than 10. There were discussions about the adequate number of criteria. Some experts, e.g., in Switzerland, suggested that three criteria were enough, others, e.g., in France, insisted on a two-criteria structure: the price and a “technical value” regrouping all other aspects as sub-criteria. In our practice, we used on average five criteria, with sometimes sub-criteria. One way or the other, two main components have to be considered: 10 In some legislations (e.g., European Union), weights can be published as intervals. Jurisprudence stated that the order of importance was not enough and put limits to the acceptable ranges.

Contrasting Applications in Environmental Planning and Public Procurement

205

• The characteristics of the product or service proposed. • The capacity of the bidder to deliver the goods. Depending on the importance of the contract—and the legal rules depending on it—the procedure to manage the second point can differ greatly. In smaller contracts, the bidders’ capacity is treated as a preliminary constraint in the bids comparison (one-step procedure). In larger ones, the first step deals with this issue; then, accepted bidders can submit a bid in the second step. In most cases, criteria deal with experience, competence, understanding of the problem, and organization of the team/company. Characteristics of the offer usually involve criteria about costs, time, and quality of the outcome. Note that the two first types of criteria are usually decreasing and the third one is associated with an ordinal scale. A proper handling of the evaluation process is therefore crucial. The handling of the price was an almost unique subject of interest on the topic of evaluation. Most countries favor the simplest formulas, that are local and non-linear (creating an uncontrolled asymptote) (Fig. 6). It is based on the assumption that the least expensive acceptable offer should absolutely receive the maximal evaluation on the price, and not only the highest awarded one.11 In other words, it induces the use of a local scale—at the upper end—and prohibits the use of a global one. The specific treatment of the price as a local scale—most other criteria, including decreasing ones (e.g., based on time) use global scales—creates a heterogeneity that nobody has been able—or even willing—to justify. To this day, this inconsistency is fully respected in Switzerland and France, and certainly elsewhere in Europe. Dozens of formulae for the transformation of the price have been proposed; a few have been deemed unsuitable. For instance, those based on a distance from the average of the offered prices were rejected in favor of those using the lowest bid as a reference. Nowadays, for instance, two formulae are recommended in Switzerland (Conférence romande des marchés publics 2021) that are also used in France (Direction des affaires juridiques 2013): In Pictet and Bollinger (2003), we recommended the following formula (see Fig. 6, solid line): Offer Y grade =

Grademax if Offer Y price ≤ Low reference price Grademin if Offer Y price ≥ High reference price High reference price - Y price elsewhere ðGrademax - GrademinÞ ðHigh reference price - Low reference priceÞ

where

11 In Switzerland, it stems from an early interpretation of one of the goals of the law stating that “the adjudicator must choose the most economical offer” (our translation). In other cases, people informally refer on a regular basis to the “price-quality ratio,” coming from the cost-benefits analysis, which cannot be applied—and is not—in this field.

206

J. Pictet and D. Bollinger

Grademax if Offer Y price < = Low reference price Offer Y grade = Grademin if Offer Y price > = High reference price ðGrademax—GrademinÞ ðHigh reference price—Y priceÞ= ðHigh reference price—Low reference priceÞ elsewhere • • • •

Grademax: Maximum grade Grademin: Minimum grade High reference price: Reference price awarded the minimum grade Low reference price: Reference price awarded the maximum grade

A strange transformation formula, seemingly stemming from Germany, also appeared in France: it mixes weights and notes beyond recognition before the final aggregation (Direction des affaires juridiques 2013).

3.3

Process Management

The groups managing these processes were generally small and their members were working together on a regular basis: elected people, public servants, and their experts. In the simpler cases, they were DMs and experts at the same time. Discussions about the objectives of the project were behind them and the focus was on who would be awarded the contract. To facilitate the elicitation of common weights from a group of often like-minded people, Pictet and Bollinger (2005) designed a procedure named silent negotiation, based on the modified Simos’ card method (Figueira and Roy 2002): decisionmakers could in turn modify the present rank of a couple of criteria, over several rounds, without discussion (Box 3). The outcome could then be discussed and the remaining divergences managed, sometimes as intervals of weights or by compromise. It is a time-saving procedure that focuses the discussion on solving divergences and limits its less useful functions (e.g., establishing the picking order, perturbing other members’ expression). Box 3 Silent negotiation The silent negotiation is an iterative voting procedure to obtain shared information that follows the principles of Simos’ card procedure. Decision-makers can, in turn, promote or demote a given number of items—criteria to weight or alternatives to note on a given criterion—to fit their preferences. It is a negotiation, as DMs can react to the other’s moves and adopt a strategic behavior. Some features will be quite consensual, others will need some back and forth to settle (Fig. 7). It works well in the context of “light (continued)

Contrasting Applications in Environmental Planning and Public Procurement

207

Box 3 (continued) negotiation”: DMs agree that they need this collective information to go ahead. The number of rank changes on each round depends on the number of items and DMs, and the number of rounds on the willingness of the DMs to go on. The remaining disagreements can then be discussed and usually treated as uncertainty (e.g., range of values). Time pressure was always present, due to the need to start the contract as soon as possible. A recourse was seen as catastrophic, as it often signifies months of delay. Moreover, the legislation mentions several strict deadlines. The usual solution is a (simple) PERT diagram that clarifies when to start each specific task.

3.4

Outcomes

5 4 3 2 1 0 -1 -2 -3 -4

0

2

Rounds 4

6

8

A B C D E F

Cumulated moves

Cumulated moves

A decision should be quite straightforward and based on the MCDA results. After all, offering some certainty to bidders on this issue was one of the objectives of the GATT’s rules reform. But at least two factors interfere: the clarity of the results and the risk of recourse. It can be hard to choose among two overall close offers. The problem was not anticipated by anybody, but the Tasmanian authority (Australia) who recognized that a small difference in overall scores opens the possibility to adjudicate to any of the bidders. Later, Swiss tribunals acknowledged the problem and proposed several acceptable ways to award the contract, e.g., chance, choosing the less expensive or the one who beneficiated the least of public contracts so far. In our cases, no legal recourse was ever filed, and all contracts were signed, but it cannot be seen as a general case. In the early years of implementation in Switzerland, recourse became a kind of national sport. To avoid being overflown, tribunals decided at some stage to examine

7 6 5 4 3 2 1 0 -1 -2 -3 -4

1

6

Rounds / Participants 11 16

21

A B C D E F

Fig. 7 Example of a silent negotiation process. The relative position of each item is reported after each move performed by a DM. The final choice about items D and E might need some discussion (Pictet and Bollinger 2005)

208

J. Pictet and D. Bollinger

first whether the author of the recourse had any chance of winning; if not (e.g., a better notation on a minor criterion that would not significantly impact the overall score), independently of the merit of the case, the contract would be awarded as proposed by the adjudicator. Due to the losing tenders’ tendency to file a complaint, mathematical issues were more frequent in tribunals. Many adjudicators’ basic mistakes were corrected, and some principles were adopted. There were different strategies about the adequate level of transparency in the communication of the decision that would minimize the risk of recourse. Some authorities provided no details beyond the ranking of the offers before the recourse deadline, to avoid providing “ammunition” to the bidders. Many others provided some details with the same purpose while seemingly “being cooperative.” Rarely they provided the evaluation matrix, not to mention the whole evaluation report. A lot of effort were put into the decision, but much less into its implementation. Some companies adapted their practice. For instance, they offer a knockout price and, after the contract is signed, manage to add invoices for tasks allegedly not specified in the call for tenders. The tendency of public authorities to be soft on such abuses is rather well known. A strict “no excuse” policy, with punishing penalties, would certainly level the field (we did it in some cases). But it would require a form of aggressiveness better known in the private procurement area. Authorities usually miss the fact that weighting the criteria is a political act. A Swiss city adopted a guidance, defining weights as intervals for most types of contracts, and then leaving to the people in charge the choice of the precise values. This would have come in handy for a public servant in another country who tried— unsuccessfully—to conciliate her boss’ tune “Money, Money, Money” with her boss’ boss one “Quality, Quality, Quality.” A minimal consistency in that regard— that could vary from authority to authority and be revised—would make many people’s life much easier.

4 Discussion These two types of application—Environmental Planning (EP) and Public Procurement (PP)—have little in common beyond the basic MCDA tenets. The main differences are discussed below, followed by a couple of topics of interest.

4.1

Spiral of Decisions

The two types of application are situated at distinct phases of a project history and near the extremes of a spiral of decisions representing a succession of similar steps— applied in specific ways—of decision cycles in a shrinking free space—within which a given application has to be limited—while moving from the ideal to the real

Contrasting Applications in Environmental Planning and Public Procurement

209

(Fig. 8). Real cases usually deal with a section of the spiral, which can lead to sequels in another section that are not necessarily managed by using MCDA. The EP cases were set at the planning phase while the PP ones were set at the interface with implementation. The playing field and the time pressure were quite different. The different worldviews were very visible in EP cases, about objectives, alternatives, criteria, and weights, but much less so in PP ones. There were some disagreements in PP cases, but they were quite easy to overcome, even about the weights.

4.2

Contexts

The contexts are quite different too. In the Environmental Planning cases, the context is at the interface between the political and technical spheres. To ensure the adequacy of the analysis, a good understanding of the politics and policies is required. Having a good grasp of the (recent) history of a topic (who said what, etc.) is very helpful to navigate in an ocean of information. Leaving space for a democratic debate does not mean that one should come unprepared. In a recent case, a county minister decided, against our explicit advice, to call a meeting with the municipalities that could potentially be impacted by a major project. His rationale was that he was “one of them,” having been a mayor in the area. As the project was just starting, his side, including the analysts, was largely unprepared. When precise questions started to be asked, his sidekicks could only tell the truth—that they did not have an answer yet—and the minister had to put a quick end to the meeting. In the Public Procurement cases, the context was mostly legal—applying the regulations and jurisprudence applicable at the time—as the policy and technical decisions were already made. Nowadays, the situation is more stable, but it still requires a capacity to command the “juridical lingo.” Beyond that, the analysts are often collaborating with people with a legal background, i.e., a strong sense of logic, but different from theirs. Free space Alternaves

Decision

Criteria Evaluaons

Aggregaon

Alternaves

Decision

Criteria Evaluaons

Aggregaon Time

Fig. 8 Spiral of decisions

210

4.3

J. Pictet and D. Bollinger

Aggregation Methods

The aggregation methods were different and, in our opinion, adapted to each situation. In EP cases, ELECTRE methods might better cope with the diversity of worldviews, the complexity of the case, past experience, and the level of uncertainty. Moreover, the way they were used certainly contributed to their acceptability. Even if the decision-makers did not understand all their intricacies, they could broadly link their results to their own weights and the evaluation matrix. Avoiding complete aggregation, results allowed a persistence of their differences up to the recommendation that made them more at ease at the beginning of the final discussion. In the PP cases, a clearer result was expected to justify the adjudication. In that sense, the use of the weighted average was perhaps more sensible, not to mention its massive use in the past.

4.4

Overall Consistency

Ensuring overall consistency is a primary task in any case. It stems from constraints moving in opposite directions (Fig. 9). Mathematical constraints are moving “upstream” to ensure the soundness of the results: the aggregation method conditions the information needed (e.g., evaluations and weights) and the tools used to elicit it. At some stage, they meet the “downstream” constraints resulting from the chosen procedure (e.g., group decision), the legal requirements, and the partial choices made in problem, alternatives, and criteria structuring. The potential discrepancies have to be handled, as early as possible in the process, to avoid backpedaling and changes that might destabilize the actors. A common mistake found in practice and in the literature is the use of more than one aggregation method in a specific case, sometimes without even considering the adaptation of the information to each method’s constraints. In practice, we strongly advocate against this and refused to do so when the suggestion occurred. The potential gain of having converging results does not, in our opinion, compensate for the complexity to handle different sets of information that could bring more confusion than clarity. Managing the content and the process often occur in parallel and they influence each other. In the beginning, many actors know little or nothing about MCDA and need to be convinced about its interest. During the process, they need to be reassured about potential misuse. Helping them to see the connection between the matrix (es) and the result(s) plays a central role in that regard. In some cases, this can be done by providing them with the files used—which implies using generic tools like spreadsheets—in others, this can only be done by highlighting how the main features of the input are visible in the output.

Contrasting Applications in Environmental Planning and Public Procurement

211

Process management

Content management Stakeholder(s)

Program

Perceived reality Perceived reality Problem structuring

Decisionmaker(s) Criteria structuring

Criteria weighng

Performances evaluaons

Aggregaon

Results

Aggregaon method

Recommendaon

Alternaves structuring Analyst(s)

Expert(s)

Fig. 9 Downstream and upstream constraints and interfaces with actors

4.5

New Interfaces

While working on the cases, the analysts tried to improve the interfaces they used with the decision-makers to elicit different types of information and to analyze and present results. In the elicitation phase, the modified Simos’ cards method was developed to elicit individual weights in EP cases.12 When shared weights were needed, as in PP cases, it was adapted as the “silent negotiation” to provide support. Some experts easily adopted another adaptation for evaluations and were quite happy to have their “qualitative” evaluations treated on the same footing as the more “quantitative” ones. The visual overlay of results (Fig. 4) was especially helpful for representing multiple ELECTRE results. It helped DMs to grasp at once each alternative position in the partial preorders, its sensitivity to the weights and the uncertainties, and its incomparability with the other ones. We also tried other approaches, like counting the number of results showing the outranking of an alternative by each of the other ones. This is important as we do not choose the people we work with. Some may share our type of education, others may have a different one, or no substantive one. Some 12 Usually, the procedure was presented in meetings and DMs started to work on it immediately. The quickest ones provided their answers and received their proposed weights on the spot, while others finished it as “homework.” In any case, the interaction improved greatly when it became possible to send them a small file via e-mail allowing to “play” with the figures before choosing a final set. This procedure was even used with some Chinese regional officials, with the help of very competent translators, within a large clean energy project [Haldi and Pictet 2003].

212

J. Pictet and D. Bollinger

people are not at ease with symbols, and others with numbers. It is our duty to make sure everyone understands enough to contribute usefully to the project. For instance, presenting the evaluation matrix(es) with a color code that facilitates the rough assessment of the strengths and weaknesses of each alternative contributes to the understanding and acceptance of the results (Fig. 2). Some forms of dominance analysis were, for example, proposed in Bana e Costa et al. (2012). Such tools contribute to obtain the needed commitment toward the implementation. There is little interest in mathematically perfect decisions that remain without consequence. Listening is the first communication skill needed. A central tenet of MCDA is putting our tools at the service of people who need them, not using people as mere sources of information. That was one of the stated goals when MCDA took some distance from Operational research (Roy 1985). Problem structuring (e.g., Eden and Ackermann (1998)) has been proposed as a good complement to MCDA model structuring (Montibeller et al. 2008), but was not used in our cases, at least explicitly with the actors.

4.6

MCDA and Democracy

Historically developed as an analytical approach—usually for a single decisionmaker, real or alleged—MCDA was later extended to (small) groups—as exemplified in this chapter—and a more recent trend tries to extend it further to larger groups and possibly a more deliberative approach (see, e.g., Proctor and Drechsler (2006) and Rauschmayer and Wittmer (2006)). We lack a typology that would clarify how to implement MCDA with larger groups, up to an entire population, depending on the settings. In many countries, some forms of consultation exist besides elections. For instance, in France, a national commission for public debate organize meetings since 1995 for projects of national importance, such as a new railway. It could contribute, on the positive side, to achieve a more direct democracy, but it also shows, on a negative side, a defiance toward representative democracy. Long-time experience in Switzerland (Box 4) shows how they both play a role. Box 4 Initiatives and referendums in Switzerland If Switzerland can serve as an example, direct democracy rarely deals with concrete projects. At the federal level, initiatives propose changes to the constitution and referendums deal with laws proposed by the parliament. At the county level, initiatives can propose, modify or suppress laws, with the extra case of financing laws that could deal with a single project. At the community level, some decisions by the legislative body can be submitted to a vote. In a way, direct democracy completes indirect democracy, and does not (continued)

Contrasting Applications in Environmental Planning and Public Procurement

213

Box 4 (continued) replace it. The rationale for this could be summarized that way: (1) The format of the vote (basically yes or no) does not fit the MCDA one; (2) Necessary resources to allow a full participation seem overwhelming; (3) Swiss citizens already vote four times a year, each time about several federal, county, and/or community texts; most people agree that we are close to the limit; (4) Often, direct democracy is animated by non-political entities, such as labor unions, corporate lobbies, environmental associations, who also provide the—paid or unpaid—workforce to collect the signatures needed; (5) Transparency is far from perfect, as it is the case for the whole political field. More broadly, it calls into question the often unclear role of economy in politics. Several cases of the EP type can be seen as a form of deliberative democracy. Its drawbacks are often cited along with calls for more direct democracy. Switzerland, for instance, shows that public participation can go beyond “mere” elections, with initiatives and referenda. But, in our opinion (Bollinger and Pictet 2003), it is hard to integrate the general public into MCDA applications outside its participation in the initial diagnostic (questions, concerns, proposals), the presentation of the results, and sometimes a final vote. Documenting what happens in between is important to connect these two steps and convince people that their initial input was properly taken into account. A central issue deals with the role of a representative. Public officers usually represent an elected official who represents its electors. Delegates of an organization represent its members. In the process, many things can happen (Eden and Ackermann 1996). A mandate, or a clearer definition of it, could simplify the delegate’s positioning. And all can forget about their duty to report back unless accountability is required. Democracy is perhaps the only regime that allows to discuss the Common Good—or at least the allocation of the common good—at the intersection of the people’s aspirations and the constraints of reality. MCDA can help in that regard, if its proponents participate in the discussion about its place within this framework.

