Epistemic economics and organization forms of rationality and governance for a wiser economy

In each part then, the missing and new epistemic perspective, on economic activity and its organiza-tion as a problem of knowledge, is outlined: decision making as a process of imaginati

ROUTLEDGE STUDIES IN GLOBAL COMPETITION

Epistemic Economics and

Organization

Forms of rationality and governance for

a wiser economy

Anna Grandori

Epistemic Economics and

The implications for contracts and organizations, sustained also by insights from law, are shown to be far reaching, including a new view of the nature of the firm as an entity-establishing agreement under which to discover uses of resources under uncertainty, and as a democratic institution

Anna Grandori is Professor of Business Organization at Bocconi University,

Milan, Italy

1 Japanese Firms in Europe

Edited by Frédérique Sachwald

2 Technological Innovation,

Multinational Corporations and

New International

Competitiveness

The case of intermediate countries

Edited by José Molero

3 Global Competition and the

Edited by Frédérique Sachwald

10 Multinational Firms and Impacts on Employment, Trade and Technology

New perspectives for a new century

Edited by Robert E Lipsey and Jean-Louis Mucchielli

11 Multinational Firms

The global–local dilemma

Edited by John H Dunning and Jean-Louis Mucchielli

12 MIT and the Rise of Entrepreneurial Science

15 European Union Direct

Investment in China

Characteristics, challenges and

perspectives

Daniel Van Den Bulcke,

Haiyan Zhang and

Maria do Céu Esteves

16 Biotechnology in Comparative

Perspective

Edited by Gerhard Fuchs

17 Technological Change and

Economic Performance

Albert L Link and

Donald S Siegel

18 Multinational Corporations and

European Regional Systems of

20 Local Industrial Clusters

Existence, emergence and evolution

Thomas Brenner

21 The Emerging Industrial

Structure of the Wider Europe

Edited by Francis McGowen,

Slavo Radosevic and

Nick Von Tunzelmann

The US Advanced Technology

Program’s Intramural Research

Initiative

Albert N Link and John T Scott

24 Location and Competition

Edited by Steven Brakman and Harry Garretsen

25 Entrepreneurship and Dynamics

in the Knowledge Economy

Edited by Charlie Karlsson, Börje Johansson and Roger R Stough

26 Evolution and Design of Institutions

Edited by Christian Schubert and Georg von Wangenheim

27 The Changing Economic Geography of Globalization

Reinventing space

Edited by Giovanna Vertova

28 Economics of the Firm

Analysis, evolution and history

Edited by Michael Dietrich

29 Innovation, Technology and Hypercompetition

Institutional framework and learning in information technology

in Japan, the US and Germany

Edited Cornelia Storz and Andreas Moerke

32 Entry and Post-entry Performance of Newborn Firms

Marco Vivarelli

33 Changes in Regional Firm

Founding Activities

A theoretical explanation and

empirical evidence

Dirk Fornahl

34 Risk Appraisal and Venture

Capital in High Technology New

Ventures

Gavin C Reid and Julia A Smith

35 Competing for Knowledge

Creating, connecting and growing

Robert Huggins and Hiro Izushi

36 Corporate Governance, Finance

and the Technological

Advantage of Nations

Andrew Tylecote and

Francesca Visintin

37 Dynamic Capabilities between

Firm Organisation and Local

Systems of Production

Edited by Riccardo Leoncini and

Sandro Montresor

38 Localised Technological Change

Towards the economics of

Edited by Jorge Martinez-Vazquez

and François Vaillancourt

42 Evolutionary Economic Geography

Location of production and the European Union

45 Innovation, Knowledge and Power in Organizations

Evidence from Europe

Edited by Franco Malerba

49 Innovation in Complex Social

Systems

Edited by Petra Ahrweiler

50 Internationalization, Technological Change and the Theory of the Firm

Edited by Nicola De Liso and Riccardo Leoncini

51 Territory, Specialization and

Continuing and emerging patterns

in Japan and China

Cornelia Storz and

Sebastian Schäfer

53 Innovation and Economic Crisis

Daniele Archibugi and

Andrea Filippetti

54 The Communications Industries

in the Era of Convergence

60 The Economics of Creativity

Ideas, firms and markets

Edited by Thierry Burger-Helmchen

61 Epistemic Economics and Organization

Forms of rationality and governance for a wiser economy

Anna Grandori

Epistemic Economics and

Organization

Forms of rationality and governance for a wiser economy

Anna Grandori

First published 2013

by Routledge

2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN

Simultaneously published in the USA and Canada

by Routledge

711 Third Avenue, New York, NY 10017

Routledge is an imprint of the Taylor & Francis Group, an informa business

© 2013 Anna Grandori

The right of Anna Grandori to be identified as author of this work has been asserted by her in accordance with the Copyright, Designs and Patent Act 1988.

All rights reserved No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers.

Trademark notice: Product or corporate names may be trademarks or

registered trademarks, and are used only for identification and explanation without intent to infringe.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloging in Publication Data

Grandori, Anna

Epistemic economics and organization: forms of rationality and

governance for a wiser economy/Anna Grandori.

p cm

1 Knowledge management–Economic aspects 2 Epistemics

3 Organizational learning–Economic aspects I Title

HD30.2.G723 2013

ISBN: 978-0-415-57565-2 (hbk)

ISBN: 978-0-203-78677-2 (ebk)

Typeset in Times New Roman

by Wearset Ltd, Boldon, Tyne and Wear

1 ‘Models of man’ and the ‘rationality divide’ 9

2 Savage and Simon revisited: how both ‘maximizing’ and

‘satisficing’ simplify problems 13

3 Endogenizing assumptions: contingent rationality 19

4 The ‘psychology’ versus the ‘logic’ of judgement and

Contracts and the firm beyond transactions: the

1 Contract incompleteness and the rationality divide 44

2 How both relational contracting and authority relations have

limited capacity of governing uncertainty 47

3 Contracting without knowing 52

4 Ten theses on the nature of the firm 63

5 Relations with extant views of the firm 73

6 Summary 77

x Contents

PART III

Organization design beyond comparative assessment: the

1 Organization forms and forms of rationality 81

2 How both markets and hierarchies decompose problems 83

3 Missing alternatives: non decomposable systems and panarchic

7 The negotiated discovery of organizational arrangements 112

8 Properties of robust economic organization in uncertain

II.1 Contractual specification of property rights and action rights

and obligations in multi-party projects 56–57II.2 Incidence of capital invested by type, and property rights

assignments in inter-organizational projects 69 A.1 Organizational elements and practices 127

A.4 Formulas for high efficiency (complexity as contingency) 128A.5 Formulas for high efficiency (uncertainty as contingency) 129A.6 Formulas for innovation (complexity as contingency) 129A.7 Formulas for innovation (uncertainty as contingency) 130A.8 Formulas for both high efficiency and innovation (complexity

A.9 Formulas for both high efficiency and innovation (uncertainty

BATNA Best alternative to a negotiated agreement

CM Coordination mechanism

CQA Comparative qualitative analysis

DSA Discrete structural alternatives

ER Epistemic rationality

GDP Gross domestic product

HC Human capital

IPO Initial public offering

KGP Knowledge, governance and projectsL&E Law and economics

OD Organization design

OE Organizational economics

OT Organization theory

PR Property rights

PRT Property rights theory

TCE Transaction cost economics

VC Venture capitalist

The book sets out the main tenets of a research program in economics and ization capable of explaining and designing actions and structures in a discovery oriented, knowledge intensive and risky economy

It is commonly recognized that the modern economy and society are ingly characterized by those traits The recent and increasingly frequent eco-nomic ‘crises’ are also widely recognized as signals that there is something wrong in the way decisions are taken and economic organization designed It is also increasingly recognized that the traditional economic models fall short of providing a good guide for action What is less clear or convincingly developed

increas-is what the alternative models might be

In fact, economics has so far been ‘reformed’ mainly by having been made more ‘psychological’, ‘behavioural’ and ‘experimental’ That route is guided by the good intention of making economics more ‘realistic’ and better able to

‘predict’ the most common behaviours This approach is certainly useful for the prediction of normal behaviours, and therefore also for designing some aspects

of economic organization that should take real average behaviour into account (such as laws and rules, or incentives) By contrast, the approach entails a risk of anchoring institutions to ‘wrong’ or biased behaviours that could be improved to start with; the risk of being founded on ‘reduced’ standards of rationality Organ-izational and behavioural models have neglected to a large extent the problem of

valid knowledge construction and effective knowledge governance; and have

focussed on ‘experiential learning’ as an alternative to thinking ahead and to foresight However, in a strongly uncertain and variable world, experience, at least used in the inductive way often implied by the ‘common patterns of behav-iours’ incorporated in behavioural models, is anything but a logically sound basis; actually it is a most unreliable and dangerous way of proceeding (as it has been pointed out in major classic works – e.g Bandura 1986)

This book – and indeed all my work, on which it builds – takes another route The aim is to contribute to making economic organization ‘wiser’ and ‘more robust’, and in that sense ‘more rational’ rather than less rational, in the face of uncertainty A main shift, or reform of assumptions, on which the attempt rests

is a disentanglement between what the state of knowledge (risk and uncertainty, unpredictability) is, from what rationality is Not knowing is one thing, not being