4.7

Future Areas of Research

It might be a reminder of scientism to believe that providing better aggregation methods will be enough. The fact is that, in practice, most people can hardly handle methods more complex than the weighted sum. Improvements are possible, as exemplified by MACBETH. In settings with more expertise, there is certainly room for more sophisticated methods.

214

J. Pictet and D. Bollinger

What is much needed is an engagement to promote MCDA in the various fields it can be applied to, that are legion. It requires to enter into the arena, discuss the proposed methodologies, understand the power struggles, and try to be victorious. In that regard, we consider that Public Procurement is a largely lost opportunity to do so, as it is a mandatory field in which MCDA was in a very favorable position to win the contest. But overcoming the inertia would have required a concerted effort. Adapting to new knowledge is very important to identify new venues of research for our field. An interesting one, in our opinion, deals with dual processes (see, e.g., Kahneman (2011)). It is very much a work in progress, but it brings new light on issues that are at the core of MCDA, such as reasoning, biases, and heuristics, as they are in other fields. Research in that direction already started (see, e.g., Montibeller and von Winterfeldt (2015)). In return, MCDA can propose reasonably simple solutions to ensure a dialogue about the two processes in practice.

5 Conclusion Belton and Pictet (2002) argued that real-world applications of MCDA were relatively rare in the literature. Some of our applications were published (mainly in French) and a full account of one application is not easy, mainly because it would contain elements that belong to different fields of knowledge. Moreover, clients might have been reluctant to have their work put into the spotlight. These features might not be specific to MCDA, but it does not help promoting this field. Instead of detailing one case, as originally requested, the authors thought, perhaps wrongly, that it would be more useful to address issues arising from numerous applications they were involved with in two distinct areas. Practicing MCDA is an art using science, among other things. Its success depends on the “glue” used to bind the patchwork of contributing elements together. As we tried to exemplify in this chapter, this patchwork varies from case to case, even though specific methodologies can be designed to tackle certain types of cases. Consistency and honesty are at the heart of this practice. Consistency deals with the use we make of existing methodologies and tools, and the efforts we make to remain within their respective admissibility. Honesty is our only remaining guide when we have to push beyond that. That way, we might be able to satisfy our clients and the unwritten mandate from the general public. We cannot help noticing the widening gap between what research has to offer and what is needed in practice. If MCDA is just a branch of applied mathematics and neglects the social sciences, it will be harder and harder to claim it is appropriate to handle real-world situations that could benefit from it. It is a challenge for all of us to maintain the connection. Acknowledgments The authors would like to thank the editors for their suggestions to improve the readability of this chapter.

Contrasting Applications in Environmental Planning and Public Procurement

215

References Ballesteros-Pérez P, Skitmore M, Pellicer E, González-Cruz MC (2015) Scoring rules and abnormally low bids criteria in construction tenders: a taxonomic review. Constr Manag Econ 33(4): 259–278. https://doi.org/10.1080/01446193.2015.1059951 Bana e Costa CA, Corrêa EC, De Corte JM, Vansnick JC (2002) Facilitating bid evaluation in public call for tenders. A socio-technical approach Omega 30(3):227–242. https://doi.org/10. 1016/S0305-0483(02)00029-4 Bana e Costa CA, De Corte JM, Vansnick JC (2012) Macbeth. Int J Inf Tech Decis 11(2):359–387. https://doi.org/10.1142/S0219622012400068 Belton V, Pictet J (1997) A framework for group decision using a MCDA model: sharing, aggregating, or comparing? J Decision Systems 6(3):283–303. https://doi.org/10.1080/ 12460125.1997.10511726 Belton V, Pictet J (2002) Talking about the practice of MCDA. In: Bouyssou D, Jacquet-Lagrèze E, Perny P, Slowinski R, Vanderpooten D, Vincke P (eds) Aiding decisions with multiple criteria. Essays in honor of Bernard Roy. Kluwer, Dordrecht, pp 71–88. https://doi.org/10.1007/978-14615-0843-4 Bollinger D, Pictet J (2003) Potential use of e-democracy in MCDA processes; analysis on the basis of a Swiss case. J Multicrit Decis Anal 12:65–76. https://doi.org/10.1002/mcda.348 Bollinger D, Pictet J (2008) Multiple criteria decision analysis of treatment and land-filling technologies for waste incineration residues. Omega 36:418–428. https://doi.org/10.1016/j. omega.2006.07.008 Confédération helvétique (2000) Ordonnance sur l’aménagement du territoire, Berne Conférence romande des marchés publics (2021) Guide romand pour les marchés publics, Lausanne Conseil d’Etat (2012) Département de la Guadeloupe. Décision N°362532. Paris Direction des affaires juridiques (2013) Le prix dans les marchés publics. Ministère de l’économie et des finances, Paris DTAP (2017) La pesée des intérêts en aménagement du territoire. Rapport du groupe de travail, Conférence suisse des directeurs cantonaux des travaux publics, de l’aménagement du territoire et de l’environnement, Berne Eden C, Ackermann F (1996) “Horses for courses”: a stakeholder approach to the evaluation of GDSSs. Group Decis Negot 5:501–519. https://doi.org/10.1007/BF02404647 Eden C, Ackermann F (1998) Making strategy: journey of strategic management. SAGE Publications, London, UK Figueira J, Roy B (2002) Determining the weights of criteria in the ELECTRE type methods with a revised Simos' procedure. Eur J Opl Res 139:317–326. https://doi.org/10.1016/S0377-2217(01) 00370-8 Genard JL, Pirlot M (2002) Multi-criteria decision-aid in a philosophical perspective. In: Bouyssou D, Jacquet-Lagrèze E, Perny P, Slowinski R, Vanderpooten D, Vincke P (eds) Aiding decisions with multiple criteria. Essays in Honor of Bernard Roy, Kluwer, Dordrecht, pp 89–117 Haldi PA, Pictet J (2003) Multicriteria output integration analysis. In: Eliasson B, Lee YY (eds) Integrated assessment of sustainable energy systems in China: a framework for decision support in the electric sector of Shandong province. Kluwer, Dordrecht, pp 661–701. https://doi.org/10. 1007/978-94-010-0153-3 Kahneman D (2011) Thinking, fast and slow. Penguin Books, London Maystre LY (1985) Initiation aux calculs économiques pour l’ingénieur. Illustrée par des exemples du génie de l’environnement. PPUR, Lausanne Maystre LY, Bollinger D (1999) Aide à la négociation multicritère. Pratique et conseils. PPUR, Lausanne Maystre LY, Pictet J, Simos J (1994) Méthodes multicritères ELECTRE. Description, conseils pratiques et cas d'application à la gestion environnementale. PPUR, Lausanne

216

J. Pictet and D. Bollinger

Montibeller G, Belton V, Ackermann F, Ensslin L (2008) Reasoning maps for decision aid. An integrated approach for problem-structuring and multi-criteria evaluation. J Multicrit Decis Anal 59(5):575–589. https://doi.org/10.1057/palgrave.jors.2602347 Montibeller G, von Winterfeldt D (2015) Cognitive and motivational biases in decision and risk analysis. Risk Anal 35(7):1230–1251. https://doi.org/10.1111/risa.12360 Pictet J (1996) Dépasser l’évaluation environnementale. Procédure d’étude et insertion dans la décision globale. PPUR, Lausanne Pictet J, Bollinger D (2000) Aide multicritère à la décision: Aspects mathématiques du droit suisse des marchés publics. Baurecht/Droit de la construction 2:63–65 Pictet J, Bollinger D (2003) Adjuger un marché au mieux-disant. Analyse multicritère, pratique et droit des marchés publics. PPUR, Lausanne Pictet J, Bollinger D (2005) The silent negotiation or how to obtain collective information for group MCDA without excessive discussion. J Multicrit Decis Anal 13(5–6):199–211. https://doi.org/ 10.1002/mcda.392 Pictet J, Bollinger D (2008) Extended use of the cards procedure as a simple elicitation technique for MAVT. Application to public procurement in Switzerland. Eur J Opl Res 185:1300–1307. https://doi.org/10.1016/j.ejor.2006.05.051 Pictet J, Maystre LY, Simos J (1994) Surmesure. An instrument for presentation of results obtained by methods of the Electre and Prométhée families. In: Paruccini M (ed) Applying multiple criteria aid for decision to environmental management. Kluwer, Dordrecht, pp 291–304 Pictet J, Simos J (1996) Pesée globale des intérêts. Apports de l’aide multicritère à la décision. In: Morand CA (ed) La pesée des intérêts. Droit de l'environnement et de l'aménagement du territoire. Helbing & Lichtenhahn, Bâle/Faculté de droit de Genève, pp 173–188 Proctor W, Drechsler M (2006) Deliberative multicriteria evaluation. Environ Plann C: Government and Policy 24:169–190. https://doi.org/10.1068/c22s Rauschmayer F, Wittmer H (2006) Evaluating deliberative and analytical methods for the resolution of environmental conflicts. Land Use Policy 23:108–122. https://doi.org/10.1016/j.landusepol. 2004.08.011 Rogers M, Bruen M, Maystre LY (2000) Electre and decision support. Methods and application in engineering and infrastructure investment. Kluwer, London. https://doi.org/10.1007/978-14757-5057-7 Roy B (1985) Méthodologie multicritère d'aide à la décision. Economica, Paris. English edition. In: Roy B (ed) (1996) Multicriteria methodology for decision aiding. Kluwer, Dordrecht Roy B, Bouyssou D (1993) Aide multicritère d'aide à la décision: Méthodes et cas. Economica, Paris Ruegg J, Mettan N, Vodoz L (eds) (1992) La négociation: son rôle, sa place dans l’aménagement du territoire et la protection de l’environnement. PPUR, Lausanne Simos J (1991) Evaluer l’impact sur l’environnement. Une approche originale par l’analyse multicritère et la négociation, PPUR, Lausanne

Social Multi-Criteria Evaluation of Policy Options Giuseppe Munda

1 Introduction When one wishes to formulate, evaluate and implement public policies, the existence of a plurality of social actors, with interest in the policy being assessed, generates a conflictual situation. Any policy decision affects individuals, regions or groups in different ways; consequently, public support for any policy depends upon the distributional effects it entails. In my view, a proper public policy assessment process should always consider a plurality of social values, perspectives and interests (Munda 2016, 2022). Real-world problems and their complexity change in relation with the nature of the specific policy and the geographical and cultural contexts (see e.g. Etxano and Villalba-Eguiluz 2021, where 42 worldwide case studies are reviewed in fields such as rural and urban planning, water and energy resource management, waste management and so on). The capacity of dealing with real-world complexity is one of the strongest arguments in favour of a multi-dimensional approach such as multi-criteria evaluation; however, in the framework of policy analysis, the traditional evaluation tool is cost-benefit analysis (CBA), which focuses on efficiency criteria only. Here, I will then first introduce CBA, as an economic-oriented approach, and after social multicriteria evaluation as a more comprehensive evaluation framework. Some basic mathematical issues will be also considered and finally some conclusions will be drawn.

G. Munda (✉) European Commission, Joint Research Centre (JRC), Ispra, Italy e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_8

217

218

G. Munda

2 Democracy and Policy Assessment The main methodological assumption behind cost-benefit analysis is that any individual makes rational decisions only if she/he weighs up the advantages and disadvantages of a particular action. The question then is if this individual rationality applies in the social context too; according to CBA, it does apply. If individuals can carry out their own personal cost-benefit analyses in respect to a given policy, then we can aggregate the results to secure a social assessment. The notion of individual preference used in the Kaldor (1939)–Hicks (1939) compensation principle, which is the economic theoretical foundation of cost-benefit analysis, is the preference expressed on the market place by consumers (or which would be expressed if there were a market) (see e.g. Mishan 1971; Pearce and Nash 1989). This kind of “economic democracy” is preferred to classical political voting on the following grounds: 1. The Kaldor–Hicks compensation principle declares a social state S1 “socially preferable” to an existing social state S0 if those who gain from the transition to S1 can compensate those who lose and still have some gains left over. In political democracy, minorities must accept decisions taken by majority, on the contrary, in the framework of the Kaldor–Hicks compensation principle, losers receive compensation; this appears an improvement of the fairness of the policy process. 2. Economic democracy always reflects voters’ preferences. If a voter can be considered as a consumer, then if she/he does not like a good, she/he does not buy it on the market. 3. To observe consumers’ behaviour on the market is much cheaper, quicker and easier than political referenda on any specific policy option. 4. “The use of money values permits some expression of the intensity of preference in the vote: it enables the individual to say how deeply he wants or does not want the project or good in question” Pearce and Nash (1989, p. 7). Although Kaldor and Hicks were interested in implementing objective Pareto efficiency, explicitly not grounded on egalitarian considerations, economic democracy appears to perform better than political democracy. But, is this really true? By only taking preferences expressed on the market into account, one compares individuals according to one objective and one institution only, i.e. economic efficiency and markets. Different objectives and values, e.g. sustainability or fairness are not considered (Munda 2016; Sagoff 1988; Sen 2009). Economic efficiency tries to answer to the following question: does society wish to assign any resource to a given policy objective? It is impossible to avoid the economic problem of “opposition between tastes and obstacles”, as Pareto made clear. Cost-benefit analysis deals with this issue correctly. However, if the losers are poor (or even not yet been born), the compensation is always low. Contemporary welfare economics deals with distributional consequences of policy options by attaching different weights to different income groups (Bojo et al. 1990). However, it is not clear how to derive such weights and who should

Social Multi-Criteria Evaluation of Policy Options

219

attach them.1 On the other hand, not using any weighting system implies making the implicit assumption that the existing distribution of income is ideal. If, and only if, one is happy with such a value judgement, it is reasonable to use un-weighted market valuations to measure costs and benefits. Therefore, there is no escape from value judgements. The compensation principle is not the positivistic objective evaluation criterion Hicks hoped to be. On the other side, it does not consider individuals as equal, thus the compensation principle can be considered a direct application of the ancient principle that property owners should count more (Munda 2019). The most widespread non-monetary approach to public policy assessment is multi-criteria evaluation (MCE). The basic methodological foundation of MCE is incommensurability, i.e. the notion that in comparing options, a plurality of dimensions and perspectives is needed (Chang 1997; Frame and O’Connor 2011; Lo and Spash 2013; Martinez-Alier et al. 1998; Munda 2016; O’Neill 1993; Spash 2008). The fact that “one’s welfare economics will inevitably be different according as one is a liberal or a socialist, a nationalist or an internationalist, a Christian or a pagan” (Hicks 1939, p. 696) is the normal state of affairs in public policy. There is no obvious reason why this issue of existence of a plurality of values should be considered a problem that can be solved by considering consumers’ preferences as the only relevant social values. A question arises here: is it more scientific (and fair) an approach dealing with such a plurality of values explicitly or one which solve all conflicts by imposing a perspective considered superior on some ethical or technical grounds? The point is that different metrics are also linked to different social objectives and values; in this context, the statement “x is better than y” implies an answer to two questions: (1) according to what? (2) According to whom? To use only one measurement unit for incorporating a plurality of dimensions, objectives and values, implies reductionism necessarily. Incommensurability is the only rational way to compare various objects under different methodological assumptions than traditional optimisation grounded on the use of one criterion and one measurement unit only. The basic idea of multi-criteria evaluation (MCE) is to achieve the comparability of incommensurable metrics. From an operational point of view, the major strength of MCE is its ability to deal with policy issues characterised by various contradictory evaluations, thus allowing for an integrated assessment of the problem at hand. Being a decision tool, MCE focuses on the issue of the opportunity cost connected to

1

From the technical point of view, one should note that the fact that intensity of preference is taken into account inside a linear aggregation rule has the consequence that weights must be considered as trade-offs. A question then arises: in their standard use, are distributional weights used as importance coefficients or as trade-offs? The basic idea underlying all the different weighting methods can be summarized by quoting the following sentence: “if the decision-maker considers individual 2 more ‘deserving’ than individual 1 he will weight 2’s losses more heavily than 1’s gains i. e. 2 > 1” (Dasgupta and Pearce 1972, p. 65), thus weights should be considered as importance coefficients. Unfortunately, since CBA is based on a completely compensatory mathematical model, weights can only have the meaning of a trade-off ratio, as a consequence a theoretical inconsistency exists (see Munda 1996 for more details on this issue).