2 Introduction

rational is quite another The challenge is precisely how to address rationally a situation of high uncertainty and risk The ‘original sin’, in a way shared both by economics and behavioural approaches, is to equate the two things, and assume that if knowledge is imperfect, than decision makers are forced to be ‘less rational’ This widespread assumption has blocked the development of rational approaches for a high risk, high uncertainty world – the core theme of this book and the approach that has always inspired my work The approach is based on a broadening of the sense in which decisions can be said to be rational, connecting that sense to the meaning that rationality has in philosophy and science – where

it emphatically does not mean ‘knowing everything’ The foundations of nomic reasoning have been divorcing from this wider and wiser form of rational-ity and thought, in which the core problem is one of generating and testing valid knowledge about what we are doing, before calculating the costs and benefits of action This book is about how to reconnect economics and organization to the problem of discovery and of knowledge validity and growth Therefore, the approach is both empirically grounded and prescriptive, as it is in all those con-tributions to the theory of knowledge interested in detecting, in science and other discovery activities, ‘the best patterns of thought and judgement’ (Lakatos 1970a; Kiss 2006) that humans are capable of, not the most frequent patterns Once modelled, the superior patterns can become more diffuse and common The three parts of the book critically revisit the main working hypotheses that are currently dominating, respectively, at the micro-level of assumptions about economic rationality, at the level of coordination and contracting among actors, and at the level of governance and organization structures In each part then, the missing and new epistemic perspective, on economic activity and its organiza-tion as a problem of knowledge, is outlined: decision making as a process of imagination and discovery, contractual structures able to govern processes of collective discovery, and organization structures as a set of mechanisms for gov-erning the growth of knowledge

The first part of the book integrates inputs from economic and behavioural science with insights and models from the philosophy of knowledge and of eco-

nomic philosophy, to define new micro-foundations: neither a calculative,

deduc-tive and omniscient ‘rational actor’; nor an experiential, adapdeduc-tive and biased

‘behavioural actor’; but a theory building, knowledgeable, imaginative stemic actor’ A model of behaviour that in the course of being heuristic (in the sense of based on a logic of discovery) is also rational (in the sense of logically sound according to the canons of reason) has been called for by only a couple of outstanding thinkers in economic sciences (Shackle 1972; Loasby 2004), and little attention and effort has been invested in this direction The time is more than ripe now though, as testified by the increasing attention paid to knowledge processes in economic, organizational and management fields Hence, I thought

‘epi-it sensible to bring together all my work inspired by an ‘epistemic’ approach to economic behaviour and organization in this book In fact, I embarked on my scientific enquiry with the thesis that economic decision making and coordina-tion is first and foremost an epistemic and discovery problem, before being a

Introduction 3

problem of allocating given scarce resources and of conciling given interests and proposing an ‘extension’ of the available Simonian heuristic decision model to incorporate more innovation oriented discovery heuristics (Grandori 1984) Later

on, I mobilized some scholars, sharing the idea of developing the neglected sides

of Simon’s legacy, to contribute in this direction (Grandori 2001a) and started work on a repertory of ‘rational heuristics’ for innovation (Grandori 2010a)

On the basis of those micro-foundations, the book then presents the ‘new facts’ that a research program in epistemic economics and organization can explain and predict as to the nature and shape of contracts, firms and economic organization structures, using available evidence in organizational economics (OE) and organization theory (OT) research, as well as the empirical ad hoc research conducted by myself and associates over the years, as follows

Part II focuses on contracts and the firm It reviews the limits of the classic

solutions of ‘relational contracting’ and of authority or power relations in aging uncertainty The notions of ‘constitutional governance’ and ‘contracts of society’ are then introduced to overcome those limits They turn out to provide

man-a foundman-ation to the firm thman-at is linked to uncertman-ainty man-and knowledge but pendent from power or authority – so commonly linked to the ‘nature’ of the firm The implications for the nature of the firm are drawn and exposed in ten theses that overall should free our view of the enterprise from the nineteenth-century dichotomies between central planning and ‘spontaneous’ market order, and more recent dichotomies between contracts and organizations These prop-ositions on the nature of the firm are based on a ‘rational reconstruction’ of the need for firm-like associations, rather than starting from the average large capi-talistic firm as it has become diffused in the last century The result is to rescue the firm from being a special world based on power rather than on right, violat-ing not only the constitutional principles applied to any other association, but even the Hayekian law that no complex system can be governed by planning and command

Part III develops the implications of the two former parts in terms of zation forms and their design This analysis ‘closes the circle’ in that it substi-

organi-tutes the classic nexus between bounded rationality as a micro-foundation and

‘discrete structural alternatives’(DSAs) as a macro-structural consequence, with the nexus between the more rational but still heuristic notion of epistemic ration-ality (ER) as a micro-foundation and a more continuous space of qualitatively different structural configurations as a macro-structural consequence Consistent with the generative and discovery-based approach outlined in the book, some

guidelines for design-as-discovery of new configurations are specified The

sub-stantive features of the configurations more suitable for sustaining discovery and managing uncertainty are also illustrated They include ‘nearly-indecomposable’ rather than nearly-decomposable forms; ‘panarchic’ rather than hierarchic and polyarchic governance regimes; ‘robust’ rather than adaptive structures; and

‘multimodal’ rather than ‘coherent’ forms

In disciplinary terms, the analysis builds on, and contributes in integrating,

OE, L&E (law and economics), OT, and behavioural science as applied to

4 Introduction

understanding and designing economic organization I have tried to implement (who else if not me?) the ‘methodological options for an integrated view of eco-nomic organization’ that I myself proposed in the Millennium Special issue of

Human Relations (Grandori 2001): the ‘endogenization of assumptions’; the

‘disentanglement of concepts’; and the ‘rejuvenation of design’.

In my writing, I have also tried to keep faith to, and hopefully fulfil, the judgements and expectations on the positioning and potential of my contribu-tion provided by some of the anonymous reviewers of this book Thanking them heartily for their confidence in the project and their ameliorating sugges-tions, I will let their words define the positioning and potential interested readership:

This book is positioned at the interfaces of economics, strategic ment, entrepreneurship, and innovation It thus addresses a wide range of audiences The book will also raise interest in the growing behavioural eco-nomics community and in constitutional economics, as well as neighbouring social sciences such as psychology, sociology, and philosophy of knowledge

manage-It tackles questions largely debated by articles published in highly ranking international Reviews, in the fields of Organization Theory, Economics, and Management

The material is highly innovative since it envisages a new perspective on the creation and the evolution of firms on the basis of three fundamental ingredients: (1) recent advances in the theory of rationality (2) A peculiar expanding of the vision of firms according to a ‘generative’ and ‘combinato-rial’ mental frame, which models decision making under structural uncer-tainty, viewed as discovery process, with multifaceted organizations forms (called multimodal and multifunctional) (3) A sound endeavour to over-come the prevailing theoretical approach, based on “discrete structural alternatives”

Dedication

This book is dedicated not to the ‘memory’ but to the ‘living legacy’ of two fathers, both recently disappeared I cannot even trace all the debts I have toward them, but I can perceive with clarity that the debts are large and that my grati-tude is immense

One is my real father, Giuseppe Grandori Perhaps nothing more than the intense relation and endless discussions with him, one of the pioneers of modern seismic engineering, on how to frame and solve complex problems marked my development and thought Certainly, the ‘generative and combinatorial mental frame’ that has been noticed in my work has a root in the kind of engineering as

‘ingeniousness’ in which my father was a master

Introduction 5

Perhaps, among the reasons for studying decision making and ‘adopting’ Herbert Simon as a scientific father there has been the similar interest he cultivated for ‘the world as it might be’ This election or adoption was somehow recog-nized by Simon, who wrote to me on the occasion of the publication of my 1984 ASQ paper on contingent decision making that it was ‘a step forward in our understanding of decision making’ One can imagine what a motivational and reassuring effect those simple words had on a young Italian female researcher at the beginning of her scientific enquiry more than 30 years ago; on top of that daring to criticize, since that very first article, the ‘satisficing’ model for being a particular and ‘too behaviourist’ version of a more important ‘heuristic-as- epistemic’ approach to decision making But, as in any healthy inheritance relationship, the acknowledgment of the importance of the heritage should never become the acceptance of everything received Actually this book, and most of

my work of which it expresses the ‘fil rouge’, builds on an active interpretation,

a revitalization through reinterpretation, of Simon’s work and of the perspectives

it originated in economics, organization and management

Acknowledgements

For this kind of book, I should actually acknowledge all colleagues in my research network who continue to nurture my thought and work Some research and edit-orial projects have been prominent in the formation and evolution of the core ideas, especially of the second and third parts of the book Early salient networked expe-riences in the 1990s included the series of workshops that led to my becoming one

of the co-founders of the EGOS Standing Group on Network Research; and having been one of the scientific directors of the ESF Research Program on European Management and Organization in Transition’ – the conjecture that ‘breaking’ DSAs and resorting to a combinatorial approach was ‘necessary’ for understanding and designing economic organization, matured there