220

G. Munda

the choice of any policy option, thus efficiency is surely an important objective. Differently from economic efficiency assessment tools such as CBA or frontier methods such as Data Envelopment Analysis (DEA), often used by operational research analysts (see Emrouznejad and Guo-liang 2018 for a recent overview), MCE is based on a multi-dimensional framework. CBA or DEA can also be one of the criteria used in a MCE exercise, but never the only ones (Agasisti et al. 2019). A clear advantage of MCE is that it can tackle different objectives, such as efficiency, equity, or sustainability separately in a transparent way. Historically, the first stage of the development of MCE is characterised by the so-called methodological principle of multi-criteria decision-making (MCDM). The main objective of this approach is first to elicit preferences from a decision-maker and then solve a well-structured mathematical decision problem (see e.g. Keeney and Raiffa 1976). The limitations of the classical concept of an optimum solution and the consequential importance of the decision process were emphasised by authors such as Herbert Simon (1976) and Bernard Roy. According to Roy (1996) saying that a decision is a good or bad one is in general impossible referring only to a mathematical model. All aspects of a decision process, which leads to a given decision, also contribute to its quality and success. The final solution is more like a “creation” than a discovery. Under the concept of a Multiple-Criteria Decision Aid (MCDA) the principal aim is not to discover a solution, but to construct or create something which is viewed as liable to help an actor taking part in a decision process either to shape, argue, and/or transform her/his preferences, or to make a decision in conformity with his/her goals (Roy 1996). A positive aspect of CBA is that in theory economic votes of all individuals might be used (of course under the assumption that the current distribution of income is considered acceptable). On the other hand, MCE may be based on the preferences, perspectives and interests of a restricted number of policy-makers only. Public policy analysis has increasingly recognised the need for public participation (Guimarães-Pereira et al. 2006; Funtowicz and Ravetz 1991; O'Neill 2001). Social Multi-Criteria Evaluation (SMCE) tries to extend MCDA by incorporating the notion of social actor. Thus, an SMCE process must be as participative and as transparent as possible; although, participation is a necessary but not a sufficient condition for successful evaluation (Munda 2004, 2008). This is the main reason why the concept of SMCE is proposed. In an SMCE framework, fairness is an ethical obligation to take a plurality of social values, perspectives and interests into account, in a coherent and transparent manner.2 The main accomplishment of SMCE is that a wide range of evaluation criteria (incommensurable from a technical point of view) has a direct translation in 2

It has to be clarified that the concept of fairness is different from the one of an equal distribution of income. A society could have a fair inequality if the economic system promotes and rewards individual efforts. Clearly ethical connotations are there; this implies that people, social scientists and governments differ significantly on what they consider to be fair. Overall there is agreement on the fact that evaluation of fairness should be linked to the social process leading to a certain

Social Multi-Criteria Evaluation of Policy Options

221

terms of the plurality of values and perspectives (incommensurable from a social point of view) used in the evaluation exercise.

3 Down to Earth Social Multi-Criteria Evaluation Processes In the context of public policy, the issue of weights is particularly relevant; consequently, SMCE tries to put transparency on this issue, starting with the relation dimensions/criteria. A dimension is the highest hierarchical level of analysis and indicates the scope of criteria and criterion scores. The general categories of economic, social and environmental impacts are dimensions. Weights are often used to represent the relative importance attached to dimensions and criteria. The idea behind this practice is very intuitive and easy, that is, to place the greatest number in the position corresponding to the most important factor. A common practice is the pragmatic solution of no criterion weighting. However, the fact that all criteria have the same weight does not guarantee at all that dimensions have the same weight. This would be guaranteed only under the condition that all the dimensions have the same number of criteria; this of course is quite unnatural and artificial, and even dangerous. On the contrary, different criterion weights can guarantee that all the dimensions are considered equal. A reasonable practice can be to start by giving the same weight to each dimension and then splitting each weight among the criteria of any dimension proportionally. Figures 1 and 2 (produced by means of the SOCRATES software described in the next section) represent these situations in a graphical way. As one can see in this case the relation dimensions/ criteria is a very peculiar one. In fact, most of criteria belong to the economic dimension, while other dimensions are much less populated. This implies that the starting weighting assumption can be only equal dimension weights because otherwise (under the equal criterion weighting assumption) the economic dimension would dominate since its weights would be higher than 50% of all dimensions considered (in technical terms it would become a dictator). Of course, one could assume that some dimensions are more important than other ones, and thus their weight should be higher, but this should be justified. Finally, one should note that weights can be used in the way described here, only if they have the meaning of importance, which depends on the fact that they are combined with noncompensatory aggregation mathematical rules. One should note that policy evaluation is not a one-shot activity; on the contrary, it evolves as a learning process. It has to be realised that the evaluation process is usually highly dynamic, so that judgements regarding the political relevance of items, alternatives or impacts may display sudden changes, hence requiring a policy

outcome and not to the outcome itself (i.e. when differences in the final income distribution of a society exist, this does not mean that the society has unfair rules).

222

G. Munda

Fig. 1 Equal criterion weighting—economic dimension receives 61.54%

analysis to be flexible and adaptive in nature. This is the reason why evaluation processes have a cyclical nature; that is the possible adaptation of elements of the evaluation process due to continuous feedback loops between the various steps and consultations among the actors involved (Nijkamp et al. 1990). In operational terms, the application of an SMCE process involves the following seven main steps (Munda 2008): (i) Description of the relevant social actors.3 For example, institutional analysis4 may be performed on historical, legislative and administrative documents to provide a map of the relevant social actors.

3

The social actors are those who can influence or whose interests are affected by the policy options. The social dimension of a problem can be explored by using institutional analysis, an approach that can illuminate values, interests, roles, possible alliances and available resources of the social actors involved in a policy problem. Different written and oral sources are used for carrying out an institutional analysis. In the first category there are local and national press, specialized magazines, official and informal documents produced by the social actors in order to explain their position, 4

Social Multi-Criteria Evaluation of Policy Options

223

Fig. 2 Equal dimension weighting—economic dimension receives 33.33%

(ii) Definition of social actors’ values, desires and preferences can be achieved using focus groups and develop a set of policy options. The main limitations of focus groups are lack of statistical representation of the population and occasional reluctance of people to participate or state publicly what they really think (e.g., in small towns and villages). For this reason, anonymous questionnaires and personal interviews are an essential part of the participatory process. (iii) Generation of policy options and selection of evaluation criteria is a process of co-creation resulting from a dialogue between analysts and social actors. For example, potential sites for the location of wind parks generating renewable energy could be found by considering factors relevant for investors only, such as technical and economic feasibility depending on wind availability. However, local people may raise other concerns such as visual impact or closeness

books, articles, and so on. In the second group, there are individual interviews with key agents or with a random sample or focus groups. Normally, many complementary sources are used simultaneously.

224

G. Munda

to places of a high symbolic value for the community (e.g. an ancient monument or a peculiar landscape). Should the evaluation criteria then come directly from the public participation process or be mediated by the research team? I think that the rough material collected during interviews and focus groups could be used as a source of inspiration but the technical formulation of criteria is a task for expert researchers. In this way, evaluation criteria become a technical translation of social actors’ needs, preferences and desires. For example, if a local community has worries about the possible noise produced by windmills, a possible evaluation criterion is sound pressure computed in decibels; if a desire is to keep younger generations in a rural area, a clear relevant criterion is people employed by the wind park, and so on. Of course, in this step, judgement is unavoidable, e.g. which is the best way of measuring visual impact? For this reason, a widespread information campaign— including local people, regional and national authorities, international scientists and even children at school—covering the assumptions and conclusions of the study is, in my opinion, always highly recommended. (iv) Construction of the multi-criteria evaluation matrix synthesising the scores of all criteria for all alternatives, i.e. the performance of each alternative according to each criterion. (v) Construction of an equity matrix. Criteria and criterion scores are not determined directly by social actors. The multi-criterion impact matrix is a result of a technical translation operationalised by expert analysts. Even if the criteria are exactly the ones agreed with the social actors, the determination of the criterion scores is independent of their preferences. This is the main reason why I recommend combining a “social impact matrix”, where each option is assessed according to social actors` preferences or foreseen direct impacts on them, with the multi-criteria evaluation matrix; this approach is a peculiarity of SMCE. Of course, the criticism applies here that a subjective component of the research team is always present when summarising the impacts of the various options on the different social actors. This is obviously true, although in my experience, social scientists greatly appreciate the possibility of working with an operational framework which allows them to synthesise the large amount of non-formalised information collected during their field investigations. Another possibility for deriving the social impact matrix is to ask directly to social actors their preferences towards the policy options considered; in this way a kind of voting matrix is obtained. This second approach eliminates the subjective component due to researchers and relates all subjectivity to social actors directly. Some techniques, such as “opinion mining”, allowing for statistical analysis of opinions can be useful here. This makes a clear distinction between opinions contained in the social impact matrix and evidence contained in the multi-criteria evaluation matrix. Obviously, social actors might have a strategic and lobbying behaviour in evaluating policy options. (vi) A mathematical procedure (or algorithm) is applied in order to aggregate criterion scores and obtain a final ranking of the available alternatives. The importance of mathematical approaches is their ability to allow a consistent

Social Multi-Criteria Evaluation of Policy Options

225

aggregation of the diverse information contained in the multi-criterion impact matrix and in the equity matrix. (vii) Sensitivity and robustness analysis aims addressing aspects of abstraction from reality required by any modelling exercise, i.e., checking the relevance and the explicative capacity of the theoretical framework used to structure and understand a policy problem. One approach is to look at the sensitivity of results to the exclusion/inclusion of different criteria, criterion weights and dimensions (see Saltelli et al. 2008, 2010). While such analysis may look very technical, in reality a social component is also present. That is, inclusion/exclusion of a given dimension, or set of criteria, normally involves a long story of social, political and scientific controversy, and involves social values and social actors. These seven steps are not rigid. On the contrary, flexibility in real-world situations is one of the main advantages of social multi-criteria evaluation (see e.g. Vargas Isaza 2004 for an application of SMCE in Colombia, where there was an extreme situation involving social actors belonging to various informal armies (the so-called “actor armado”); Martí 2005, who conducted a study with indigenous communities in Peru; Sittaro 2006 who applies SMCE in the context of indigenous communities in the Amazonian region of Ecuador). As a tool for policy evaluation and conflict management, SMCE has demonstrated its applicability to real-world problems in various geographical and cultural contexts (e.g., Cerreta and De Toro 2010; Corral and Hernandez 2017; Gamboa 2006; Gamboa and Munda 2007; Garmendia and Stagl 2010; Lerche et al. 2017; Monterroso et al. 2011; Özkaynak 2008; Scolobig et al. 2008; Soma and Vatn 2009; Straton et al. 2010; Zendehdel et al. 2010). A recent and exhaustive overview of worldwide SMCE applications can be found in Etxano and Villalba-Eguiluz (2021). In my opinion, some interesting lessons emerge from these case studies: • SMCE is a powerful framework for policy analysis since it accomplishes the goals of being inter-multi-disciplinary (with respect to the research team), participatory (with respect to the local community) and transparent (since all criteria are presented in their original form without any transformations into money, energy or any other common measurement rod). • Transparency is an essential feature to guarantee the quality of any study based on science for policy. In fact, all such studies should be accountable to the public at large for peer review. • One should not forget that the classical schematized relationship between decision-maker and analyst is embedded in a social framework, which is of a key importance in the case of public policy. • The use of a cyclical evaluation process allows for the incorporation of everything learnt during the study. It is extraordinarily important that one uses different participatory and interaction tools at different points in time. This allows for continuous testing of assumptions and unavoidable biases of the research team.

226

G. Munda

• The combination of various participatory methods, which has proved powerful in sociological research, becomes even more so when integrated within a multicriterion framework. • According to the geographical scale chosen, institutional analysis helps in finding the relevant social actors with interests at stake. However, besides the unavoidable mistakes that may occur when carrying out an appropriate institutional analysis, there are even stronger reasons why a purely participatory study is undesirable. • The scientific team cannot accept uncritically all input of a participatory process, since: a. In a focus group, powerful social actors may influence all the others quite strongly. b. Some social actors might not desire or be able to participate, but for ethical reasons the scientific team should not ignore them. c. Focus groups are not a representative sample of the population. Consequently, they can be a useful means of improving the researchers’ knowledge of the institutional and social dimensions of the problem at hand, but they are never a way to derive consistent conclusions on social preferences. • Since decision-makers require legitimacy for the decisions taken, it is extremely important that public participation and scientific studies do not become the justification for a lack of political responsibility. I strongly believe that the deontological principles of the scientific team and policy-makers are essential for assuring the quality of the evaluation process. Social participation does not imply that scientists and decision-makers have no responsibility for policy actions defended and eventually taken. • Ethics matter. Let us imagine an extreme case in which a development project in the Amazon forest could affect an indigenous community with little contact with other civilisations. Would it be ethically more correct to invite them to a focus group. . . or to take into account the consequences of the project for their survival? And what about future generations and non-humans? • A positive externality of participatory approaches is that the results obtained by the research team, i.e. data, findings, interpretations and insights, can sometimes also be returned to the community, which may then use them not just as a given, but rather as an input for deliberative democracy. According to Etxano and Villalba-Eguiluz (2021), SMCE is useful in public policy for three main reasons: (1) operationalisation of sustainability principles, (2) incorporation of social actors’ views through participative processes, and (3) search for compromise solutions (in the social and political meaning). However, these authors also point out some difficulties in applying SMCE in real-world problems:

Social Multi-Criteria Evaluation of Policy Options

227

1. Participatory processes are often linked to the local scale, but implications at a larger scale of the policy options should also be taken into account, which is a challenge from the operational point of view. 2. The influence of the social actors in selecting the evaluation criteria is decisive as it reflects their vision of the policy problem, but how to deal with the issue of criterion weights is not entirely clear. 3. Complete transparency of the policy process allowing full understanding for all the social actors involved of the technical parameters used in a multi-criterion algorithm is very difficult to achieve. 4. Reaching compromise solutions in practice is not easy. Social actors may have power and lobbying-strategic behaviour for vetoing some policy options or influencing their selection and implementation. 5. An innovative policy-making practice incorporating elements such as complexity, incommensurability and uncertainty is desirable. The probability of creating such a practice is not very high without making SMCE and other MCDA participatory approaches a legal requirement when evaluating public projects.

4 Mathematical Approaches in the Social Multi-Criteria Framework In the SMCE framework, mathematical models still play a very important role, i.e. the one of guaranteeing consistency between assumptions used and results obtained. This is a key success factor since multi-criteria mathematics does answer to the standard objection that the aggregation of apples and oranges is impossible in a definitive way. As the reader knows, a “discrete multi-criterion problem” can be described as follows (see e.g. Figueira et al. 2016; Ishizaka and Nemery 2013; Roy 1996; Vincke 1992): A is a finite set of N feasible actions. M is the number of different points of view, or evaluation criteria, gm, that are considered relevant to a specific policy problem. In this way a decision problem may be represented in an N by M matrix P called an evaluation matrix. In general in a multi-criterion problem, there is no solution optimising all the criteria at the same time (ideal or utopia solution), and therefore “compromise solutions” have to be found. Arrow and Raynaud (1986) analysed the formal analogies between the discrete multi-criterion problem and the social choice one. They concluded that the main results of social choice also apply to MCDA; in particular Arrow’s impossibility theorem stating that there is no perfect mathematical aggregation rule (Arrow 1963). Thus, unlike other mathematical fields, neither approximation (i.e., discovering pre-existing truths) nor convergence (i.e., does the decision automatically lead, in a finite number of steps, to the optimum a*?) criteria can be used. Only “reasonable” mathematical procedures make sense in this field. Reasonable here means that algorithms must be evaluated not only according to the formal properties they

228

G. Munda

present, but overall according to the empirical consequences implied by their use too. In synthesis, the information contained in the multi-criteria evaluation matrix is: • • • •

Intensity of preference (when quantitative criterion scores are present) Number of criteria in favour of a given alternative Weight attached to each single criterion Relationship of each single alternative with all the other alternatives

Combinations of this information generate different aggregation conventions, i.e. manipulation rules of the available information to arrive at a preference structure. The aggregation of several criteria implies taking a position on the fundamental issue of compensability (see Podinovskii 1994; Vansnick 1990; Vincke 1992). Compensability is a very important concept when SMCE is applied to integrate various policy dimensions. For example, in evaluating a policy option that presents a very bad environmental impact and a very good economic impact, it is clear that allowing or not for compensability and to which degree is the key assumption (complete compensability implies that a good performance on the economic side would offset a very bad one on the environment or vice versa). An aggregation rule that is relatively simple and non-compensatory (thus also allowing the use of weights as importance coefficients) is the Kemeny rule (Kemeny 1959). Its basic idea is that the maximum likelihood ranking of policy options is the ranking supported by the maximum number of criteria (or criterion weights) for each pair-wise comparison, summed over all pairs of options considered (Munda and Nardo 2009). For example, let us assume that in comparing three options according to 60 criteria, the following pair-wise comparisons are obtained:

a

a 0

b 33

c 25

b c

27 35

0 18

42 0

Then, the corresponding Kemeny scores of all possible rankings are the following ones, where the ranking b → c → a is the best result. a b c b c a

b c a a b c

c a b c a b

100 104 86 94 80 76

Moulin (1988, p. 312) clearly states that the Kemeny method is “the correct method” for ranking options, and that the “only drawback of this aggregation method is the difficulty in computing it when the number of candidates grows” (for

Social Multi-Criteria Evaluation of Policy Options

229

example with ten alternatives, the number of permutations is 10! = 3,628,800).5 A numerical algorithm solving this computational drawback in an efficient way has been developed recently (Azzini and Munda 2020) and it has been implemented in a software tool called SOCRATES (SOcial multi-CRiteria AssessmenT of European policieS) (all methodological and mathematical details behind the SOCRATES software can be found in Azzini and Munda 2020; Munda 2004, 2009, 2012, 2022).6 Overall, the objective of SOCRATES is NOT substitution of policy-makers through a mathematical model, on the contrary, the objective is to improve their understanding of the main features of the problem at hand, such as key assumptions, degree of uncertainty, robustness of results and overall technical and social defensibility of options chosen. The philosopher Socrates said “I cannot teach anybody anything. I can only make them think”. This is the main inspiring principle of the SOCRATES software too. Three main components constitute the core of SOCRATES: multi-criteria, equity and sensitivity analyses. Multi-criteria analysis requires the definition of relevant dimensions, objectives and criteria. It uses weights as importance coefficients and clarifies their role in the hierarchical structure. The multi-criteria evaluation matrix may include either quantitative (including also stochastic and/or fuzzy uncertainty (see Munda 1995)) and qualitative (ordinal and/or linguistic) measurements of the performance of an alternative with respect to an evaluation criterion. It supplies a complete ranking of the alternatives according to the set of evaluation criteria (i.e. the technical compromise solution/s), computed by using the Kemeny non-compensatory aggregation rule. Equity analysis requires as input a set of social actors and their evaluation of the alternatives considered in the multi-criteria analysis. Weights to social actors can be attached, if needed; the starting point is the equal weighting assumption. The equity analysis produces the following information: • Indications of the distance of the positions of the various social groups (i.e. possibilities of convergence of interests or coalition formations) • Ranking of the alternatives according to actors’ impacts or preferences (social compromise solution) In real-world applications in average three situations emerge: 1. Both multi-criteria and equity analyses indicate the same preferred option. This is a win-win situation since policy-makers can implement an option, which is the best-ranked option from the technical point of view and a very defendable option from the social point of view. 2. Equity analysis indicates as a possible social compromise solution an option that is not the first one in the multi-criteria ranking but still is a top-ranked alternative 5

An exhaustive discussion on pros and cons of the Kemeny approach can be found in Chapter 6 of Munda (2008). 6 See also https://knowledge4policy.ec.europa.eu/modelling/topic/social-multi-criteria-evaluationpolicy-options_en/socrates_en

230

G. Munda

(the typical second best). In this case, policy-makers can evaluate the trade-off between the social and technical rankings and consequently make an informed and transparent decision. In my experience, if an option is not the best in the multi-criteria ranking but still is a technically defendable one and is the best candidate for a social compromise, this will be the one chosen. 3. The situation is a very polarised one, that is the most defendable option from a technical point of view is also the most conflictual one. If this applies to secondbest options too, this situation is the worst of the possible worlds for policymakers, a decision should be grounded on clear assumptions and weights used for technical and social concerns. Sometimes, these situations are useful to put light on lobbying behaviour, that is powerful social actors that try to keep the status quo just because any change will make them worse-off. The objective of sensitivity analysis is to check if the rankings provided are stable and to determine which of the input parameters influence more the model output. Local sensitivity analysis looks at the sensitivity of results to (a) the exclusion/ inclusion of different criteria and dimensions and (b) dimensions, criteria or social actors weight changes. All parameters are changed one per time. Global sensitivity analysis focuses on all the possible combinations of criterion weights; all parameters are changed simultaneously. The whole information produced by local and global sensitivity analyses is synthesised into simple graphics. Recently the whole SMCE framework and the SOCRATES software tools have been applied in a set of official impact assessments of the European Commission. A first public study is the “IMPACT ASSESSMENT REPORT Accompanying the document Proposal for a Regulation of the European Parliament and of the Council on standards of quality and safety for substances of human origin intended for human application and repealing Directives 2002/98/EC and 2004/23/EC EUR-Lex - SWD:2022:190:FIN - EN - EUR-Lex (europa.eu)”.

5 Conclusions Real-world problems and their complexity change in relation with the nature of the specific policy and the geographical and cultural contexts. The capacity of dealing with real-world complexity is one of the strongest arguments in favour of a multidimensional approach such as multi-criteria evaluation. Social multi-criteria evaluation (SMCE) has been purposefully designed to deal with public policy problems. In an SMCE framework, fairness is an ethical obligation to take a plurality of social values, perspectives and interests into account, in a coherent and transparent manner. The main accomplishment of SMCE is that a wide range of evaluation criteria (incommensurable from a technical point of view) has a direct translation in terms of plurality of values and perspectives (incommensurable from a social point of view) used in the evaluation exercise. We may also conclude that the SMCE approach shows some commonalities with recent research lines in welfare

Social Multi-Criteria Evaluation of Policy Options

231

economics and public policy, trying to tackle political constraints, interest groups and collusion effects explicitly (see e.g. Laffont 2000; Stiglitz 2002). Acknowledgements This article has been developed in the context of the activities of the Competence Centre on Modelling. The views expressed are purely those of the writer and may not in any circumstances be regarded as stating an official position of the European Commission.

References Agasisti T, Hippe R, Munda G (2019) Measuring the efficiency of European education systems by combining data envelopment analysis and multiple-criteria evaluation. J Product Anal 51:105– 124. https://doi.org/10.1007/s11123-019-00549-6 Arrow KJ (1963) Social choice and individual values, 2d edn. Wiley, New York Arrow KJ, Raynaud H (1986) Social choice and multicriterion decision making. M.I.T. Press, Cambridge Azzini I, Munda G (2020) A new approach for identifying the Kemeny median ranking. Eur J Oper Res 281:388–401. https://doi.org/10.1016/j.ejor.2019.08.033 Bojo J, Maler KG, Unemo L (1990) Environment and development: an economic approach. Kluwer, Dordrecht Cerreta M, De Toro P (2010) Integrated spatial assessment for a creative decision-making process: a combined methodological approach to strategic environmental assessment. Int J Sustain Dev 13: 17–30. https://doi.org/10.1504/IJSD.2010.035096 Chang R (ed) (1997) Incommensurability, incomparability, and practical reason. Harvard University Press, Cambridge Corral S, Hernandez Y (2017) Social Sensitivity analyses applied to environmental assessment processes. Ecol Econ 141:1–10. https://doi.org/10.1016/j.ecolecon.2017.05.022 Dasgupta AK, Pearce DW (1972) Cost-benefit analysis. Macmillan, London Emrouznejad A, Guo-liang Y (2018) A survey and analysis of the first 40 years of scholarly literature in DEA: 1978–2016. Socio-Econ Plan Sci 61:4–8. https://doi.org/10.1016/j.seps.2017. 01.008 Etxano I, Villalba-Eguiluz U (2021) Twenty-five years of social multi-criteria evaluation (SMCE) in the search for sustainability: analysis of case studies. Ecol Econ 188:107–131. https://doi.org/ 10.1016/j.ecolecon.2021.107131 Figueira J, Greco S, Ehrgott M (eds) (2016) Multiple-criteria decision analysis. State of the art surveys. Springer International Series in Operations Research and Management Science, New York Frame B, O’Connor M (2011) Integrating valuation and deliberation: the purpose of sustainability assessment. Environ Sci Pol 14:1–10. https://doi.org/10.1016/j.envsci.2010.10.009 Funtowicz SO, Ravetz JR (1991) A new scientific methodology for global environmental issues. In: Costanza R (ed) Ecological Economics. Columbia, New York, pp 137–152 Gamboa G (2006) Social multi-criteria evaluation of different development scenarios of the Aysén region, Chile. Ecol Econ 59:157–170. https://doi.org/10.1016/j.ecolecon.2005.10.014 Gamboa G, Munda G (2007) The problem of wind-park location: a social multi-criteria evaluation framework. Energy Policy 35:1564–1583. https://doi.org/10.1016/j.enpol.2006.04.021 Garmendia E, Stagl S (2010) Public participation for sustainability and social learning: concepts and lessons from three case studies in Europe. Ecol Econ 69:1712–1722. https://doi.org/10. 1016/J.ECOLECON.2010.03.027 Guimarães-Pereira A, Guedes S, Tognetti S (eds) (2006) Interfaces between science and society. Greenleaf Publishing, Sheffield

232

G. Munda

Hicks JR (1939) The foundations of welfare economics. Econ J 49:696–712. https://doi.org/10. 2307/2225023 Ishizaka A, Nemery P (2013) Multi-criteria decision analysis: methods and software. Wiley, New York Kaldor N (1939) Welfare comparison of economics and interpersonal comparisons of utility. Econ J 49:549–552. https://doi.org/10.2307/2224835 Keeney R, Raiffa H (1976) Decision with multiple objectives: preferences and value trade-offs. Wiley, New York Kemeny J (1959) Mathematics without numbers. Daedalus 88:571–591 Laffont JJ (2000) Incentives and political economy. Oxford University Press, Oxford Lerche N, Wilkens I, Schmehl M, Eigner-Thiel S, Geldermann J (2017) Using methods of multicriteria decision making to provide decision support concerning local bioenergy projects. SocioEcon Plan Sci 68:1–15. https://doi.org/10.1016/j.seps.2017.08.002 Lo AY, Spash CL (2013) Deliberative monetary valuation: in search of a democratic and value plural approach to environmental policy. J Econ Surv 27:768–789. https://doi.org/10.1111/j. 1467-6419.2011.00718.x Martí N (2005) La multidimensionalidad de los sistemas locales de alimentación en los Andes peruanos: los chalayplasa del Valle de Lares (Cusco). Ph.D. Dissertation, Doctoral Programme in Environmental Science. Univ. Autónoma de Barcelona, Barcelona Martinez-Alier J, Munda G, O’Neill J (1998) Weak comparability of values as a foundation for ecological economics. Ecol Econ 26:277–286. https://doi.org/10.1016/S0921-8009(97)00120-1 Mishan EJ (1971) Cost-benefit analysis. Allen and Unwin, London Monterroso I, Binimelis R, Rodríguez-Labajos B (2011) New methods for the analysis of invasion processes: multi-criteria evaluation of the invasion of Hydrilla verticillata in Guatemala. J Environ Manag 92:494–507. https://doi.org/10.1016/j.jenvman.2010.09.017 Moulin H (1988) Axioms of co-operative decision-making. Econometric Society Monographs, Cambridge University Press, Cambridge Munda G (1995) Multicriteria evaluation in a fuzzy environment. Physica-Verlag, Contributions to Economics Series, Heidelberg Munda G (1996) Cost-benefit analysis in integrated environmental assessment: some methodological issues. Ecol Econ 19:157–168. https://doi.org/10.1016/0921-8009(96)00048-1 Munda G (2004) Social multi-criteria evaluation (SMCE): methodological foundations and operational consequences. Eur J Oper Res 158:662–677. https://doi.org/10.1016/S0377-2217(03) 00369-2 Munda G (2008) Social multi-criteria evaluation for a sustainable economy. Springer, Heidelberg, New York Munda G (2009) A conflict analysis approach for illuminating sustainability distributional issues. Eur J Oper Res 194:307–322. https://doi.org/10.1016/j.ejor.2007.11.061 Munda G, Nardo M (2009) Non-compensatory/non-linear composite indicators for ranking countries: a defensible setting. Appl Econ 41:1513–1523. https://doi.org/10.1080/ 00036840601019364 Munda G (2012) Intensity of preference and related uncertainty in non-compensatory aggregation rules. Theory Decis 73:649–669. https://doi.org/10.1007/s11238-012-9317-4 Munda G (2016) Beyond welfare economics: some methodological issues. J Econ Methodol 23: 185–202. https://doi.org/10.1080/1350178X.2016.1157199 Munda G (2019) Multiple-criteria evaluation in public economics and policy. In: Doumpos M, Figueira JR, Greco S, Zopounidis C (eds) New perspectives in multiple criteria decision making. Springer International Series in Operations Research and Decision Theory, New York, pp 297–313 Munda G (2022) Qualitative reasoning or quantitative aggregation rules for impact assessment of policy options? A multiple criteria framework. Quality & Quantity Int J Methodol 56:3259– 3277. https://doi.org/10.1007/s11135-021-01267-8

Social Multi-Criteria Evaluation of Policy Options

233

Nijkamp P, Rietveld P, Voogd H (1990) Multicriteria evaluation in physical planning. NorthHolland, Amsterdam O’Neill J (1993) Ecology, policy and politics. Routledge, London O'Neill J (2001) Representing people, representing nature, representing the world. Environ Plan C: Government and Policy 19:483–500. https://doi.org/10.1068/c12s Özkaynak B (2008) Globalisation and local resistance: alternative city developmental scenarios on capital's global frontier-the case of Yalova, Turkey. Program Plann 70:45–97. https://doi.org/10. 1016/2Fj.progress.2008.04.002 Pearce DW, Nash CA (1989) The social appraisal of projects. Macmillan, London Podinovskii VV (1994) Criteria importance theory. Math Soc Sci 27:237–252. https://doi.org/10. 1016/0165-4896(93)00737-F Roy B (1996) Multicriteria methodology for decision analysis. Kluwer, Dordrecht Sagoff M (1988) Some problems with environmental economics. Environ Ethics 10:55–74. https:// doi.org/10.5840/enviroethics198810128 Saltelli A, Ratto M, Andres T, Campolongo F, Cariboni J, Gatelli D, Saisana M, Tarantola S (2008) Global sensitivity analysis. The primer. Wiley, New York Saltelli A, Annoni P, Azzini I, Campolongo F, Ratto M, Tarantola S (2010) Variance based sensitivity analysis of model output. Design and estimator for the total sensitivity index. Comput Phys Commun 181:259–270. https://doi.org/10.1016/j.cpc.2009.09.018 Scolobig A, Brot VC, Zabala A (2008) Integrating multiple perspectives in social multicriteria evaluation of flood-mitigation alternatives: the case of Malborghetto-Valbruna. Environ Plan C: Government and Policy 26:1143–1161. https://doi.org/10.1068/c0765s Sen A (2009) The idea of justice. The Belknap Press of Harvard University Press, Cambridge Simon HA (1976) From substantive to procedural rationality. In: Latsis JS (ed) Methods and appraisal in economics. Cambridge University Press, Cambridge Sittaro F (2006) Participative multicriteria planning in community development. A case study: the Cuyabeno wildlife reserve, Ecuador. Ph.D. dissertation, Venice International University, International Ph.D. programme in “Analysis and Governance for Sustainable Development”, Venice Soma K, Vatn A (2009) Local democracy implications for coastal zone management-a case study in southern Norway. Land Use Policy 26:755–762. https://doi.org/10.1016/j.landusepol.2008. 10.002 Spash CL (2008) Deliberative monetary valuation and the evidence for a new value theory. Land Econ 84:469–488. https://doi.org/10.3368/le.84.3.469 Stiglitz JE (2002) New perspectives on public finance: recent achievements and future challenges. J Public Econ 86:341–360. https://doi.org/10.1016/S0047-2727(01)00193-1 Straton AT, Jackson S, Marinoni O, Proctor W, Woodward E (2010) Exploring and evaluating scenarios for a river catchment in northern Australia using scenario development, multi-criteria analysis and a deliberative process as a tool for water planning. Water Resour Manag 25:141– 164. https://doi.org/10.1007/s11269-010-9691-z Vansnick JC (1990) Measurement theory and decision aid. In: Bana e Costa CA (ed) Readings in multiple criteria decision aid. Springer-Verlag, Berlin, p. 81–100 Vargas Isaza OL (2004) La evaluación multicriterio social y su potencial en la gestión forestal de Colombia. Ph.D. Thesis, Doctoral programme in Environmental Sciences,. Universitat Autonoma de Barcelona, Barcelona Vincke P (1992) Multicriteria decision aid. Wiley, New York Zendehdel K, Rademaker M, De Baets B, Huylenbroeck V (2010) Environmental decision making with conflicting social groups: a case study of the Lar rangeland in Iran. J Arid Environ 74:394– 402. https://doi.org/10.1016/J.JARIDENV.2009.09.011

GIS Based/MCDA Modelling for Strategic Environmental and Social Assessment of Land-Use Planning Scenarios in Conflictual Socioecosystems Jean-Francois Guay and Jean-Philippe Waaub

1 Introduction The operational context of regional planning has become increasingly complex in the last 20 years or so. Regional planners, territorial managers as well as environmental psychologists have witnessed the emergence of unprecedented changes in societal Weltanschauung or worldviews while cognitive and ideological barriers are increasingly falling or changing, raising up more environmental awareness and several modifications in worldviews by actors and stakeholders (Gifford 2011). At the societal level, there is a fast-growing demand for stewardship from actors, including stakeholders,1 and individual citizens towards the decision maker. In this context, conflicting goals and aspirations are clashing with the goal to improve one’s environmental choices. Moreover, in the field, the outcome of these cultural changes brought up a complex contemporary challenge for local and regional planners who henceforth need to promote and apply innovative territorial planning solutions. When combined with the scarcity of land resources for development and growth, this challenge becomes intricate and requires unprecedented planning models. Finally, assessing the effects of a planning decision and possibly their impacts

1

Actors are individuals or groups of individuals in a decision-making process. Through their value system, they directly or indirectly influence the decision, be it in the first degree because of their intentions, or in the second degree because of how they involve the intentions of others (translation from Roy and Bouyssou 1993). Here, we associate the expression of stakeholders with organized groups of civil society and reserve the expression of public to individuals (Côté et al. 2017a, b).