In the last decade, three projects have had a particularly important impact, and are in fact widely cited in the book One has been the ‘Knowledge, Governance and Projects’ research programme, promoted by, and conducted in collaboration with the research centre (CRORA) I was directing at the time

The other two projects have been editorial projects mobilizing researchers to whom I felt akin in both concern and approach (at least in the effort of integrating rather than dividing ‘schools’ and ‘disciplines’) around a joint output They were

the edited volumes Corporate Governance and Firm Organization in 2004, and the Handbook of Economic Organization currently in press I warmly thank all the

colleagues who accepted the invitation to join those projects; their contributions and discussions form a very direct pillar on which this book rests

A particular and personal thanks goes to the younger scholars who have been working and co-publishing with me over the years – Giuseppe Soda, Massimo Neri, Santi Furnari, Marco Furlotti, Magdalena Cholakova (more or less in a gen-erational order) –: their contribution is quite intermingled within the research on which this book is based

Part I

Micro foundations

From bounded to epistemic rationality

1 ‘Models of man’ and the ‘rationality divide’

2 Savage and Simon revisited: how both ‘maximizing’ and ‘satisficing’ simplify problems

3 Endogenizing assumptions: contingent rationality

4 The ‘psychology’ versus the ‘logic’ of judgement and discovery

5 The logic of economic discovery: an epistemic decision model

6 Conclusions

7 Summary

The term rationality, (from the Latin ‘ratio, rationis’) according to its

encyclo-ing the activity of knowing’: something is rational if ‘it proceeds from reason’, if

fol-8 Micro foundations

logically prior to the questions that are typically addressed in extant decision making models, where problems are defined to start with, and alternatives are at best ‘searched’, not ‘researched’

‘Logically sound procedures’ can be defined with respect to various logical operations relevant in the activity of knowing and deciding In particular, two distinctions are particularly relevant for discussing and renewing how the notion

of rationality is employed in economic and behavioural sciences, and will be widely used in this book First, knowing activity involves both a ‘logic of justifi-cation’ and a ‘logic of discovery’ of any proposition (a central distinction in phi-losophy of science, as is well known) (e.g Lakatos and Musgrave 1970; Simon 1977) Second, decision making involves two differing broad categories of judgements, both calling for logically sound procedures: rationality in ‘instru-mental’ judgements (what are the best means to ends) and rationality in ‘epis-temic’ judgements (is it true that those means have those effects? Is it true that those ends are desired?) (Foley 1987)

In economics, the meaning of rationality has been restricted to refer only to the logically sound procedures for ordering alternatives according to preference and to the logic of justification of choice among them (Shackle 1972) Hence, it can be said that economics has been concerned with instrumental rationality – a set of log-ically sound procedures for consistency in utility judgements and means to end choices – not with epistemic rationality – a set of logically sound procedures for knowing what the preferences, alternatives and consequences are in the first place Second, a judgement expressed by Lakatos (1976) about mathematics could be applied equally well to economics: it has been concerned only with the ‘deductive’ rather than with the ‘heuristic’ component of rationality, that is only with the ‘logic

of justification’, not also with the ‘logic of discovery’

This use of the term ‘heuristics’ also throws light on another, related, match of meaning between the philosophy of science and knowledge on one side, and economic and social sciences on the other ‘Heuristics’ in the philo-sophy of science is a branch or sub-discipline concerned with ‘the logic of dis-covery’: the set of ‘logically sound procedures for researching and finding’, for generating conjectures and controlling them, the very method of empirical science ‘Heuristics is neither psychology nor logics, but an independent discip-line, the logic of discovery’ (Lakatos 1976, appendix II, note 4) As such (some)

mis-‘heuristics’ can be qualified as ‘rational’, in the same sense as the methods of scientific empirical research can be defined to be rational: they are logically correct methods for discovery (Grandori 2010a) Hence, the concept of ‘heuris-tics’ has also been restricted, and ‘downgraded’, in behavioural and economic sciences usage: it is used to refer to psychological devices, substituting cognitive

‘shortcuts’ and ‘fast and frugal’ thinking to deductive rationality and statistical decision theory As such, ‘heuristics’ have been associated with quick and low

effort choice (Gigerenzer et al 1999) or with ‘biases’ (Kahneman et al 1982)

rather than with rational discovery

Building on those observations, and on a long lasting research program oriented by them (Grandori 1984, 2001a, 2010a), Part I of this book aims at

Micro foundations 9

oretically, then, the view proposed in this first part of the book originates from the observation of the mismatches between the broader philosophical meaning of rationality and the narrower meanings attributed to the term in economic and behavioural sciences; and will show how the latter reduced meanings are at the origin of many puzzles and problems in important areas such as contract theory and innovation management Empirically, it originates from the observation that actual behaviours inspired by instrumental rationality only, on calculating costs and benefits, all too often lead to proceeding in quite a distorted manner, disre-garding all the important considerations about the validity of action hypotheses that are considered, and on ‘simplifying’ problems too much for making those calculations possible (rather than implying ‘wide’ or even ‘complete’ knowledge

providing renewed and enlarged micro-foundations to economic behaviour The-of the world, as they are commonly said to imply) On the other side, actual behaviours saving on effort and based on limited search and limited understand-ing of the problem at hand, all too often ‘leave resources on the table’ This expression, commonly employed in negotiation analysis is evocative of the real problems I am pointing at: suppose that a negotiator or a designer of a new product adopts the rule of stopping at the first satisfactory alternative found, or lowers their aspirations in the face of difficulty Even in problems with no closed boundaries, with infinite possible solutions, much better can be done, and

is often done

The analysis will feature the following aspects and be developed through the following analytical steps:

• revisiting and criticizing the standard ‘rationality divide’ between the

‘rational actor paradigm’ and the ‘bounded rationality paradigm’, offering a different account or interpretation of the two models, allowing to interpret them not as incomparable or rival ‘paradigms’ but as comparable and con-tingently rational decision strategies;

• specifying a further, missing, ‘epistemic’ model of decision behaviour that,

in the course of being heuristic (aimed at discovery) is also ‘rational’ cally correct); capable of explaining and guiding behaviour in the rather uncharted and neglected territory of the rational discovery of economic action

(logi-1 ‘Models of man’ and the ‘rationality divide’

The model of rational decision making employed in economic sciences typically includes the statement that the set of possible actions A and the set of possible

‘states of the world’ S (or the ‘contingencies’ under which actions generate their

consequences) are ‘known’ (the elements in the set are enumerable or a criterion

of belongingness to the set is given) An expected utility can be defined ‘over’

the combinations of a and s belonging to A and S respectively: by ordering the

values of outcomes according to preference and by assigning probabilities when their value is not known for certain The only rational learning method

10 Micro foundations

considered is the Bayesian updating of probabilities, if they can be assigned,

upon the observation of new information on a and s.

If analysed from an epistemological standpoint, those standard assumptions

on rationality in economics should be characterized as a set of ‘logically sound procedures’ for the calculative component of decision only (Shackle 1972)

If analysed from an epistemological standpoint, it may also be noticed,

however, that these assumptions allow two different interpretations – a tionalist and a realist version.

conven- Inconven- aconven- conventionalistconven- interpretation,conven- theyconven- canconven- beconven- readconven- asconven- follows:conven- let’sconven- assumeconven- that actors behave ‘as if ’ they have knowledge of the inputs and are calculating and choosing the option with the maximum expected value in a given set of options, and explore and test the predictions (Friedman 1953) The prediction, as known, is that, under those restrictive conditions approximating perfect competi-tion, if actors behave ‘as if ’ they were value maximizers, then equilibrium prices will be reached and resources allocated efficiently at their best use

In the most common interpretation, though, the conventionalist assumption that inputs are known and calculations are performed according to value maxi-mizing calculation procedures, has been transformed into the descriptive and realist assumption that actors do in fact possess a complete and infallible know-ledge of the possible states of the world and their possible actions Economists have apparently started to believe that the admission that knowledge is fallible and that actors do not have perfect foresight, would amount to abandoning the rational actor paradigm for accepting a ‘bounded rationality’ assumption: e.g

‘with rational agents contingencies are never unforeseen, they are at worst scribable’ (Tirole 1999); ‘If parties do not foresee even relatively obvious events,

inde-it would seem necessary to assume that they are boundedly rational’ (Hart and Moore 1999)

This realist version of the rational actor paradigm entails some cal problems that a conventionalist version does not

epistemologi- Ernstepistemologi- Nagelepistemologi- offersepistemologi- anepistemologi- interestingepistemologi- startingepistemologi- point.epistemologi- Heepistemologi- observedepistemologi- (Nagelepistemologi- 1963:epistemologi- 214) that a ‘trivial way in which any statement can be said to be unrealistic is that it can never give a complete description of all the infinite aspects of any real objects or situation’ This criticism applies to the realist version of economic rationality, according to which the rational actor actually has a complete know-ledge of the world, and actually foresees all possible contingencies The criticism

is consistent with and rooted in the tradition of the philosophy of knowledge and science, to which Nagel himself contributed According to that tradition (after Russell and Popper), in fact, any assertion or assumption about knowledge being

‘objective’, ‘complete’, ‘perfect’, or assertions like ‘there are no unforeseen

con-tingencies’, would be considered nạve and even illogical statements A core

argument, due to Popper, is that one can never prove, according to logically sound procedures, that a statement about reality is true or to have considered ‘all the possible contingencies’ because they are infinite not only in number but also

in kind Hence, one can provocatively say that the assumptions of classic economic rationality in their realist version are not rational, in the sense of

Micro foundations 11

ally defending the legitimacy of making ‘unrealistic assumptions’ in economic modelling, but interestingly concluded that this methodological approach can be defended precisely because there can be no ‘complete’ description of reality and

applying ‘logically sound procedures’ (Grandori 2010a) In fact, Nagel was actu-no ‘perfect foresight’ of the world!

The alternative notion of ‘bounded rationality’ (Simon 1955), has been defined in strong reference with the economic model intended in its realist version, and developed by relaxing some of its core assumptions about the knowledge of decision inputs In Simon’s words, in the ‘classical’ (economic) concept of rationality,

the organism must be able to attach definite payoffs (or at least a definite range

of payoffs) to each possible outcome This, of course involves also the ability

to specify the exact nature of the outcomes – there is no room in the scheme for ‘unanticipated consequences’ The payoffs must be completely ordered – it must always be possible to specify in a consistent way, that one outcome is better than, or as good as, or worse than any other And, if the certainty or probabilitic rules are employed, either the outcomes of particular alternatives must be known with certainty, or at least it must be possible to attach definite probabilities to outcomes My first empirical proposition is that there is a com-plete lack of evidence that, in actual human choice of any complexity, these computations can be, or are in fact, performed

(Simon 1955: 104)The alternative ‘satisficing’ model was then defined in terms of ‘partial’ and

‘sequential’ exploration of problem spaces, guided by ‘aspiration levels’ (ALs)

or ‘thresholds’ on the expected utility of the sought alternatives, and a choice rule stated in terms of ‘acceptability’ – an alternative is acceptable if its payoff is superior to the threshold

In this way, the main alternative to classic economic rationality was born as a

‘daughter of a minor godness’, a form of rationality ‘trying’ to approximate the classic economic template, but in almost all situations falling short of accom-plishing the task and ending up in a ‘reduced’, more ‘limited’ form of rationality,

of complete description and knowledge of the world may be raised from an stemic viewpoint, as argued above

epi-12 Micro foundations

On the other side, it seems that Simon’s criticism, based on the ‘empirical proposition’ that humans ‘are not able’ to perform all the calculations involved

in economic models was in a sense too weak and in a sense too strong It was too weak because it criticized the classic assumptions interpreted in their realist and descriptive version, with an empirical and descriptive argument: human behav-

iour, descriptively and empirically, is commonly ‘less’ knowledgeable and

cal-culative It did not criticize whether, normatively and logically, the way of being knowledgeable and calculative of economic classic rationality was actually ‘the’ model toward which to ‘tend’ in our intended rationality The acceptance of the equation between ‘complete knowledge’ – or ‘omniscience’ – and value maxi-mizing rationality, present in the realist version of the rational actor model, pro-duced too strong a criticism of value maximizing as an empirically inapplicable decision process or strategy; a behaviour for which there is ‘no evidence’ The observation I advance here and develop in the following paragraphs is that, while there is no evidence (and actually no logical foundation) for complete knowledge, there is plenty of evidence (and strong logical foundations) for the applicability of value maximizing strategies to specified classes of structured and stylized problems (Savage 1954) It can be noticed here that Simon himself pro-vided an outstanding example of the applicability of value maximizing reasoning with incomplete knowledge, in his analysis of the employment relationship (1951) The model reconstructs the authority relation as a Pareto-efficient con-tract among two actors maximizing their respective utility, which is specified as

a function of the benefit or cost of performing the task x, and of the wage w to be paid or gained: for the employer (e), max U e = U e (x) – U e (w); for the labour pro- vider (l) max U l = U l (x) + U l (w) The attractiveness of a contract giving the

employer the right to choose the task is conditional to the circumstance that there

is uncertainty about which task will be better (contingencies cannot be foreseen), and the utility of having option rights is high for the employer and low for the worker

Hence, a utility maximizing logic does not imply or require perfect knowledge

and complete foresight Actually, an important contribution by Simon himself has been to demonstrate that the authority relation (as well as other possible decision procedures, such as codetermination, or worker self-determination; cf Simon 1951:

304) can be understood as Pareto-efficient solutions among utility maximizing actors, precisely under conditions of unforeseeable contingencies.

Both the complete knowledge rationality and the bounded rationality approaches can be criticized from an epistemological standpoint also in another respect Both have neglected, albeit in different ways, the problem of how eco-nomic decision makers may construct, in a valid and reliable way, the knowl-edge (voluminous or not) on which any decision is based In other words, there

has been little concern, in either tradition, with rational knowledge construction:

in the economic approaches because knowledge is assumed to be there, in the behavioural tradition because the concern with the ‘common patterns of think-ing’ has crowded out the concern with the ‘best patterns of thinking’ humans are capable of More practically, questions as simple and important as the following

Micro foundations 13

cedures for discovering new products and services? Is there any method in

are seldom posed: What are the better and worse, effective and ineffective pro-having new good entrepreneurial ideas? How can economic actors improve their

exploratory decision processes? Paradoxically, neither of the two major digms about rationality in economic decision making is able to respond to these core and simple questions, that any decision maker daily faces and are so central

para-in the currently much emphasized ‘knowledge intensive’ economy

The two rival approaches to rationality are therefore both suffering some problems in their current formulations They are therefore revisited and reinter-preted here in a new light, indicating how (1) they can be seen as alternative decision behaviours or ‘strategies’ rather than rival assumptions on human behaviour in general; (2) both types of decision strategies are ‘possible’, i.e empirically feasible; and (3) the conditions under which they are applicable and are comparatively superior can be specified That reinterpretation paves the way

to see how and why the two classical forms of rationality are not the only ble ones In particular, it will be argued that these two basic models of decision making can be complemented by a third model specifying how the inputs enter-ing decision processes – judgements on alternatives, utility, consequences, prob-lems themselves – can be constructed in a logically sound way in the first place

possi-2 Savage and Simon revisited: how both ‘maximizing’ and

‘satisficing’ simplify problems

The classic economic rational actor model can be (re)formulated in a still further way, that differs from the conventionalist, ‘as if ’, ‘Friedmanian’ mode; as well

dations of statistical decision theory (Savage 1954) is useful to see how those foundations are actually compatible with, and even suggestive of, a notion of utility maximizing rationality different from both versions which became diffused

Savage’s criterion for constructing problem models to which a rational mizing approach can be applied can be interpreted as stating almost the opposite with respect to the ‘omniscience-like’ criterion of including everything (eventu-ally taking into account the costs of search) Rather, he emphasized ‘the practical necessity of confining attention to, or isolating, relatively simple situations in almost all applications of the theory of decision developed in this book’ (1954: 83) Savage consequently also posed the further question of ‘what constitutes a satisfactorily isolated decision situation’, of when a problem model is satisfacto-rily constructed so that the approach can be applied He responds: a problem model is satisfactory (acceptable) if the preference judgements over the alterna-tives and consequences within the ‘small world’ of the problem model do not change if transferred into the real ‘grand world’ of which they are a ‘partition’ (Savage 1954: chapter 5.5) Then, we can observe that, first, utility maximizing procedures are conceived as ‘applicable’ to ‘simplified problems’; and second, that the acceptance of a problem model is based on an ‘epistemic’ criterion

maxi-14 Micro foundations

rather than an instrumental criterion: the predictive validity and representational stability of the model with respect to reality (Grandori 2010a)

Another quite neglected, and quite interesting epistemic feature of Savage’s foundations is to make it clear that a maximizing logic is compatible with ordinal preference orderings He clearly distinguishes ‘The simple ordering of acts with respect to preference’ (Savage 1954: chapter 2.6) involving only an ordinal judgement over the utility of alternatives, from the

arithmetization of comparison among acts achieved by attaching a number

U( f ) to each consequence f, g, so that f ≤ g if and only if the expected value of U( f ) is numerically less than or equal to that of U(g).