J.-F. Guay (✉) Institut des sciences de l’Environnement, Université du Québec à Montréal, Montreal, Canada e-mail: [email protected] J.-P. Waaub Geography Department, Université du Québec à Montréal, Montreal, Canada e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 M. F. Norese et al. (eds.), Multicriteria Decision Aiding Interventions, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-28465-6_9

235

236

J.-F. Guay and J.-P. Waaub

related to local issues, on a city, a regional county municipality, or a region, requires a rich picture of the outcomes of this decision. The consequences of any planning decision are deeply rooted in a territorial basis. For this reason, it affects the lives of thousands if not millions of individuals but on a different timescale as it has different impacts. The decisions made are effective at the very moment they are endorsed by the decision maker but in an almost static environment so the effects of these decisions on the overall dynamic of the socioecological system (Biggs et al. 2021) are only partially known at the time the decisions are made. It is therefore of paramount importance to assess the impact of planning scenarios on the structure, processes, and dynamics of the socioecological system, using the appropriate prospective sectorial planning tools such as land-use plan and strategic environmental assessment. We present a prospective framework that involves the use of formalization and adapted modelling tools. These tools are dedicated to very specific premises in the situation we are addressing here. They are: (1) the interactions between natural and man-made components of regional space; (2) the transformation feedbacks between these components; (3) the general organization levels of the problem which ranges from spatial-temporal to political/normative in an almost bijective set of relations; (4) the use of different scales of measurement (cardinal, ordinal, and nominal, for objective and subjective issues and apparently uncountable entities); (5) the existence of several actor groups with differential belief/desire/intention schemes that might vary rapidly in time and space; (6) the overall objective of valuable knowledge generation starting from simple empirical observations. We have chosen a case study approach to illustrate the workflow, and methodology walk-through, intended for both experienced planners and new practitioners entering the land-use planning field, who need new tools to cope with the evergrowing complexity of the planning process. The authors’ proposition relies on experiences and practices in the field. Jean-François Guay is a regional planner agent accompanying Regional County Municipalities (RCMs), a GIS (Geographical Information System) analyst, and a decision modelling expert. He is working for the Quebec government. The Bellechasse RCM (object of the case study) is part of the territory he covers in his duties. His previous and actual experiences were widely used to obtain an accurate picture of the planning issues and for the elaboration of the decision modelling process. Jean-Philippe Waaub has extensive experience in environmental assessment and application of multicriteria methods in various contexts. Personally, he knows the region very well. We propose a methodology that facilitates the planning process, based on the principles of knowledge-based decision and expression of preferences and values by actor groups, including stakeholders. Our proposal relies on a systemic and constructivist way that allows the emergence of collective intelligence (“order”) from fuzzy and ill-defined situations (“disorder”), a new situation, characterized by intuition, creativity, and flexibility. It is important to mention that this full proposition is not mandatory in any RCM or municipalities since they do not have to implement one such approach in their planning process. However, it is recommended, in the highly transdisciplinary (Nicolescu 1996) field of regional

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

237

planning, to address complex and conflictual issues of this problem using adequate tools. An analysis of the context complexities is proposed in the next section, the proposed methodology is described in Sect. 3 and the case study in Sect. 4. Some recommendations and remarks conclude the chapter.

2 A Brief Overview of One Complex Problem The operationalization of strategic planning topic is a highly complex and conflictual one: plurality of administrative authorities and processes; multiple spatial levels and times horizons; multi-actors and diversity of participants including decision makers; conflicting opinions, perceptions, beliefs, values; types of knowledges; multiples dimensions, preoccupations, needs, issues including environmental and social impacts; links and interconnections. All these characteristics of a typical planning context are combined to form an almost inextricable problem for the territorial manager. In the absence of specific and interdisciplinary data to address these questions, the decision-making process becomes thus vulnerable to arbitrariness, particularly regarding the decision to be taken by accountable decision makers in a representative democracy. In the province of Quebec in Canada, the complexity of the local and regional planning process takes root in the legal apparatus of the province mainly by the expressions of legislative corpuses, a complicated and multilevel territorial organization and a differential composition of the territorial subdivisions, not only for the geographical and environmental conditions but also in terms of demography, labor market and economic sectors, education level, health and wellbeing conditions.

2.1

Legal Apparatus by the Expressions of Legislative Corpuses

The Act respecting land-use planning and development of 1979 (LegisQuebec 2021), the Act to preserve agricultural land and agricultural activities adopted in 1981 (Quebec 1999), the Environment Quality Act adopted in 1978 (Blakes 2018; LegisQuebec 2021), and their updates, are three major laws that dictate most of the regional planning scheme contents in the province. The Act respecting land-use planning and development a-19.1 is one of the most important laws in the province of Quebec regarding land-use planning and territorial management. It establishes the legal bases for the preparation and administration of the rules and obligations governing land-use planning and development. Under the Act respecting land-use planning and development (LAU), the government adopts planning policies that regional authorities must consider when they prepare planning documents (Ministère

238

J.-F. Guay and J.-P. Waaub

des Affaires municipales, des Régions et de l’Occupation du territoire 2010). In the province of Québec in Canada, in accordance with this Act, every Regional County Municipality (RCM; including about 10 to 15 municipalities) must maintain in force, always, an RCM plan, and every included local municipality’s urban plan, taking into consideration the whole of its territory in a coherent way. This RCM plan, and related urban plans, must determine the general aims of land development policy and identify the public policies on land use of the territory for its different parts. Moreover, it must allow the identification of zones where land occupation is subject to special restrictions for public safety or environmental issues. Accessorily, the RCM plan may identify any zone, mainly within an urbanization perimeter, which is likely to be, as a priority, the subject of land development or redevelopment, establish the rank of priority between zones thus identified and determine for such a zone, or for its different parts, the land uses and the approximate density of occupation. This RCM plan must be revised every 5 years or so. The planning objectives are mandatory and broadly the same for every RCM of the province: urban sprawl management, agricultural and forestry zone management, natural resources protection and preservation, biodiversity conservation, local development, and empowerment of actors and citizens, are among the most important issues to be addressed in this plan. The other mandatory competencies of the RCMs include land-use planning and development, residual materials management planning, fire planning, civil protection planning, assessment, and watercourse management. The optional powers of the RCMs generally include the adoption of a territorial development plan, the regulation on the felling of trees in private forests, the establishment and management of regional parks, the management of residual materials, the management of local roads, the management of public transit, the management and municipal financing of social housing, and supralocal infrastructures. The optional powers of the RCMs include assistance to economic development organizations, the acquisition of skills of local municipalities, and the delegation of certain powers by local municipalities. One of the important features of this Act is given by the nature of the Government’s involvement in the planning process. The government adopts land-use policies and guidelines to reflect the prescription of the Act and to which the local and regional authorities must conform in their planning documents. The main guidelines document of the Quebec government is the Government land-use policy statement. These strategies were built upon expertise and consensual work as they identified the problems faced by local municipalities, regional county municipalities, and metropolitan communities. This problem-setting tool is the vehicle of the government’s concerns and an instrument for exchange between the government and the regional county municipalities and metropolitan communities on land-use planning issues (Table 1).

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

239

Table 1 Governmental land-use guidelines statement and content (MAMH 2021) Groups of issues Urbanistic

Economic growth Natural resources management

Public safety

2.2

Content of the guidelines Urban growth distribution Improving general housing conditions Improving built/natural environments in urbanized environments Maintaining/improving services to citizens Integrated planning of the location of equipment /infrastructure Energy development Strategic planning of industrial/commercial spaces Land management of the public domain Protection of agricultural land Planning of mining activities Protection/management of forest environment Biological diversity conservation Natural and anthropogenic risks and nuisances

Territorial Organization

The territorial organizational level of the province reveals to be another challenge for the practitioner in regional planning. In 2020, the province of Quebec has three organizational levels consisting of administrative regions, regional county municipalities and metropolitan communities, and municipalities. The 87 regional county municipalities (RCMs) are entities responsible for the regional management of the included local municipalities, each with a power of jurisdiction and regulation devolved by the Government of Québec. The equivalent territory (ET) is a territorial collectivity whose administration is vested with the competences generally attributed to the RCM. The Quebec City ET and the Montreal ET are two of the most important ET in the province, which includes five ETs. There are two Metropolitan Communities (MCs) in the province, the Montreal metropolitan community and the Quebec metropolitan community. An MC is required to maintain in force, always, a metropolitan land-use planning and development plan for its territory, and to equip themselves with tools to monitor and implement it and to evaluate the progress made towards achieving its objectives and carrying out the actions it proposes. The metropolitan plan makes it possible to make choices and decisions in terms of land-use planning and development that affect all the RCMs, and agglomerations of a metropolitan community. Local municipalities are administrative entities ensuring the territorial management of public authorities that are established in a specific place and that enjoy power of jurisdiction and regulation devolved by the Government of Québec. As of April 2020, 1108 local municipalities are incorporated in Quebec. Within the framework of the territorial administration of Quebec, the powers of local municipalities are generally related to urban planning, economic development, roads,

240

J.-F. Guay and J.-P. Waaub

public transit, public safety, water distribution, the disposal of residual materials, recreation, and community life. Any part of the territory of Québec that is not that of a local municipality is an unorganized territory (French acronym: TNO) or an Indian reserve. The administration of a TNO is one of the mandatory competencies of an RCM. These territorial entities are different in the sense of geographical scale and environmental conditions, but their composition is also different in terms of demography, economic sectors, education level, health and wellbeing conditions, employment rate, and labor market. This situation obviously induces many discrepancies between citizens’ way of living, perceptions, and overall Weltanschauung or world view of their living environment. Contrasting beliefs, desires, intentions, and value schemes bring out conflict in interfacing environment.

2.3

Complexities and Proposal

This planning context, with the many planning documents and shared competencies between these organizational levels and the Legislator, is extremely difficult to manage, as the rule of compliance (MAMH 2021) forces the harmonization between all these documents.2 As a matter of fact, planners and territorial manager are undoubtedly facing complex situation on both societal, legal, and administrative sides of their practice. Their main challenges are to promote and apply innovative territorial planning solutions in this context where there is a growing demand for stewardship of the planning resources process from citizens, and the emergence of new societal value schemes with concurrent conflicting points of view. These elements lay the foundation for unprecedented planning models. In a world where complexity is already quite high and still increasing with time, normative potential planning combined with decision-making autonomy is required. The planners and decision makers continuously seek a territorial balance between order and disorder, between bureaucracy and policy, and between statics and dynamics. The multifaceted and interdisciplinary process by which the RCM plan is elaborated refers to regional planning and relates nowadays to what Proulx (2008) describes as an innovative territorial planning. The aims of the plans (“what to”) and the operational guidelines (“how to”) are all prescribed by the Land Use Planning and Development Act and by every provincial agency policy guideline.

2

The compliance rule is a mechanism that ensures consistency between the metropolitan land use and development plan, the land-use planning and development plan, the planning plan and planning by-laws and government interventions in the territory of a metropolitan community, an RCM, or a municipality. Given that land use planning is a function shared between various decision-making levels (government, metropolitan community, RCM, local municipality), the Act respecting land use planning and development introduces the compliance rule that ensures the concordance of the objectives and projects of the various decision-making levels through the various land use planning and urban planning tools provided for by law.

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

241

The territorial modalities (“where to”) are modulated or should be, in part, upon societal preferences. We suggest that these modalities rely on a co-constructed vision with local actors, including the stakeholders, regrouped on a restricted societal representativeness basis, concerned by the planning process, and forming a core working group. They are thus much more involved in the process than in the traditional approach consisting of mandatory public consultations at the very end of the planning process.

3 An Adapted Methodology Combining Tools The proposed approach and related models are named SOMERSET-P (Soft Multicriteria, Simulation, Environmental, Territorial—Planning). It innovates by combining, in an iterative process of contributory participation: (1) the design of planning scenarios; (2) their Regional Strategic Environmental and Social Assessment (R-SESA) based on issue-based methodology (Côté et al. 2017a, b); (3) the use of GIS and spatial analysis to model the scenarios and to perform their assessment; (4) a multicriteria decision aid methodology to compare scenarios in a multi-actor context to support decision maker(s); and (5) soft system approach (Guay and Waaub 2015). First, solutions to complex problems depend on the ability to combine the creation of strategic visions with short-term actions (Albrechts 2004) which implies the idea of scenario design and for which GIS are dedicated tools. As sequences of events objectively built and hypothetically close to scientific methods, scenarios provide reflective and critical tools intended for experimentation of a social system, help put emphasis on causal processes and nodes of decisions (Kahn and Wiener 1967; Meyer 2008). The interest in using scenarios remains in that they allow a better structuring of thought, the elicitation of the problem under consideration, and a first glance at the systemic structuring of the problem studied (Parrott 2011). Most of the time, planners propose one single plan (i.e., scenario). This limitative approach results in highly conflictual positions between pros and cons, between and among actors, including stakeholders and owners of the planning process. In this context, the scenario approach represents an important upgrade of the land-use planning decision process. Exploring and assessing multiple scenarios representing various actor positions allows to face conflictual issues upstream and give an opportunity to construct adhesion to the plan at the end. Accordingly, the quality of the decision process is of upmost importance. In the absence of more sophisticated analysis tools to synthesize information on the outcomes of any regional plan, it becomes extremely difficult to make compromises between the widespread environmental, social, and economic considerations raised by a plan and between the interests of the parties involved. Second, one step ahead in the use of planning scenario simulations is that it allows the assessment of impacts of one scenario compared to another, not only about the magnitude of the change in the affected component of the environment and

242

J.-F. Guay and J.-P. Waaub

society, but its importance (meaning) in relation to given issues. The assessment of regional planning scenario impacts by issues has several advantages (Côté et al. 2017a, b). First, it allows the user of the information to quickly assess the main environmental and social issues of a plan and to know how the latter were processed in terms of analysis. Second, it facilitates the framing of the impact assessment, as regards the determination of the scope of the analyses to be carried out and the sources of knowledge specialized to use. Finally, the creation of a scenario impact analysis by issues improves the planning process and its transparency by clearly identifying the workflow of the process. The whole legitimacy of the process has long been the weak point of the planning process since it has been heavily criticized by various actors to rely mostly on expert judgment and arbitrariness. The scenario approach facilitates the follow-up of the regional planning process at all stages: it helps in visualizing the valuable components of the socioecological system by actors; it allows the impact assessment of scenarios on these components, i.e., environmental, and societal components in terms of issues. Therefore, impact assessment is not only about the magnitude of the change in the affected component of the environment and society by one scenario, but in its importance in relation to given issues (e.g., The magnitude of the impact of cutting forest trees is not simply proportional to the number of hectares that will be cut. It is fundamentally linked to the nature of the environmental or social issues about the socioecological systems and the actors which will be affected by these cuts, and which, for example, can be the loss of biodiversity, economic activities, or even cultural practices). Third, GIS is recognized as an essential tool for spatial planning, prediction, modelling, research, and, ultimately, decision support (Keenan 2003 in Mora et al. 2003; Rinner 2018). The idea of using GIS (Geographic Information Systems) as a tool for decision support is not new (Tomlison 1968; McHarg 1969) and coupling GIS/decision analysis became a field of study in the late 1980s and gained more popularity with the emergence of the spatial decision support systems or SDSS concept (Densham and Goodchild 1989; Crossland et al. 1995; Prévil et al. 2004). SDSS are interactive computer-based platforms designed to support a user or group of users in achieving a higher effectiveness of decision analysis while solving a semi-structured spatial decision problem (Malczewski 1999). Taking into consideration the large proportion of data owned and managed by government and business organizations—by many accounts about 80% of all data include geographic references—spatial decision analysis is now an important field of study. Moreover, the emergence of government open data catalogs broadens the application potential of SDSS. It is plausible that there will be an increasing demand for spatial decision support services in the context of regional planning, mainly due to easy access to land-use/land-cover data and to the growing citizen demand for stewardship of planning resources and issues, which leads to the complexification of the process for decision makers (Rinner 2018; Friedmann and Sorensen 2019). Fourth, multicriteria decision aid (MCDA) methods propose tools consistent with the interdisciplinary character of regional planning process, since decision analysis in that field provides answers to questions raised by several stakeholders along with several and distinct territorial issues. Among MCDA methods, outranking methods

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

243

have been recommended for situations where there are a finite number of discrete alternatives to be chosen among, several decision criteria preferably limited between 10 and 20 but not always, and a potentially large number of actors sometimes referred as decision makers even if at the end of the process, most of the time, the final decision is taken by a very limited number of persons. An advantage of outranking approaches is that they avoid compensation between criteria (Ishizaka and Nemery 2013) as compensatory methods. Outranking methods are appropriate when measurement scales of the indicators used to assess the criteria, vary over wide ranges, and different units (Linkov et al. 2004). Among them, we have selected the PROMETHEE (Preference Ranking Organization METHod for the Enrichment of Evaluations) and GAIA (Graphical Analysis for Interactive Aid) methods (Brans and Marechal 1982, 1994, 2005; Taibi and Waaub 2015; Guay 2016; Aenishaenslin et al. 2019). They allow the use of a multicriteria analysis grid with different indicators and, therefore, different units of measurement. Moreover, the availability and ease of use of the VISUAL PROMETHEE software that implements these methods (Mareschal 2013) are serious advantages. It offers tools for producing and visualizing results that are very useful for the person in charge of decision support to prepare and communicate them to the actors, including the stakeholders. Pointing out scenario strengths and weaknesses, individual and multi-actor scenario rankings, scenario conflicts, and synergies, and actor groups analysis, are the specific outputs of this procedure. In addition, sensitivity and robustness analysis scenario rankings are computed to give this comprehensive process its full effectiveness and credibility among all the actors, including stakeholders and the decision maker. Despite those advantages of the method, Guitouni et al. (2010) have documented some weaknesses: the preference functions are defined on differences in performance although “these differences do not necessarily have the same meaning depending on their position on the scale associated with the criterion”; “the assumption of independence in the sense of preferences is (implicitly) made”; the method is not totally compensatory. The method has been criticized regarding the phenomenon of rank reversal (ranking can change when another scenario is added to or deleted from the set of scenarios), even if it is “inherent to pairwise-comparisons-based multicriteria decision aid methods such as for instance PROMETHEE, ELECTRE, AHP or Macbeth” and it is limited in the case of the PROMETHEE methods (Mareschal 2013). The multicriteria assessment process is at the same time a constructivist and cognitivist approaches. The key difference between constructivism and cognitivism is that constructivism explains that learners use prior knowledge to understand new knowledge, while cognitivism explains that learning takes place through the internal processing of information. Both intellectual postures form the backbones of intelligence creation in that context. It is thus suggested if not strongly recommended to any organization entering a multicriteria decision aid process to provide a formal training for each actor involved in the decision modelling process and tools used for decision analysis. This ensures the best possible transparency on the approach and avoids “black boxing” the outcomes of the process. Moreover, it is not recommended that decision makers and analysts perform “live” multicriteria

244

J.-F. Guay and J.-P. Waaub

processing. This should be avoided to bring as much as objectivity possible in the approach. Fifth and finally the modelling of a decision process involves stages where judgment must be made namely on weighting method, weights, aggregation method, and indicator construction. All these sources of subjective judgment might affect the ranking of the scenarios. However, judgmental aspects on the weights, on the construction and validity of the criteria and related measurement indicators, and on the overall assessment model, have all been validated by expert consensus and participatory process among the actor groups using one soft system approach (Agnew 1984; Checkland and Scholes 1990; Guay and Waaub 2015).

4 Real-World Application The case study was performed from a realistic territorial and environmental planning situation in the municipality of Sainte-Claire in the Bellechasse RCM, in the province of Quebec, Canada. Four planning scenarios were designed to represent different and contrasted planning visions built according to several planning objectives based on RCM plan requirements and guidelines. These scenarios (and the criteria and indicators used for their assessment) are realistic representations of different planning views in accordance with the accurate translation of actors’ preferences heard during hearings held by the Commission on the future of agriculture and agri-food of Quebec (Québec 2008a, b). We build four land-use planning scenarios and assessed their impacts, externalities, and amenities in accordance with 12 decision criteria and related performance indicators. The spatial translation and spatial analysis of impacts of the scenarios were performed within the ArcGIS geographic information system. Results were integrated into VISUAL PROMETHEE multicriteria analysis software implementing the PROMETHEE and GAIA methods for scenario comparisons and rankings.

4.1

Selection of the RCM, Logic of the Planning Process, and Data Sources

The selection of the Bellechasse RCM for our case study is justified by several geographic, territorial, and economic factors. The RCM is located on the south shore of the St. Lawrence River, south of the agglomeration of Levis in the Quebec metropolitan community. This situation gives the RCM a strong suburban character. The interfacing of rural areas with urban areas brings out likelihood for disruptive conflicts since the urban immigration process—increasing over the years—appeal to the polarization between more traditional sector activities (agriculture) and periurban activities like residential clusters buildup and the disseminating of

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

245

non-farming uses activities. Although the economic and environmental costs of urban sprawl are well documented (Brabec and Smith 2002; Wilson and Chakraborty 2013), the attraction to living in the countryside has not wavered and agricultural land and immediately adjacent areas are affected by this longing to move beyond urban boundaries. The following territorial information give a short portrait of the Bellechasse RCM (Guay and Waaub 2015): an area of 1760 km2; 20 municipalities; 35,000 inhabitants (growth rate for the 2011–2016 horizon is 17%); four important peri-urban municipalities including Sainte-Claire; nearly 85% of the MRC’s area is zoned for agriculture; 866 farms in 2018 (48 certified organic in 2019); agricultural sector is very dynamic in the northern part with a more intensive and specialized agriculture; greater fragmentation of cultivated acreages is observed in the southern part with a more marginal and more diversified agriculture. This case involves actual issues identified by real actors, including stakeholders, and gathered from previous works such as public hearings, and data along with empirical observations. More specifically, actors’ system, problem elicitation, issues selection, deliberation, negotiation, and conflict resolution were processed from the work of the Commission on the Future of Agriculture and agri-food of Quebec that was held across the whole province from 2006 to 2009 including in the concerned Bellechasse RCM. These public hearings and the resulting Pronovost report (Quebec 2008a, b) are acknowledged today by everyone as the cornerstone for the modernization of Quebec’s agriculture policies. The full planning process revolves around eight steps regrouped into three major phases. Phase 1, problem structuring concerns (1) the problem setting and the actor’s system definition, (2) the construction of scenarios, and (3) the structuring of issues translated in terms of a limited set of decision criteria. Phase 2, assessment concerns (4) the evaluation of the scenario’s performance by criterion (indicators, scales, preferences); (5) the formalization of the existing value systems, currently referring to the weighting of criteria, and (6) the performance of all the criteria on every scenario, followed by the aggregation of the overall preferences of the scenarios which provide the ranking of the scenarios. Phase 3, choice and decision refer to (7) the construction of a robust group of scenarios, and (8) the recommendations toward the selection of the most consensual scenario(s) submitted to the final decision makers who will decide the one to be implemented. It consists of the synthesis of all the information gathered from the two previous steps, namely scenario modelling and simulation, and their assessment, to the complete ranking of all of them. Several data sources (Table 2) were utilized for the first two phases. For the identification and structuration of issues and criteria, a third source of data was derived from expert advice from the Territoire database which is maintained by the MAMH. This web-based platform contains all official notices of the Quebec’s government produced over the years in response to planning decisions taken by the region’s RCM and MC. This advice focuses on the judgments by regional planning specialists, of detailed planning decisions and scenarios and their aftermath on the region’s agricultural zone in terms of losses or gains of agricultural acreages. We

246

J.-F. Guay and J.-P. Waaub

Table 2 Data sources for the problem-setting step (Guay and Waaub 2015) Data source • Briefs/reports • Plan for agricultural zone of the RCM • Experts’ advice

• Agricultural zone planning policy guidelines and regional reforesting policy guidelines

Inferred information contents Typology of actors Concerns, issues, societal values. General structures of the system. Interactions between urban zones/agricultural protected zones (APZ) Actors’ interactions Spatial interactions between actors/land Conflicting area issues Deforesting issues General processes of the system Non-agricultural use planning limitations in the APZ APZ planning standards Farmers and foresters’ rights Intention of the owner of the system

extracted and reviewed all the advice that Jean-François Guay performed over the years on many issues. Such decisions include, among several others, agricultural spot zoning, territorial land-use dispositions and modifications, non-agricultural uses of agricultural acreages, deforesting policies, and shoreline protection policies among others. For the fine detail of the main interaction between the different actors of the system, we collected several valuable information from agricultural regional development plans (Agriculture Protection Zoning: APZ). As a prominent actor in the overall accomplishment of these plans, Jean-François Guay was involved in the negotiation process with the main actor in this sector. These plans permitted to gather intelligence since they focused more precisely on the agri-food and agriculture sector of the region. These APZ put emphasis on the interactions between farmers, foresters, public authorities, and stakeholders specifically in the agricultural zone. Preoccupations and issues of the agricultural sector were gathered from these plans. Finally, the governmental guidelines on agricultural planning and the regional reforesting policy allowed us to make a more accurate assessment of the actors, including the stakeholders, involved in the systems and more specifically in their prerogatives, rights, or constraints. In accordance with these guidelines, we were able to point out important characteristics and processes that are undergoing in the agricultural zone: diffuse urbanization, water resource, woodlands preservation, and land occupancy distribution. This analysis gave us sufficient knowledge to identify the actors in the system, their roles, functions, concerns, and values as well as their interaction. Moreover, it allows us to get insights into general structures and processes of the socioecological system and its environment and to apprehend what would be the best criteria/ indicators combinations. The careful analysis of the professional advice combined with the information gathered from participation in agricultural regional

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

247

development plans committees enabled us to point out some emergent phenomena like land-use conflicts that are a part of the overall dynamic of the socioecological system.

4.2

Problem Structuring

Step 1, concern problem setting and identification of the actors. Gray literature used to build the actor system and get an adequate portrayal of the structures and processes of the system, was based on the works of the Pronovost Commission held in 2007 in the province of Quebec. This Commission on the future of agriculture and agri-food of Quebec had the mandate to review the issues and challenges facing the agriculture and agri-food sectors in Quebec through environmental, health, land use, and regional development. In a thorough consultation process, the Commission met with farmers, economic development organizations, ecologists, elected municipal officials, professionals involved in agricultural production, researchers, educators, citizens, and consumers. The 660 briefs and presentations reflected a great diversity of viewpoints. These documents expressed the concerns, hopes, expectations, and ambitions of many hundreds of people from all professional milieus (Quebec 2008a, b). The Commission also held 2 weeks of public hearings where it received 110 briefs presented mostly by regional or Quebec-wide organizations. The Chaudière-Appalaches region was the host of one of the regional hearings as well, during which 50 briefs and reports were received from all the stakeholders. This information from secondary sources constitutes a valid basis to identify the main actors of the system (Cote and Waaub 2000). They are classified according to the four categories of the typology proposed by Prades et al. (1998): • Public authorities: Represented by civil servants and deriving their legitimacy from the elected power, i.e., ministry agencies and territorial manager. This category was defined as the OWNERS group. • Economic actors: Deriving their legitimacy from their economic power, i.e., new rural dwellers. This category was defined as the NEORURALS. • Groups of stakeholders: Deriving their legitimacy from their belonging to the civil society, i.e., environmental groups and watershed groups. This category was defined as the ECOLOGISTS. • Individual experts or groups: Deriving their legitimacy from science or their professional practice. This category was defined as the FARMERS and FORESTERS. Regarding Step 2, the design of four planning scenarios was performed by creation and/or manipulation, selection and/or reclassification, or generation of new attribute layers from the state variables. Each of the scenarios has its own planning logic (Table 3) and is based on the beliefs, desires, and intentions of every actor group, in accordance with the type of general worldview prospective:

248

J.-F. Guay and J.-P. Waaub

Table 3 Design rules supporting the logic of the scenarios (Guay and Waaub 2019) Issues Economic prosperity

Criteria Industrial crops Agroindustry localization

Agritourism localization

Urbanistic and cohabitation

Resources preservation and biodiversity Forestry and agricultural land development

Territorial vitality and health

Concentric sprawl

Scenarios Status Quo 75% of all cultivated surfaces Low agricultural activity sector 225 m buffer from urban center 250 m buffer from existing UP

Growth 95% of all cultivated surfaces High agricultural activity sector 290 m buffer from urban center

Ecotopia 0% of all cultivated surfaces Very high agricultural activity sector 500 m buffer from urban center

Exurbia 50% of all cultivated surfaces Devitalized sector

900 m buffer from urban center

500 m buffer from existing UP

No sprawl

10 ha lots 2500 m2/ home devitalized sectors only 1m 0%

50 ha lots 9500 m2/ home + agricultural plan

500 m buffer from existing UP only in forested sectors 50 ha lots 2500 m2/home devitalized sectors only

10 m 50%

5m 25%

No None

Diffuse sprawl

5 ha lots 2500 m2/ home

Riparian strips Organic crops

3m 5%

Deforestation Reforestation

Yes Low agricultural activity sector 30%

Yes Every sector

50%

No Very high agricultural activity sector 10%

Moderate

High

Moderate

Moderate

Low

Moderate

High

Low

Available cutovers Social harmony Empowerment

40%

(1) sustaining gains by status quo (Base), (2) economic growth (Growth), (3) environmentally friendly (Ecotopia), or (4) hybrid multifunctional (Exurbia); they correspond to several differentiated land uses and land covers. The baseline land-use/ land-cover mapping of the socioecological system of the municipality of SainteClaire consists of eight state variables: (1) current land uses (including brownfields acreages), (2) actual forest stand, (3) municipality centroid positioning, (4) agricultural exploitation units positioning, (5) industrial crop positioning, (6) road network distribution, (7) river system, and (8) cadastral subdivisions. Cartographic features of the scenarios are described hereunder.

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

249

• Point features: Farms/agribusiness/agritourism infrastructure. Imported from an existing database or digitally created and positioned in the geographical space in accordance with the actor’s preferences and the sought scenario. Farm positioning was used for agriculture activity intensity assessment with node density algorithm in the Spatial Analyst extension for ArcGIS software. • Surface features layers: Industrial crops, organic crops, diffuse sprawl, and available cutover, cadastral lots: Imported from existing cartographic databases and manipulated to simulate land-use and land-cover classes by random selection of lots in the cadastral database and reclassification to the sought classes in accordance with actor’s preferences and the sought scenario. • River system layer: Imported from existing cartographic databases and exploited for riparian strips simulation by generation of differential protective strips with the buffer algorithm in the Spatial Analyst extension for ArcGIS software. The same buffering operation was performed around urban perimeters’ land-use class to simulate concentric urban sprawl phenomena. • Road systems: Imported from existing databases and exploited to assess distance constraints for the simulated positioning of some infrastructure. The first purpose of scenario building and simulation specifically, is to compare contrasting hypothesis on evolution trends of the territory. For strategic regional planning, scenarios often consist of contrasted territorial affectations. As such, they correspond to several differentiated land uses and land covers which form the baseline maps of the territory under study. In the field of regional planning, the rationale for these simulations often relies upon one realistic hypothesis of net changes in demography during a period of reference, usually a 15-year period (e.g., 2015–2030 in the case study). Incidentally, the predicted population level is an important input into forecasts of housing demand and thus housing land requirements. It is also important in creating local demand for goods and services, and thus affects the level of local economic activity (Schmitt 1952; Field and McGregor 2018). By postulating a net positive change, we assume it will lead to densification of the urban perimeter (UP) to a saturation point around which they should be concentrically extended to absorb new housing constructions. Built-up of urbanizing pressure would lead to diffuse sprawl as well as on agricultural and forested lands and overall land-use/land-cover conflicting situations. The four scenarios are described hereunder (Adapted from Guay and Waaub 2019). Figure 1 illustrates the Growth scenario as an example. Status quo scenario (Base) has the characteristics described hereunder. • Urban perimeter can be spatially extended as needed without any territorial constraints. • Agricultural area is dedicated to traditional agricultural activities, i.e., monoculture. • There are little acreages for organic cultures. • Logging is permitted in many places as well as deforestation for agricultural purposes.