(Savage 1954: chapter 6: 69)Hence a requisite that actors should be able to attach a definite payoff or cardinal utility to each outcome is not actually a requisite for a maximizing strategy: to define at least an order of preference is a requirement

Utility maximizing can be defined as a set of logically sound procedures on how to compare defined alternatives according to preference (not necessarily according to their ‘value’: we might be able to say the we prefer partner A to partner B, without being able to size their value and without judging them in terms of value) In this formulation, there is nothing unrealistic about a maximiz-ing strategy It is an entirely possible strategy It is also compatible with situa-tions in which the ‘value’ or payoff of single acts cannot be defined, and utility

is not operationalized in a numbered cardinal function, but only in a comparative ordinal judgement Its fundamental epistemic requisite is not any form of impos-sible ‘omniscience’ but, rather, that problems are structurable and ‘isolable’ In other words, utility maximizing is a set of instrumentally rational procedures that are also epistemically rational if a good simple problem model can be constructed

If so defined, there is plenty of evidence, rather than none, that actual human beings, even in choice situations of considerable complexity, manage to select relevant information and to structure the problem so that it can be dealt with using a maximizing approach, reaching remarkable and reliable results (making airplanes fly, calculating insurance premia or optimal production and transporta-tion schemes), even though no claim of objectively complete knowledge and absolute optimality can be made about those solutions The value maximizing models typically crafted by economists themselves too, in fact, are usually very selective and stylized models, considering very few alternatives and very few possible states of the world in order to be able to make value maximizing calcu-lations; rather than models that take into account ‘all possible’ dimensions of action and contingencies

A corollary of this reinterpretation of rational choice is that it is a set of methods If there is no infallible knowledge, there is no ‘substantively’ rational choice, i.e ability to guarantee that an objectively optimal solution is determined

In other words, we should abandon any characterization of economic rationality

Micro foundations 15

doxically, quite an excessive award granted (by Simon himself; 1976) to eco-nomic rationality In an epistemic perspective, all forms of rationality are necessarily procedural, and all decision models can and should be described as a set of procedures

as ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- Theas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- logicas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- ofas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- ‘satisficing’,as ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- contrastedas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- byas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- Simonas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- withas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- ‘optimizing’as ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- andas ‘substantive’ and of other forms of rationality as ‘procedural’ That was, para- ‘maximiz-ing’,1 also requires that matters can be simplified and structured into ‘simple prob-lems’, albeit simple in a different way than problems that are solvable in a maximizing way ‘Satisficing’ requires more knowledge about the self and the problem at hand than knowledge about the environment and the possible alterna-tives Simon’s famous example of ‘searching for a needle in a haystack, sharp enough to saw with’ captures the point: the decision maker should have a clear idea

The logic of ‘satisficing’, contrasted by Simon with ‘optimizing’ and ‘maximiz-of what the problem is and of what features an adequate alternative should have – what sewing is, why a needle is what is necessary, how sharp a needle should be Even if problems are more ‘complex’, the complexity should stem only from problem size – the number of features and parameters to analyse, the variety of moves that may be checked for acceptability – as the other example most often used by Simon, the game of chess, illustrates Satisficing is a decision procedure based on an acceptability judgement based on an ordinal comparison between any act and a set of acceptability thresholds or ALs ALs can be defined as represented

by available/known alternatives (as the best known alternative to a partner or ment)2; or by desired/required levels of outcomes on a set of parameters

agree- Theagree- questionagree- weagree- areagree- interestedagree- hereagree- isagree- underagree- whatagree- conditionsagree- andagree- toagree- whatagree- extent those procedures can be said to be instrumentally and epistemically rational To that purpose, it is useful to state them more formally This is not only conducive to comparing them more clearly with utility maximizing, but is also conducive to determining that there are variants in satisficing logics, as much as in maximizing logics (cardinal and ordinal, probabilistic and deterministic)

The simplest formulation is that an alternative a i belonging to a set A,

‘known’ or ‘supposed’ to produce certain desired effects e y in E, under a range of

states of the world s j , is acceptable if a i ≥ al zmizing, alternatives are judged on their own features and attributes as predictors

As in the ordinal version of maxi-of some consequence of interest Point-wise utility judgements are not required

A more demanding cardinal version of satisficing, involving those ments, can also be defined though: an alternative is acceptable if the utility of its effects or ‘payoff ’ is higher than the utility or payoff corresponding to the

judge-acceptability threshold, i.e if U(e ai ) > U(e alz)

Finally, an expected value version of satisficing is also conceivable: an alter-native a i is acceptable if its expected utility is greater than that of the ity threshold: EU[(e ai ) · p(e ai )] > EU[(e alz ) · p(e alz)]

acceptabil- Hence,acceptabil- satisficingacceptabil- rulesacceptabil- areacceptabil- actuallyacceptabil- aacceptabil- familyacceptabil- ofacceptabil- procedures.acceptabil- Theyacceptabil- areacceptabil- notacceptabil- allacceptabil- the same in terms of information requirements The simplest rules are of the type

of ‘elimination by aspects’, the more complex of the type of acceptable expected utility judgements

16 Micro foundations

The epistemic soundness of each rule depends heavily on the reliability of knowledge of the correlations between the ‘aspects’ used for accepting/rejecting alternatives (features of the alternatives) and their effects If the problem is hiring an interpreter for an event, and it is certain that English speaking is a nec-essary cause of any level of performance in the task above zero, it is rational to reject alternatives not presenting that feature The rationality involved though, is not only instrumental (what level of performance will be guaranteed) but also epistemic (is it true that English speaking is necessary?)

The judgement becomes more complicated, if expected results and utility terms are included; but the logic of acceptability-based decision making remains

a logic of the acceptance/rejection of hypotheses or propositions on the results (effects) brought about by some alternative (cause), and on their ordinal superi-ority to some base rate level

Hence, core questions for assessing whether and when to proceed in this way

is rational, are: where do AL come from (how thresholds are defined)? should research be stopped as soon as an acceptable solution is found? These questions

in fact figured prominently in the original satisficing model The answers, as is well known, consisted in specifying some complementary heuristics regulating the learning of A and AL:

• ALs are sets of movable targets Direct experience was hypothesized to drive their adjustment in the original model: ‘if it is difficult to find, the AL

is reduced; if it is easy, it is raised’

• New sets of alternatives A1,2, n can also be generated, if the hypothesis of finding what was searched for is not supported by experience The heuristics for generating new sets, considered in the original model, were experience-driven: to consider actions that produced the sought effect in comparable circumstances (experimented in the past, by the actor or by comparable others) (March and Simon 1958; Cyert and March 1963)

• The decision maker will stop investing in search as soon as an acceptable alternative is found with respect to the movable AL

These heuristics are important as they are concerned with learning, hence with knowledge construction However, the relevant question in our perspective is: under what conditions can these experience based-heuristics, in particular the heuristics of lowering/raising AL as a function of the ‘frequency of findings’ or

‘time to finding’ – be defended as rational learning rules?

Learning rules based on ‘direct’ or ‘vicarious’ experience are defendable from an epistemological standpoint only in particular conditions (Bandura 1986; Grandori 1984):

• a problem should be defined in terms of a ‘performance gap’, whereby the performance parameters are assumed to be known;

• relevant experience should be available, whereby the problem should not be too new or unique;

Micro foundations 17

• rable to the focal actor, whereby the experimental conditions should be rather stable

the past should be comparable to the present, or the ‘relevant others’ compa-cal ‘difficulty’ of finding a solution is even more questionable as the best heuris-tics one can use: it may imply a ‘sour grapes’ fallacy, and better heuristics for testing hypotheses on preferences are conceivable and actually used (Section 5.3) Simon himself (1977) acknowledged, in his writings on methods and dis-covery that ‘persistence’ is precious in research, if not in ‘search’ Adjusting aspirations to what can be easily found is justifiable to the extent that it is possi-ble to formulate the ordinal judgement that the costs of search and effort (what-ever they might be) are in any case high in comparison to the expected difference among the alternatives that might come up (whatever they might be)

Learning by using the heuristics of lowering the AL as a function of the empiri- InLearning by using the heuristics of lowering the AL as a function of the empiri- sum,Learning by using the heuristics of lowering the AL as a function of the empiri- aLearning by using the heuristics of lowering the AL as a function of the empiri- ‘satisficing’Learning by using the heuristics of lowering the AL as a function of the empiri- strategyLearning by using the heuristics of lowering the AL as a function of the empiri- canLearning by using the heuristics of lowering the AL as a function of the empiri- beLearning by using the heuristics of lowering the AL as a function of the empiri- definedLearning by using the heuristics of lowering the AL as a function of the empiri- asLearning by using the heuristics of lowering the AL as a function of the empiri- aLearning by using the heuristics of lowering the AL as a function of the empiri- setLearning by using the heuristics of lowering the AL as a function of the empiri- ofLearning by using the heuristics of lowering the AL as a function of the empiri- logicallyLearning by using the heuristics of lowering the AL as a function of the empiri- soundLearning by using the heuristics of lowering the AL as a function of the empiri- pro-cedures on how to select acceptable alternatives while minimizing search efforts

In sum, a ‘satisficing’ strategy can be defined as a set of logically sound pro-in given problems and stable environments Under those conditions, it is not only

cally and instrumentally rational strategy Actually if the term ‘maximization/minimization’ is interpreted in the procedural and ordinal judgement version specified above, ‘satisficing’ is equivalent to a decision rule of minimizing search costs/efforts subject to an acceptability constraint.3

a possible, or empirically frequent behaviour; it can be considered an epistemi- Ina possible, or empirically frequent behaviour; it can be considered an epistemi- addition,a possible, or empirically frequent behaviour; it can be considered an epistemi- thea possible, or empirically frequent behaviour; it can be considered an epistemi- ‘behaviourala possible, or empirically frequent behaviour; it can be considered an epistemi- modela possible, or empirically frequent behaviour; it can be considered an epistemi- ofa possible, or empirically frequent behaviour; it can be considered an epistemi- rationala possible, or empirically frequent behaviour; it can be considered an epistemi- choice’a possible, or empirically frequent behaviour; it can be considered an epistemi- specifieda possible, or empirically frequent behaviour; it can be considered an epistemi- bya possible, or empirically frequent behaviour; it can be considered an epistemi- Simona possible, or empirically frequent behaviour; it can be considered an epistemi- under the label of ‘satisficing’ is a particular decision model, not a general model

of intendedly rational decision-cum-search behaviour Simon himself declared

so, while presenting it in his Models of Man, inviting to develop other ‘bounded

oped and framed as bounded rationality models For example, ‘incremental’ (Lindblom 1959), ‘cybernetic’ (Steinbruner 1974) and ‘random trial and error’

rationality models’ Those other models have been in fact subsequently devel-(Cohen et al 1972) models of decision making have been identified.