250

J.-F. Guay and J.-P. Waaub

Fig. 1 Growth scenario (Growth) (Guay and Waaub 2019)

• Environmental considerations may be summed up to the application of a protective strip on the banks of every stream and lake. • Replacement of traditional agriculture by reforestation is a conventional activity but is exceptional. • Social harmony is not a major concern and conflictual situations can arise about land uses between actor groups. Growth scenario (Growth) has the characteristics described hereunder. • Urban sprawl is intense. • Deforestation for cultivation purposes is allowed. • Cohabitation is challenged by several clashes between farmers, neo-rural dwellers, loggers, and environmentalists. • Deforestation can jeopardize thresholds for biodiversity. • Deforestation for agricultural purposes and urbanization is allowed and lowers the urban land uses/forested lands ratio to a critical level.

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

251

• Environmental considerations are minimal; there are only a few considerations for issues such as preservation and conservation of natural resources, although there are serious issues to be taken care of, but which are not. • Reforestation of fallow land is widely favored, which can lead to agriculture acreages losses. • In socially homogenous areas, low differentiation of ways of life leads to a generally peaceful cohabitation. Environmentally friendly (Ecotopia) has the characteristics described hereunder. • Biodiversity and resource protection (forested lands, organic agricultural acreages, and water quality) are paramount criteria in all the planning interventions by the final decision maker. • Preservation of the agricultural area integrity is a major concern in this scenario. • Strong ecological beliefs, desires, and intentions (BDI) are sustained by environmentalists and concomitants’ planning actions are brought to the extreme. • Wide riparian strip. • Few residential constructions are allowed on agricultural lands. • Large portion of agricultural lands is intended for organic agriculture. • Urban sprawl in agricultural zone is not allowed. • Neo-rural dwellers, farmers, and environmentalists have conflicted visions of what should a sustainable land use plan. • There is a total lack of comprehension between farmers and Neo-rural dwellers. Rebalancing scenario (Exurbia) has the characteristics described hereunder. • In this scenario, agricultural lands acquire a strong multifunctional character. • Houses can be built in the agricultural zone, yet the process is closely monitored. • Agricultural zone is acknowledged for economic externalities (agritourism among others), social externalities (dynamic and sustainable land occupancy, cultural heritage, sense of belonging, and community identification), and environmental externalities (biodiversity, landscape, and lifestyle valuation). • On the urban fringe, strong densification of urban perimeters has already occurred and extensions must be given into the agricultural area. • To avoid this situation, residential opportunities are exceptionally permitted in some less dynamic area of the agricultural zone that allows home settlings. • Social harmony is generally good. Step 3, the identification and structuration of issues and criteria, completes the problem structuring process. Some approaches allow stakeholder groups to start with their own list of criteria and indicators. On our side, we simulated a process that would have focused on clarifying conflicting issues in order to co-construct a single performance table corresponding to a common and shared vision of the problem definition. This first-hand knowledge and information obtained during the provincial hearings along with the other data sources, represent the field baseline for the present analysis on which we determined what would be the best criteria and indicators to

252

J.-F. Guay and J.-P. Waaub

Table 4 From issues to criteria/indicators (Adapted from Guay and Waaub 2019) Issues Economic prosperity (ECO)

Criteria Agricultural vitality (ViAg) Logging (coup) Agrotourism (Lcl2)

Urbanization management (URB) Biodiversity and environment (RES) Forestry and agricultural management (FOAG) Moral health of the community (TER)

Agribusiness (Lcl1) Concentric urbanization (UrC) Diffuse urbanization (UrD) Protection of water resources (Hy1) Organic crops (Cbio) Agricultural deforestation (dba) Wasteland recovery (Fri) Social harmony (Str) Contribution to empowerment (Emp)

Indicators Area under cash crops

Unit Hectares

Scale Cardinal

Available exploitable forest area Distance from a public market to the urban centroid Level of agricultural dynamism Cultivated areas lost

Hectares

Cardinal

Meters

Ordinal

Classes UEV/km2 Hectares

Ordinal

Number of residences in agricultural areas Width of riparian strips

Whole nb. Meters

Cardinal

Area under organic crops

Hectares

Cardinal

Number of residences in agricultural areas Reforestable areas

Boolean

Nominal

Hectares

Cardinal

Level of harmony

Classes

Ordinal

Value associated with contribution to empowerment

Whole nb.

Rank

Cardinal

Cardinal

assess in accordance with the issues pointed out by the actors, including the stakeholders (Table 4). Elaboration of the multicriteria performance matrix requires the conduct of sectoral studies on specific themes mobilizing both scientific knowledges carried by experts in various fields (biology, sociology, archaeology, etc.) and vernacular knowledges and concerns carried by a diversity of actors (knowledges of the territory by local populations). These criteria are known to be most relevant in several land-use management plans and planning policies in compliance with the Quebec’s Act respecting land-use planning and development, the Act to preserve agricultural land and agricultural activities and the Environment Quality Act. The problem structuring phase is fundamentally the most important in any planning and decisional process. The understanding of this problematic situation is the result of the commitment of the actors and it evolves with this commitment in an iterative and non-linear way. In accordance, the knowledge of the situation is systemic and greatly subjective by nature since it emerges from the interactions between the actors whose points of view are diversified. This so-called co-constructive approach for generation of ideas is of paramount importance at the very beginning of the process and strong considerations point out toward a human

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

253

resource formed to animation techniques, before entering the process. This seasoned facilitator whose role will be to improve the comprehension of the ill-defined situation any organization is entering is one basic requirement of problem structuring. At the same time, its contribution is mandatory to help the actors identify criteria and indicators which will be used in the assessment of planning scenarios. Once again, this phase relies on a constructivist approach. It is performed with large-scale hearings, medium-scale focus groups, or fine-scale interviews although existing issues datasets resulting from previous audits and other types of surveys can be exploited for this purpose. At the other end of the subjective/objective spectrum, scenario building requires at the same time, one deep multidisciplinary knowledge of the regional planning context as well as ease with geographical information systems to perform data acquisition and manipulations. These rule-based contrasting and shared visions of the future are indeed defined by issues that are translated into cartographic language by design variables. Thus, in order to ensure an efficient transition between the many normative elements of the planning scenarios and their spatiotemporal realization, it is necessary that the land-use planner and the geomatics specialist team up. Indeed scenario modelling is data consuming and requires top-notch technological tools such as GIS packages and database management system. In recent years, a large number of thematic data sources have been made available free of charge on government servers, which greatly facilitates the acquisition of geographic information needed for the spatial modelling process. Thus, World Map Services or WMS provides a means of accessing geo-referenced map images and/or vector and raster data over the internet. These can be viewed, manipulated, transformed, and analyzed in a geographic information system for planning purposes.

4.3

Assessment

Step 4, the evaluation of the scenario’s performance by criterion necessitates the choice of measurement indicators, and related scales (Table 4). This choice relies on data availability. Data sources are presented in Sect. 4.1. For feasibility reasons, it should also consider available expertise, time, and budget. Moreover, this step involves the determination of preference function associated to each criterion; it “defines how pairwise evaluation differences are translated into degrees of preference,” reflecting the perception of the criterion scale by the actor (Mareschal 2013). VISUAL PROMETHEE is offering six different shapes of preference functions and an assistant for helping good choices (Mareschal 2013). Thus, we have selected linear preference function (V-Shape with its preference threshold) for quantitative scales, and usual function for qualitative scales. This information is integrated in the performance matrix at step 6. Concerning Step 5, actors’ priorities reflecting value schemes were obtained by weighing each criterion. This information is integrated in the performance matrix (Step 6) and is of paramount importance since it directly affects the aggregation of

254

J.-F. Guay and J.-P. Waaub

preferences. As for PROMETHEE, “the weight of a criterion is a positive number that represents the criterion relative importance” (Mareschal 2013). Other approaches like Macbeth or AHP are also using weights but with different meanings (Guitouni et al. 2010). Weight elicitation of every criterion was performed to reflect the actual preferences scheme of every actor group. Thus, every group has its own weight sets in accordance with its own priorities and issues regarding the planning objectives. As it is obvious that the interests of actors are divergent (if not strongly polarized), the relative importance which they attach to the criteria is not the same, so each actor involved in the process is usually called to do its own criteria weighting. There are no agreed existing methodologies to weigh individual criteria. Several methods have been proposed (Nardo et al. 2008) to determine weights and the selection of the most appropriate method depends on the decision-making context. As PROMETHEE does not provide any formal guidelines for weighting, VISUAL PROMETHEE provides a hierarchical tool to help the actors. For the purposes of this demonstration, we proceeded by a direct weighting approach as made possible by the software. The low complexity/software requirements of the direct approach were best suitable and are widely used. Combined with the actor’s value scheme previously expressed in the hearings (Quebec 2008a) and our knowledge of the territorial planning processes in the agricultural zone (Guay and Waaub 2015), we have very good confidence on the weight elicitation process which we performed. It offers a good balance between the precision tolerance of the study, the complexity of questions, the size of the decision-making group, and the available resources for constructing the MCDA framework (Németh et al. 2019). As VISUAL PROMETHEE offers also to compute results for all actor groups, the question of the balance of power of each group is important. This is modelled as a weight attributed to each group. In our case, none of the actor groups were distinguished by their relative weight within the entire group. Each of the five groups receives a weight equivalent to 20% of the total weight for all actors. This reflects the representative democracy principle, at least in theory, in the decision-making process. However, this balance of power might be negotiated in advance or established by law as in the case of relative power of municipalities in an RCM (e.g., large cities benefit from more votes than little villages). Step 6, concerns the performance of all the criteria on every scenario, followed by the aggregation of the overall preferences of the scenarios which provides, among other outputs, the rankings of the scenarios. The measurement of every criterion against every scenario mobilizes sectoral expertise that might be very costly and time consuming. In our case, performance of every criterion was delivered by spatial simulation. Fortunately, GIS allows us to proceed efficiently. The results of the criteria performance are all displayed in the performance matrix. This matrix integrates all data need to compute decision aiding results within the VISUAL PROMETHEE software. It figures on the top left the name of the actor group, and then the list of criteria which can be grouped in clusters (e.g., the five issues identified in Table 4) according to a choice of symbols and colors. First column gives then the parameters: units of indicators and, cluster symbols and colors (2 lines); information about preferences, namely, criteria to be minimized or

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

255

maximized, weights, preference functions, and associated thresholds express in absolute or relative terms (Q = indifference; P = strict preference; S = Gaussian parameter) (7 lines); statistical data related to the performances, namely minimum, maximum and mean values, and standard deviation (4 lines); information on performances with identification of the scenarios. If needed, they can also be grouped in categories according to a choice of symbols and colors (not in our case). Best performances are in green, worst in red, and others in black. Our performance matrix is 12 criteria by 4 scenarios. The aggregation of preferences was carried out in three distinct procedures offered in VISUAL PROMETHEE. PROMETHEE I (partial ranking allowing for incomparability between scenarios, indicating that the choice is difficult), and PROMETHEE II (complete ranking easier to explain even with a loss of information) provide normative results. GAIA visual analysis provides descriptive results, computed from principal components analysis which is a dimension-reduction technique, representing the relative positions of the scenarios in two dimensions instead of all the dimensions related to the criteria. The GAIA plane figures the scenarios (points), the criteria (axes), and the weights of the criteria (decision axis). In PROMETHEE I graph, the intersection of two axes reveals that two scenarios are incomparable from the actor’s point of view. This incomparability situation arises when the strengths of one scenario correspond to the weaknesses of the other and inversely. Therefore, both scenarios cannot be compared because there are too many conflicting issues. As an example, at the level of analysis of all groups of actors, there is an incomparability situation between Base and Exurbia. However, PROMETHEE II provides a complete ranking of the scenarios for each actor group (Fig. 2) and for all groups. In Fig. 2, Ecotopia is ranked first for three actor groups and second for the two others. For the Farmers and Foresters groups Growth is ranked first. The complete ranking (a value from +1 to 0 in green and from 0 to -1 in red) for all actor groups indicates Growth and Ecotopia, respectively, in first and second positions. Conflicting complete ranking from one group to another is well illustrated in Fig. 2. On the one hand, it is not a big surprise that Growth is highly preferred by Farmers. On the other hand, the very slight preference for Growth over Ecotopia for Foresters, and the consensus about among Owners, Neorurals, and Ecologists about Ecotopia are interesting results emerging from the analysis. It is a critical information that should be analyzed together with other outputs, in the process to build the best possible consensus. The software computes also a PROMETHEE II complete ranking for all actors group. This global ranking reveals the standings of each scenario for all the actor groups taken together. This ranking is also confirming that Growth is strongly preferred, followed by Ecotopia, Exurbia, and then Base which have close slight negative net flows and are clearly rejected. GAIA planes are powerful graphical visualization tool for the analysis of the multicriteria problem. At the level of analysis for all groups, GAIA provides two planes. The first one is the criteria and scenarios plane representing the problem for one actor. The second one is the actor groups and scenarios plane, representing the

Owners

Foresters

Fig. 2 Complete ranking PROMETHEE II for each actor group

-1.0 Deux à deux Tous les scénarios

Base

Growth

Exurbia

Ecotopia

1.0

Farmers Neorurals

-1.0

Base

0.0 Exurbia

Growth

Ecotopia

1.0

Ecologists

256 J.-F. Guay and J.-P. Waaub

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

257

problem at the global level of all actors. Typically, GAIA planes (criteria/scenarios) are different for each actor, but in our case study, as we choose to elaborate a common and shared vision of the problem by all actors regarding the scenarios and criteria considered, only, the decision axis is pointing in a different direction as they represent the weight set attribute to the criteria for each actor. Those GAIA planes are providing a global view of criteria/actors (axis from the center of the plane), scenarios (points), and weights (decision axis). Mareschal (2013) presents in detail how to interpret the GAIA planes. The main rules are listed here. Scenarios that are similar/very different to each other appear close/far away to/from each other. Subset of scenarios can be identified. Axis is pointing in the direction of preferred scenarios. Longer the axis, the more the criterion/actor is discriminating. If no axis points in the direction of a scenario, it means that it is not preferred. Axis pointing in the same/opposite direction mean a synergy/conflict between criteria or a coalition of actors. The identification of subsets of criteria makes it easier to understand the conflicts that must be solved in deciding. The plane illustrates conflicts or synergies between criteria/actors. Orthogonal projections of scenarios on an axis (Criterion/actor/decision axis) indicate how well scenarios are performing on a criterion, or for an actor, or in the case of the decision axis, for all criteria/actors (quite equivalent to the PROMETHEE II ranking). The distance is meaningful (far from the origin and in the direction of the axis, is better). This information is of course limited by the quality of the GAIA plan (Delta is indicating the quality of information due to the projection in a plane; higher rate is better). Figure 3 illustrates the conflicting situations between actor groups, and the coalitions of actor groups, in relation to the ranking of the scenarios. The decision axis represents the weights attributed to each actor group (basically in our case, equal weights). In Fig. 3, we can see two main areas (circles) showing for one part, the existence of converging points of view between Owners, Neorurals, and Ecologists around Ecotopia and for the other part, a cluster area formed by Farmers and Foresters around Growth. The orthogonal projection of the scenarios on the decision axis illustrates the PROMETHEE II ranking with a loss of information related to the fact that it is the projection in a plane of the multidimensional space of the ranking of the scenarios for every actor group. Figure 3 confirms that Growth and Ecotopia are the first two and that the search for a compromise can be made between them. The Farmer’s group is strongly associated with Growth, and it would be more difficult to negotiate a compromise with this group than with the Foresters group. A more detailed analysis of GAIA plane with criteria and scenarios offers more insight into the subtle aspects of the problem. Figure 4 illustrates the case of the Farmers. Thus, the criteria for performing well on Ecotopia differ from the one on Growth. Indeed, the criteria dealing with resources preservation and biodiversity, as well as urbanistic and cohabitation are pointing in the direction of Ecotopia. Those criteria are pointing to the different direction, even if not in full opposition, that the criteria dealing with territorial vitality, economic prosperity, and agricultural land development are pointing in the direction of Growth. In the field, that situation could become conflicting although it offers great negotiation possibilities because these criteria are not completely in opposition. This is useful to determine the negotiation

J.-F. Guay and J.-P. Waaub

258

Fig. 3 GAIA plane for all actor groups and all scenarios

opportunities to avoid, reduce, or compensate the impacts, which could be concurrently discussed by all actor groups and should strongly be considered by the final decision maker during the final phase of the planning process. For the decision maker, it means that his/her planning could focus mostly on actions closely linked to these criteria, i.e., politics and planning orientations directed toward territorial vitality/economic prosperity and resources preservation/biodiversity/cohabitation. Figure 3 give good indications about coalition of actor groups, Fig. 4, which correspond to individual actor group, allows to identify on what they agree or not. It might happen that actor groups agree on a preferred scenario but for different reasons (criteria).