These models, at least in the way they have been formulated and presented, emphasize even more the ‘common patterns of thinking’ nature of these proc-esses rather than asking whether they can be justified as best patterns of thinking under certain circumstances Nevertheless, as in the case of satisficing, they can

be reformulated and re-evaluated as epistemically and instrumentally rational strategies under particular conditions This is the topic of the next paragraph

Before turning to that, some general comments on the two main classes of

decision models – the economic rational actor models and behavioural bounded rationality models – in the account given here Given that this account is some-what different from the standard account, let us also use somewhat different names and notations

gies: a maximization logic and an acceptability logic They are both applicable

We do not have two ‘models of man’, we have two decision logics or strate-in appropriate conditions These conditions should be such that problems can be formulated in a ‘simplified’ and ‘stylized’ way: in a maximizing logic, the world

18 Micro foundations

is simplified into a definite set of rankable alternatives and the highest ranked solutions4 are candidate for choice; in an acceptability logic the world is simpli-fied in a definite utility threshold used for searching and evaluating alternatives, and those above the threshold are candidate for choice Both these inputs to deci-sion making may be necessary in many situations: for example where potential losses are implied (whereby notions of affordable losses or acceptable risk are

necessary to represent actors’ utility well) (Simon 1955; Fischhoff et al 1984),

or in multiparty decision making where alternative deals provide a ‘best tive to a negotiated agreement’ threshold (Raiffa 1982) Hence, the two strate-gies are combinable and actually very often are combined

Both maximization-based and acceptability-based decision making strategies employ problem simplification heuristics: procedures for stating problems in the

way that is tractable with the strategy In the case of acceptability-based decision making, this feature is highly declared and commonly recognized: problems are defined as performance gaps, a lot of information is neglected, and as soon as one alternative bringing about the closure of the performance gap is found, the process is over In the maximization case, the aura of omniscience granted to it

by both its supporters and its opponents, obscures the relevant ‘neglect’ of mation implied by structuring a problem so that the strategy is applicable Prob-lems are simplified by establishing utility scales allowing the comparison among qualitatively different features of alternatives (thereby often neglecting a lot of qualitative differences); by defining preferences only ‘over’ available alterna-tives (thereby neglecting the analysis of underlying interests independently from the actions through which they may be realized); by needing to translate, for cal-culation purposes, uncertainty in probability judgements – which however require quite particular conditions for being formulated according to logically sound procedures

infor- Theinfor- simplificationinfor- ofinfor- problemsinfor- soinfor- thatinfor- theyinfor- areinfor- tractableinfor- withinfor- bothinfor- maximiza-tion and acceptability-based strategies implies the use of rational heuristics, and the ‘neglect’ of some information The type of uncertainty that can be mastered through those heuristics is however of limited intensity

The simplification of problems so that they are tractable with both maximiza- InThe simplification of problems so that they are tractable with both maximiza- fact,The simplification of problems so that they are tractable with both maximiza- theThe simplification of problems so that they are tractable with both maximiza- basicThe simplification of problems so that they are tractable with both maximiza- sourcesThe simplification of problems so that they are tractable with both maximiza- orThe simplification of problems so that they are tractable with both maximiza- typesThe simplification of problems so that they are tractable with both maximiza- ofThe simplification of problems so that they are tractable with both maximiza- uncertaintyThe simplification of problems so that they are tractable with both maximiza- withThe simplification of problems so that they are tractable with both maximiza- respectThe simplification of problems so that they are tractable with both maximiza- toThe simplification of problems so that they are tractable with both maximiza- whichThe simplification of problems so that they are tractable with both maximiza- bothThe simplification of problems so that they are tractable with both maximiza- the two main decision models have been elaborated have been the ‘difficulty of having a whole picture’ and the ‘variability of conditions’ (Hayek 1945; Simon 1955) Both the number of variables to be taken into account (‘information com-plexity’) and their variability (‘aleatory uncertainty’) enter in the type of uncer-tainty considered Uncertainty though can be stronger than that, involving the partiality and fallibility of knowledge In classic economic thought, the notion of

‘Knightian’ uncertainty is referred to In more recent scientific thought, both in economics and management sciences and in engineering and complexity sci-ences, this ‘stronger’ type of uncertainty has been called ‘epistemic uncertainty’,

as contrasted with ‘aleatory uncertainty’ (Oberkampf et al 2001) and with

‘computational complexity’ (Grandori 2010a); or ‘structural uncertainty’, as contrasted with ‘parametric uncertainty’ (Langlois 1986) All these elaborations concur in highlighting that for mastering that type of uncertainty the classic

Micro foundations 19

maximizing and satisficing logics are not sufficient Identifying rational decision procedures for that type of epistemic or Knightian uncertainty is the concern put centre stage in this micro-foundational part of the book

3 Endogenizing assumptions: contingent rationality

Thus formulated, maximizing and acceptability decision logics appear much closer than in usual accounts, in which they are supposed to be divided by ‘rival assumptions’ on nothing less than human nature In fact, there is no need for

‘rival assumptions’ on human nature for the account of decision procedures that has been given here Both strategies employ problem definition heuristics, and both strategies employ search, choice and learning procedures that may be quali-fied as logically correct and appropriate to the tasks or problems – and in that sense ‘rational’ – under certain conditions In this account, it would not be para-doxical to say that an ‘acceptability’ strategy may be ‘superior’ to a ‘maximiz-ing’ strategy, under some conditions

There have been works explicitly framing decision strategies as alternative behaviours among which an actor has to choose, that have contributed in speci-fying those conditions

In these ‘contingency models’ of decision making, in turn, different logics for strategy selection can be identified

• In economics it has been repeatedly observed that if problems are well defined and finite – even if complex – then heuristics such as satisficing and incrementalism can be justified as optimally imperfect decisions if decision

process costs are considered (Cyert et al 1978; Baumol and Quandt 1964;

Baumol 2004; Arrow 2004) In these approaches the trade-off between effort and accuracy is calculated and the logic is to choose the decision strat-egy that maximizes expected net returns Given that problem complexity generates higher decision process costs, the optimal investment in analysis and information processing is inversely related to problem complexity

• chologically informed and more empirically-based approach The hypotheses and empirical results, obtained mainly through laboratory studies, concluded

Payne (1976) and Beach and Mitchell (1978) opened the road to a more psy-that task complexity with respect to actor resources and knowledge has a major

influence on strategy selection Task complexity (as number of alternatives and information to be considered) was found a strong predictor of a shift from compensatory to non-compensatory decision procedure in Payne’s experi-ments Beach and Mitchell operationalized task complexity variables such as task novelty/unfamiliarity for the actor, ambiguity of goals, number of alterna-tives and amount of information to be processed, and instability; and modelled them as independent variables together with other situational factors such as the irreversibility of choices, the significance of outcomes, the need for accountability, and the knowledge, time and resources available to the deci-sion making process They posited and found that these factors are able to

20 Micro foundations

predict a shift from more analytic and compensatory decision procedures (such

as utility maximization), to less analytic and non-compensatory strategies (such as ‘satisficing’ and elimination by aspects), to non analytic strategies such as repeating previous actions or flipping a coin

gies is predominantly seen as deriving from an implicit trade-off between

The underlying justification of the contingent shift among decision strate-logical version of the economic approach: selecting the strategy with superior net benefits Gigerenzer and Sturm (2012) in fact, recently pointed out that strategy selection may be governed also by more experiential and ‘ecologi-cal’ processes, applied to the higher level of the selection of heuristics rather

‘effort’ and ‘accuracy’ In this version, then, the justification seems a psycho-than to the selection of actions or ‘moves’ Pattern-recognition is one of

ture are matched not only to a move (as in the classic pattern-recognition

‘didn’t have a model to follow, as there was no such thing as advertising on the Internet at that time’ At the outset the model or ‘cognitive representa-tion’ of themselves and their output was that of a ‘technology’ company and product Later, following experimentation in marketing and advertising, that model was enriched and extended to a model of a ‘techno-media’ product and service

Finally, environmental selection can also work at the level of the decision

processes followed by an actor and not only at the level of the action chosen

A pertinent example can be the classic Axelrod’s study of the superiority of tit-for-tat strategies with respect to any other (which may therefore be driven out of the game in those conditions) in prisoner’s dilemma game structures

• A different exercise in the comparative assessment of the decision strategies was performed in Grandori (1984) That analysis considered the portfolio of strategies that we have so far identified – e.g ‘maximizing’, ‘satisficing’,