4.4

Choice and Decision

Step 7 is about seeking a robust group of scenarios through sensitivity and robustness analysis. Sensitivity analysis (variation of one parameter at a time) on criteria

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

259

Fig. 4 GAIA plane, criteria, and scenarios for the Farmers

and/or actors’ weights is heavily recommended and is most helpful when the stability intervals tool available in VISUAL PROMETHEE, indicates that a little variation of one weight value results in a significant change in the ranking position of the scenarios. Most of the time, the focus is put on potential changes among the first two or three scenarios. In our case, the focus was on the first two scenarios among four, and no rank stability problems were identified. Robustness can be explored either by varying sets of parameters according to different hypotheses or by multiplying sensitivity analysis on parameters and even on performance data. To assess the robustness of the global ranking we computed, and for which the relative importance of every actor group point of view for every scenario has been equally distributed, varying the weight of actor groups allows to see how the influence of a more demanding group could change the ranking. In general, no group has the power to change the complete ranking of scenarios. In our case, even if the relative importance of each of the groups in terms of decisionmaking weight varies markedly, the ranking remains the same. There is one

260

J.-F. Guay and J.-P. Waaub

exception for the farmers: further claims by this group change the ranking but not inverse it. Growth comes first just ahead of Ecotopia only by a slight dominance. Precisely, this means that none of the groups have a decision advantage over another. This situation arose from the good performance of environmental criteria over all the others, among most of the groups of criteria. In contrast, Base (status quo) is not a viable alternative for all actor groups. Step 8, decision and recommendations necessitate to communicate the results of the decision aid process and formulate recommendations about the decision to be implemented. After performing the multicriteria analysis of the planning scenarios, we pointed out that the best-to-worst scenario, in accordance with their full ranking, would be Growth, closely followed by Ecotopia. Exurbia and Base are rejected or in other words, are strongly dominated by the two previous scenarios. Moreover, these results suggest there is a necessity for the decision maker to promote planning interventions, actions, politics, and/or strategies, which would allow, at the same time, economic prosperity to the extent where this scenario is assorted with coherent natural resources and biodiversity protection policies. Since most actors would support these decisions, at the same time it would give one certain legitimacy to the decision maker in regard to the choice made. The main challenge here would not be to find the best scenario but to point out the adequate compromise between these two major issues and among the actor groups.

5 Concluding Remarks and Recommendations for Planners The SOMERSET-P (Soft Multicriteria, Simulation, Environmental, Territorial— Planning) approach appears to offer powerful tools which can be embedded in any planning processes by many territorial organizations, as they are cost-efficient in terms of data requirements, hardware, and software resources. Besides its potential for sharp analysis of highly complex problems characterized by conflicting and/or controversial issues which arise from ecological, social, economic, and social proximity, the approach may be very useful to foresee possible outcomes of the decision taken. Within the planning procedure, monitoring is an important part of the disciplinary field of land management, and it could very well benefit from SOMERSET-P regarding this topic. The R-SESA and the application of MCDA methods in a multi-actor context make it possible to improve the territorial planning process by formulating several scenarios and assessing them by means of a multicriteria performance matrix considering the environmental, social, and economic consequences that each entails. MCDA methods allow the integration of value systems carried by the actors at each stage of the process leading to the decision (problem structuration, assessment, decision). The SOMERSET-P spatial and decision aiding models which we presented here allow to visualize the possible futures for a region and the potential outcomes—in terms of conflict/harmony—between actors, including stakeholders, according to their preferences and Weltanschauung or world view.

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

261

However, the application of such an approach needs some basic requirements. First, and foremost, the professional supporting the decision-making process must be both an expert and a mediator to bring actors together since the final decision relies inevitably on one strategic value scheme that might not reflect every preoccupation at the scale of actors’ groups and stakeholders. Moreover, from hearings and issues identification to GIS modelling and decision-making, the whole process needs to be performed by one support team including at least one regional planner and one GIS specialist. Regarding the use of existing data, those from consultations or public hearings, it is important to mention that the actors do not like to be solicited several times on the same subjects. In a planning process, it is therefore always necessary to consider the steps that have been taken previously in order to benefit from them to the maximum rather than wasting time and resources to redo what has already been done. Finally, it is suggested if not strongly recommended to any organization entering a territorial planning process to provide a formal training, for each actor involved, to the decision aid tools and procedures used for the spatial and decisional modelling steps. This ensures the best possible transparency on the approach and avoids “black boxing” the outcomes of the process. Moreover, it is not recommended that decision makers and analysts perform “live” the software in front of the actor groups and stakeholders. Often, the general perception about the tool changes rapidly from natural mistrust to the temptation to be able to manipulate the results by playing with the parameters regardless of their real meanings. It is preferable to prepare the meetings with a “participant’s notebook” (paper copy or electronic documents allowing more interactions with maps, tables, figures, photos, videos, or complementary information) which allows each actor to appropriate the problem based on their preferences, and by a “group notebook” which also allows them to have the global vision of the problem. Also, following deliberations, or negotiations, it is often useless to redo a model run to see what is happening. What counts here is the collective evolution towards a solution to be implemented with the support of the greatest possible number of actors and by treating well those who are not satisfied. Complete and total objectivity is not attainable in decision modelling and thus it is not neutral. It conveys a reference frame of values that will contribute to orient the process and thus engage the result (“the decision”). We are talking here about a “value-engaged” process, and this is very often the case in territorial planning where the final decisions are almost never apolitical. By encouraging the expression of a diversity of points of view and by allowing the actors to understand the tools used, one increases the social legitimacy of the analysis process as well as the accountability of the decision maker. That said, actors value scheme is evolving, obviously slowly but undeniably, in any organization, be it a government agency, an RCM council, or within a particular group of actors. The impacts of this BDI’s evolution at every organizational level has headmost implication on the overall dynamic of collective intelligence. This requires the planners and the decision makers to perform one such modelling process every 5 years or so to ensure they have the best situational understanding of context of their socioenvironmental context. In a world of constantly increasing in complexity, strategic decision making can hardly

262

J.-F. Guay and J.-P. Waaub

be performed without any decisional bias. Actually, no decision process is completely neutral, as policy issues become entangled with science in a situation of sometimes unclear logic. As a result, the fundamental role of the decision maker, over and above any other technico-scientific question related to the decision analysis, will be to bring the best possible consensus into a necessarily subjective process.

References Aenishaenslin C, Bélanger D, Fertel C, Hongoh V, Mareschal B, Waaub J-P (2019) Practical guide to establishing a multicriteria and multi-actor decision-making process: steps and tools. Les cahiers du GERAD (G-2019-03). Montréal, Québec, Canada. Available at https://www.gerad. ca/fr/papers/G-2019-03. Accessed 25 November 2022 Agnew CT (1984) Checkland’s soft system approach: a methodology for geographers? Area, Published by the Royal Society of Geographers 16(2):167–174 Albrechts L (2004) Strategic (spatial) planning reexamined. Environ Plan B Plan Des 31(5): 743–758. https://doi.org/10.1068/b3065 Biggs R, De Vos A, Preiser R, Clements H, Maciejewski K, Schlüter M (eds) (2021) The Routledge handbook of research methods for social-ecological systems. Taylor & Francis, London Blakes (2018) Environmental law and regulation in Quebec. Blakes Annual Overview Brabec E, Smith C (2002) Agricultural land fragmentation: the spatial effects of three land protection strategies in the eastern United States. Landsc Urban Plan 58(2–4):255–268. https://doi.org/10.1016/S0169-2046(01)00225-0 Brans J-P, Mareschal B (2005) PROMETHEE methods. In: Figueira JR, Greco S, Ehrgott M (eds) Multiple criteria decision analysis. Springer, Berlin, pp 112–131 Checkland PB, Scholes J (1990) Soft systems methodology in action. John Wiley and Sons, New-York Cote G, Waaub J-P (2000) L’évaluation des impacts d’un projet routier: l’utilité de l’aide multicritère à la décision. Cahiers de géographie du Québec 44(121):43–64 Côté G, Waaub J-P, Mareschal B (2017a) L’évaluation d’impact environnemental et social en péril. VertigO 17(3):1–31. https://doi.org/10.4000/vertigo.18813 Côté G, Waaub J-P, Mareschal B. (2017b) At risk: the environmental and social impact assessment. The need for action. Les cahiers du GERAD (G-2017-42). Montréal, Québec, Canada. Available at https://www.gerad.ca/fr/papers/G-2017-42 Accessed 25 November 2022 Crossland MD, Perkins WC, Wynne BE (1995) Spatial decision support systems: an overview of technology and a test of efficacy. Decis Support Syst 14:219–235. https://doi.org/10.1016/01679236(94)00018-N Densham PJ, Goodchild M (1989) Spatial decision support systems: a research agenda. GIS/LIS '89. Proc. annual conference, Orlando, Vol. 2 Field BG, McGregor BD (2018) Population in forecasting techniques for urban and regional planning. Routledge, London Friedmann J, Sorensen A (2019) City unbound: emerging mega-conurbations in Asia. Int Plan Stud 24(1):1–12. https://doi.org/10.1080/13563475.2019.1555314 Gifford R (2011) The dragons of inaction: psychological barriers that limit climate change mitigation and adaptation. Am Psychol 66(4):290–302. https://doi.org/10.1037/a0023566 Guay J-F (2016) Contributions méthodologiques à la modélisation intégrée pour la planification territoriale et environnementale. Thèse présentée comme exigence partielle du Doctorat en sciences de l’environnement, UQAM. Montréal, Québec, Canada Guay J-F, Waaub J-P (2015) Application of a territorial soft system approach for conceptual modelling of an agro-ecosystem. Environ Syst Decis 35(3):363–374. https://doi.org/10.1007/ s10669-015-9536-7

GIS Based/MCDA Modelling for Strategic Environmental and Social. . .

263

Guay J-F, Waaub J-P (2019) SOMERSET-P: a GIS-based/MCDA platform for strategic planning scenario’s ranking and decision-making in conflictual socioecosystems. EURO J Decis Process 7(3–4):301–325. https://doi.org/10.1007/s40070-019-00106-4 Guitouni A, Bélanger M, Martel J-P (2010) Cadre méthodologique pour différencier les méthodes multicritères. Rapport technique-DRDC Valcartier, TR 2009–386 Recherche et développement pour la défense Canada Available at https://cradpdf.drdc-rddc.gc.ca/PDFS/unc103/p534172_ A1b.pdf. Accessed 25 November 2022 Ishizaka A, Nemery P (2013) Multi-criteria decision analysis: methods and software. Wiley, Chichester Kahn H, Wiener A (1967) The year 2000. Macmillan, New York Keenan P (2003) Spatial decision support systems. In: Mora M, Forgionne GA, Gupta JN (eds) Decision making support system: achievement, trends and challenges for the new decade. Idea Group Publishing, London, pp 28–39 LegisQuebec (2021) Act respecting land use planning and development Québec (Canada). Available at: https: //wwwlegisquebecgouvqcca/fr/contenu/noteinfo. Accessed 25 November 2022 Linkov I, Varghese A, Jamil S, Seager TP, Kiker G, Bridges T (2004) Multi-criteria decision analysis: a framework for structuring remedial decisions at contaminated sites. In: Linkov I, Bakr Ramadan A (eds) Comparative risk assessment and environmental decision making. Springer, Dordrecht, pp 15–54 Malczewski J (1999) GIS and multicriteria decision analysis. Wiley, New York Mareschal B (2013) Promethee Methods. Visual promethee 14 Manuel. VP Solutions. Available at http://en.promethee-gaia.net/. Accessed 25 November 2022 McHarg I (1969) Design with nature. Natural History Press, Garden City Ministère des Affaires municipales, des Régions et de l’Occupation du territoire (2010) Planning and development powers in Québec. Direction générale des politiques, ministère des Affaires municipales, des Régions et de l’Occupation du territoire. Bibliothèque et Archives nationales du Québec. Available at https://www.mamh.gouv.qc.ca/fileadmin/publications/amenagement_ territoire/urbanisme/plan_development_powers_angl.pdf. Accessed 25 November 2022 Mora M, Forgionne G, Gupta JND (2003) Decision making support systems: achievements and challenges for the new decade. Idea Group MAMH (Ministère des Affaires municipales et de l’Habitation) (2021) Guide de la prise de décision en urbanisme, Québec (Canada). Available at https://www.mamh.gouv.qc.ca/amenagement-duterritoire/guide-la-prise-de-decision-en-urbanisme/planification/regle-de-conformite/. Accessed 25 November 2022 Meyer V (2008) La méthode des scénarios: un outil d’analyse et d’expertise des formes de communication dans les organisations. Études de communication 31:133–156. https://doi.org/ 10.4000/edc.778 Nardo M, Saisana M, Saltelli A, Tarantola S, Hoffman A, Giovannini E (2008) Handbook on constructing composite indicators: methodology and user guide. OECD publishing; JRC47008, Paris Németh B, Molnár A, Bozóki S, Wijaya K, Inotai A, Campbell JD, Kaló Z (2019) Comparison of weighting methods used in multicriteria decision analysis frameworks in healthcare with focus on low-and middle-income countries. J Comp Eff Res 8(4):195–204 Nicolescu B (1996) La transdisciplinarité. Manifeste, Éditions du Rocher Parrott L (2011) Hybrid modelling of complex ecological systems for decision support: recent successes and future perspectives. Ecol Inf 6:44–49. https://doi.org/10.1016/j.ecoinf.2010. 07.001 Prades JA, Loulou R, Waaub J-P (1998) Stratégies de gestion des gaz à effet de serres : le cas des transports urbains. Presses de l’Université du Québec, Québec Prévil C, Saint-Onge B, Waaub J-P (2004) Aide au processus décisionnel pour la gestion par bassin versant au Québec: Étude de cas et principaux enjeux. Cahiers de Géographie du Québec 48(134):209–234

264

J.-F. Guay and J.-P. Waaub

Proulx M-U (2008) Quarante ans de planification territoriale au Québec. In: Gauthier M, Gariépy M, Trépanier MO (eds) Renouveler l’aménagement et 1’urbanisme. Planification territoriale, débat public et développement durable. P.U.M. Montréal, pp 23–54 Quebec (1999) Act respecting the preservation of agricultural land and agricultural activities. A v a il ab l e at htt p :/ /w ww .c pta q. g o uv . q c . c a / fi l e a d m i n / e n / p u b l i ca t i o n s/ f o r m s/ guideforfillingoutauthorizationformLPTAA. Accessed 25 November 2022 Québec (2008a) Agriculture et Agroalimentaire: Assurer et bâtir l’avenir. Rapport final de la Commission sur l’Avenir de l’Agriculture et l’Agroalimentaires du Québec, Gouvernement du Québec. Ministère de l’Agriculture, des Pêcheries et de l’Alimentation Québec (2008b) Agriculture et Agroalimentaire: Assurer et bâtir l’avenir, ce qu’on nous a dit. Rapport de la Commission sur l’Avenir de l’Agriculture et l’Agroalimentaires du Québec, Gouvernement du Québec. Ministère de l’Agriculture, des Pêcheries et de l’Alimentation Rinner C (2018) Spatial decision support. In: Wilson JP (ed) The geographic information science and technology body of knowledge (2nd Quarter 2018 Edition). Topic GS-25. Available at https://gistbok.ucgis.org/bok-topics/spatial-decision-support. Accessed 25 November 2022 Roy B, Bouyssou D (1993) Aide multicritère à la décision: méthodes et cas. Economica, Paris Schmitt RC (1952) Demography and city planning. Soc Forces 30(3):300–304 Tomlison R (1968) A geographical information system for regional planning, Gouvernement du Canada, Ottawa Available at https://gisandscience.files.wordpress.com/2012/08/1-a-gis-forregional-planning_ed.pdf. Accessed 25 November 2022 Taibi B, Waaub J-P (2015) L’approche multicritère et la prise de décision dans les entreprises publiques, le cas de l’Algérie. Les Cahiers du GERAD (G-2015-32). Montréal, Québec, Canada. Available at https://www.gerad.ca/fr/papers/G-2015-32.pdf Accessed 25 November 2022 Wilson B, Chakraborty A (2013) The environmental impacts of sprawl: emergent themes from the past decade of planning research. Sustainability 5(8):3302–3327. https://doi.org/10.3390/ su5083302