‘incrementalism’, ‘reinforcement-based’ – as dependent variables, similarly

to previous contingent decision making studies The reasons for the relative superiority of each strategy were then derived from an ‘epistemic failures’ framework rather than from a cost–benefit criterion As there are ‘market failures’, so there are decision strategy failures if information requirements for applying the strategy are not present Two main independent variables were considered: the state of knowledge and the structure of interests The less knowledge relevant to the problem at hand is available/acquirable, and the higher the conflict among relevant interests, the narrower the set of

Micro foundations 21

applicable decision strategies becomes, up to random strategies in the more severe conditions For example, value maximizing may fail for unsolvable conflict situations (e.g prisoner’s dilemma games); or because alternatives are incomparable and not rankable; or because it is not clear what their utility is Acceptability-based strategies may also fail, especially in very new fields, where there is no sound basis for defining ALs Incremental strategies may fail when the problem space is discontinuous, so that either action cannot be reduced to incremental steps, or the effects of incremental steps cannot be reliably assumed to be incremental The principle of rein-

forcement and ex post feedback upon experience fails in any situation

fea-turing variability of experimental conditions As a consequence, strategies that use information more selectively, do not employ probability judge-ments, or avoid certain types of analyses as trade-offs among outcome parameters, can be ranked as epistemically superior to strategies which do perform those operations For example, if a problem involves multiple and incomparable dimensions, modelling the problem in a non-compensatory way – i.e so to speak, without performing inter-criterion comparisons of

utility – may be more rather than less accurate (Grandori 1984, 2010a) In

recent studies on ‘complex’ decision making as early stage investment in new entrepreneurial projects (Grandori and Cholakova 2013), evidence has been provided that experienced and successful investors do proceed without attempting to force comparisons or trade-offs between problem dimensions that are as incomparable as the ‘potential of technology’, the ‘quality of entrepreneurial teams’, the ‘attractiveness and growth potential’ of a sector, the expected size of financial returns

Some observations on these three approaches to ‘contingent rationality’ are useful at this stage to the purpose of the present analysis First, it is worthwhile noticing (given how ingrained the universalistic idea that more information is better than less) that all the three approaches, being ‘contingent’, admit the pos-sibility that ‘less’ information processing can be ‘more’ appropriate than more

It is also noteworthy that they generate similar predictions, in the domain that is

common to all approaches, that is the solution of given problems: in that domain,

the more complex the problem (hence the higher the process costs), the less complex the superior decision strategy is (keeping the importance/size of conse-quences constant)

Conversely, the logic of (meta) choice among decision behaviour is different across the three approaches In the classic ‘economic’ approach, if a problem is closed and structured, the choice among defined decision strategies can be based

on value maximizing and cost minimizing reasoning In the second approach, rooted in cognitive psychology, decision strategies are learned by experience over series of similar problems, and can therefore be defined ‘experiential’ The third approach to strategy selection is ‘epistemic’, as it relies on the possibility

of constructing a valid problem model characterized by the type of uncertainty and interest structure that can be treated with a decision strategy

22 Micro foundations

Hence, we find the same three fundamental types of decision strategies that were possible in the selection of actions, also at the meta-level of the selection of heuristics or decision procedures This may be seen as a sign of strength, rather than of weakness, of the framework In fact, more than a case in ‘infinite regress’, it may be seen as a case in the ‘fractal nature’ of complex phenomena After all, we might well start accepting that problems are never ‘really’ closed and ‘infinite’ levels are not impossible to conceive In the case of decision rules,

if by shifting to higher order choices the logic of decision became indefinable, or contrasting with that at lower levels, than inconsistency objections would be pos-sible By contrast, if the overall complex picture of decision making can be gen-erated by applying the very same set of criteria at any level, it only means that the phenomenon is governed by the same ‘recursive’ and ‘fractal’ structure exhibited by many complex phenomena

Then, it may be said that the metaphor of the alternative ‘models of man’ may

be dangerous if intended in a ‘realistic’, ‘rival assumptions on human nature’ interpretation, as it obscures the capacity of any one actor of adopting different, and contingently rational, decision strategies

The review conducted in this section also showed that the portfolio of sion models has been enlarged, becoming a ‘continuum’ rather than a dichot-omy The type of uncertainty considered also extended to include epistemic aspects, such as the ‘lack of knowledge’ on cause–effect relations, or on goals themselves However, in the face of those stronger forms of uncertainty, the models so far developed have prospected responses going in the direction of lowering even further the requirements in terms of knowledge inputs and ration-ality of procedures with respect to the original ‘satisficing’ model What has not been done is, so to speak, to go in the opposite direction: investing in research and constructing valid knowledge, rather than reducing thought in the face of uncertainty, where problems should be defined and can be changed, and experi-ence is a faulty guide

deci- Thedeci- taskdeci- addresseddeci- nextdeci- isdeci- thereforedeci- todeci- reconstructdeci- adeci- precisedeci- profiledeci- fordeci- thisdeci- usually neglected further class of rational discovery-based decision strategies – the need for which has been detected in various places in the above discussion

chological approach to heuristics is needed

In order to do so, a premise on the difference between an epistemic and a psy-4 The ‘psychology’ versus the ‘logic’ of judgement and

of economic models, and have opened the entire new field of behavioural econom-Micro foundations 23

problems to subjects, the common and systematic use of a variety of ‘heuristics’ has been detected, including: considering ‘available’ rather than ‘relevant’ infor-mation; judging probabilities by case ‘representativeness’ and neglecting base rate distributions; sampling on the dependent variable; looking only for positive evidence; fundamental attribution errors The term heuristic has therefore been associated with that of ‘biases’, as in fact, that kind of heuristics does include

‘logically unsound procedures’ and generates systematic ‘biases’ – such as local search, over-confidence, self-confirmation, escalation of commitment and the illusion of control

Therefore, in most cases, these heuristics cannot be qualified as ‘intendedly rational’ procedures, in the Simonian sense They are even weaker forms of thought, ‘shortcuts’ taken for effort saving reasons, but demonstrably dangerous

as they include logical mistakes and optical illusions

These ‘heuristics’ have been detected in conditions where a ‘right’ statistical procedure is available for actually detecting the ‘deviation’ or bias in a precise way This is a useful and rigorous method for detecting a phenomenon It has a long history in behavioural and social science: for example in group decision making research we can demonstrate that on average the quality of group decisions is supe-rior to individual solutions in experiments where a technically optimal or superior solution can be calculated or is scientifically known This method does not imply that the algorithm for calculating the right solution is always available in reality, nor does it imply that one can or should revert to classic statistical decision theory for correcting them I am afraid that, instead, that interpretation has been quite diffuse, generating, on one side, a reinforcement of the belief that classic economic and statistical rationality is the cure for all evils; and, on the other side, a generator

of attempts to justify those heuristics with the argument that ‘fast and frugal thought’ can ‘outsmart’ or be more efficient than value maximizing or statistical

decision processes, that are seen as more cumbersome (Gigerenzer et al 1999)

‘rational’ or ‘non biasing’?

24 Micro foundations

The underlying problem is that in psychology (as well as in the combinations between psychology and economics) the dominant research question has been: how do people actually behave? The implication has been either to ‘correct’ behaviours by means of statistical and economic decision theory (Tversky and Kahneman 1974) or to adjust economic models to account for observed behav-iours by inserting ‘behavioural parameters’ into utility functions (Tversky and Kahneman 1981; Akerlof and Kranton 2000; Thaler 1991) The question that is typically not posed in psychology-based research is: how might people behave

tions has a perfect parallel in the debate between the psychology and the logic of science The ‘logic of scientific discovery’ (Popper 1935, Lakatos 1970a; Simon 1977) is an empirically-based but ‘rationally reconstructed’ and prescriptively filtered notion: the relevant research question is not ‘how do scientists actually think?’ but ‘which are the best patterns of thinking that can be found in scientific thinking?’ (Kiss 2006) The logic of science is distinguished from the sociology and the psychology of science precisely on that ground (Lakatos and Musgrave 1970)

to the best of their cognitive possibilities? The difference between the two ques- Into the best of their cognitive possibilities? The difference between the two ques- ato the best of their cognitive possibilities? The difference between the two ques- similar,to the best of their cognitive possibilities? The difference between the two ques- andto the best of their cognitive possibilities? The difference between the two ques- brilliantlyto the best of their cognitive possibilities? The difference between the two ques- polemic,to the best of their cognitive possibilities? The difference between the two ques- vein,to the best of their cognitive possibilities? The difference between the two ques- theto the best of their cognitive possibilities? The difference between the two ques- psychologicallyto the best of their cognitive possibilities? The difference between the two ques- informedto the best of their cognitive possibilities? The difference between the two ques- logi-cian Simon (1977) in fact criticized the rationalist logician Popper (1935) for a psychological view of discovery:

In a similar, and brilliantly polemic, vein, the psychologically informed logi-It is unusual for an author, less than one-tenth of the way through his work, to disclaim the existence of the subject matter that the title of his treatise

announces Yet this is exactly what Karl Popper does in his classic, the logic

of scientific

discovery, announcing in no uncertain terms on p 31 that scien-tific discovery has no logic The disclaimer is so remarkable that it deserves to

be quoted at length: ‘The question how it happens that a new idea occurs to a man – whether a musical theme or a scientific theory – may be of great interest

to empirical psychology, but it is irrelevant to the logical analysis of scientific knowledge Accordingly I shall distinguish sharply between the process of conceiving a new idea and the methods and results of examining it logically’ This mystical view toward discovery, while shared by most of the world, has not gone without challenge Pierce coined the term ‘retroduction’5 as a label for the systematic process leading to discovery; while Hanson, revived that term and gave us a careful account of the retroductive path that led Kepler

to the elliptical orbits of the planets It is instructive to confront Popper’s view with Hanson’s: ‘The initial suggestion of a hypothesis is very often a reasona-ble affair It is not so often affected by intuition, insight, or other impondera-bles Disciples of the H-D (hypothetico-deductive) account often dismiss the dawning of an hypothesis as being of psychological interest only, not of logic They are wrong If establishing an hypothesis through its predictions has

a logic, so has the conceiving of an hypothesis (Hanson 1958: 71)’.6 ‘It is the aim of this paper to explain in what sense one can speak of a “logic of discovery” or “logic of retroduction” ’

(Simon 1977: 326–327)

Micro foundations 25

The methodological works by Simon then provide a different, and quite neglected, side or interpretation of ‘intendedly rational’ behaviour: including not necessarily only behaviour that is uncertainty-avoiding, saving on cognitive and search effort, and problem space-reducing; but also behaviour that is uncertainty-modelling, investing in research effort, and problem-expanding; a behaviour that has the same nature of scientific research behaviour (Hatchuel 2001; Grandori 2001a) Actually Simon (1977: 286) considered ‘scientific discovery as a form

ing the converse framing of considering ‘problem solving as a form of scientific discovery’ (Grandori 1984)

of problem solving’ In the next section, the journey enters this territory, follow-5 The logic of economic discovery: an epistemic decision

model

In the repertory of decision making models there is a conspicuous hole, then: the

‘missing model’ is that of a ‘heuristic’ and yet ‘rational’ behaviour A model of behaviour interested in distinguishing true from false statements (‘epistemic rationality’) in addition to being interested in distinguishing positive from nega-tive consequences (‘instrumental rationality’) (Foley 1987) and interested in doing both things correctly, whereby both behaviours can be qualified as

‘rational’ A model of behaviour capable of describing and prescribing how a rational actor can address epistemic or ‘Knightian’ uncertainty

A point of departure for such an endeavour can be precisely to notice how different the meaning of the term ‘heuristic’ is in epistemology from what it has taken to mean in cognitive psychology: ‘heuristics is that part of a science that has as an objective the discovery of facts or truths’, the ‘very method of research’ (Lakatos 1970a: appendix II)

In Simon’s methodological writings on scientific method (Simon 1977), the notion of heuristics is closer to this philosophy of science notion than to the notion of a ‘shortcut’ or a ‘rule of thumb’ often found in his economic and organizational writings Heuristics, in Simon’s ‘models of science’ writings, are conceived as methods of discovery that can be meaningfully compared in terms

of their effectiveness and efficiency For example, he reported the results of a computer simulation on the task of detecting a law in a series of letters in which patterns or regularities can be observed (e.g ABMCDMEFMGHM ) The results showed that the ‘heuristics’ of ‘pattern recognition’, and of the ‘abduc-tion’ of hypotheses about the laws that might have regulated the recognized pattern, was more efficient in discovering the law than the heuristics of scanning

‘all conceivable laws’ in a blind trial and error, sequential way (Simon 1977: 332) Interestingly the latter heuristics – the ‘British Museum Algorithm’ – was

assessed as inferior in spite of processing more information than the former.

We can enrich the argument by observing that this finding has parallels in real world decision making processes For example, the shift from ‘random screen-ing’ heuristics to ‘rational drug discovery’ heuristics in the pharmaceutical industry, was described as follows:

26 Micro foundations

In earlier phases (1950s–1990s) the prevailing approach was ‘random

screening’: natural and chemically derived compounds randomly screened

in test tube experiments and laboratory animals for potential therapeutic activity Pharmaceutical companies maintained enormous ‘libraries’ of chemical compounds added to their collections by searching for new com-pounds in places such as swamps, streams, and soil samples Thousands if not tens of thousands, of compounds might be subjected to multiple screens before researchers honed in a promising substance ( .)

The more traditional path of starting with a problem and searching for a solution was also common, especially in commissioned research, e.g ‘Find

Proceeding in this direction, a portfolio of ‘rational heuristics’ can be recon-‘biasing’ methods becomes meaningful and clear These heuristics are to be intended, empirically, as the ‘best patterns of thinking’, and, theoretically, as the

temic’ uncertainty can be mastered The approach is different from, and somehow opposite to, reacting to the complexity of a (given) problem by reduc-ing the complexity of the strategy; it involves adopting a more complex strategy, including problem modelling heuristics as ingredients

‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- A‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- variety‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- of‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- heuristics‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- of‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- that‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- type‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- have‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- actually‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- been‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- identified,‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- both‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- in‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- field‘logically sound procedures’ through which conditions of ‘Knightian’ or ‘epis- and laboratory studies on discovery-intensive decision making They are exposed here showing that they can be used to compose a new, epistemic model of economic decision.7

5.1 Problems as conjectures

mulating them?

Where do problems come from? Are there any logically sound methods for for- InWhere do problems come from? Are there any logically sound methods for for- anWhere do problems come from? Are there any logically sound methods for for- epistemicWhere do problems come from? Are there any logically sound methods for for- perspective,Where do problems come from? Are there any logically sound methods for for- theWhere do problems come from? Are there any logically sound methods for for- firstWhere do problems come from? Are there any logically sound methods for for- logicalWhere do problems come from? Are there any logically sound methods for for- operationWhere do problems come from? Are there any logically sound methods for for- inWhere do problems come from? Are there any logically sound methods for for- anyWhere do problems come from? Are there any logically sound methods for for- decisionWhere do problems come from? Are there any logically sound methods for for- isWhere do problems come from? Are there any logically sound methods for for-

problem modelling Second, in an epistemic perspective, a model of a situation

is made of conjectural knowledge and needs testing for reliability

Let’s then define a problem model as a set of hypotheses on cause–effect tionships, where alternatives are causes (explanans) and consequences are

rela-Micro foundations 27 effects (explananda), conjectured to be valid under some conditions (states of the world or contingencies).

Examples are propositions such as: analytical skills sustain performance in high education programmes in science-based degrees; investments in IT increase decision making quality/efficiency if decision makers use them; information dis-closure attracts investments if investors read reports

The impact of the quality of those causal hypotheses, entering almost any decision processes, on the quality of economic decision making cannot be overstated An example can be the (periodically reappearing) discussion on how to increase the productivity of work A lot of causal modelling is involved, but rarely is a rigorous modelling and testing of these hypotheses undertaken before taking action (sometimes with very high costs) For example: the management of Fiat recently raised a dramatic incident in instru-mental rationality deciding to reduce production pauses for increasing pro-ductivity (and threatening to close production in Italy if the agreement was not signed) Little consideration, in that decision, has been paid to research and theory, on how a reduction of production pauses may increase productiv-ity (actually, it does so only up to a point in fatigue and attention curves, as demonstrated by Taylorist researchers, an irony of history) Apparently no consideration has been given to the known fact that productivity effects of

logistic factors can be expected only under ‘ceteris paribus’ conditions – in

particular to the fact that social rewards may overshadow material conditions effects – (as demonstrated by human relation researchers, in the same setting

tric in the 1950s was deciding more rationally on the organization of work than Fiat in the 2000s – in spite of both wishing to ‘maximize productivity’ – because General Electric was doing so by gathering more valid and reliable knowledge, both general on the phenomenon of productivity and specific on the system under ‘reform’

of the famous Hawthorne experiment) So, it may be said that General Elec- Ifof the famous Hawthorne experiment) So, it may be said that General Elec- aof the famous Hawthorne experiment) So, it may be said that General Elec- problemof the famous Hawthorne experiment) So, it may be said that General Elec- isof the famous Hawthorne experiment) So, it may be said that General Elec- aof the famous Hawthorne experiment) So, it may be said that General Elec- conjecture,of the famous Hawthorne experiment) So, it may be said that General Elec- thenof the famous Hawthorne experiment) So, it may be said that General Elec- itof the famous Hawthorne experiment) So, it may be said that General Elec- shouldof the famous Hawthorne experiment) So, it may be said that General Elec- beof the famous Hawthorne experiment) So, it may be said that General Elec- askedof the famous Hawthorne experiment) So, it may be said that General Elec- howof the famous Hawthorne experiment) So, it may be said that General Elec- itof the famous Hawthorne experiment) So, it may be said that General Elec- canof the famous Hawthorne experiment) So, it may be said that General Elec- beof the famous Hawthorne experiment) So, it may be said that General Elec- generatedof the famous Hawthorne experiment) So, it may be said that General Elec- (the ‘logic of discovery’ of the conjecture) and how it can be tested (the ‘logic of justification’ of the conjecture) in a valid way

The empirical, pattern-recognition-based, abduction of ‘laws’ of causal

rela-tion between alternatives–causes and outcomes–effects is an epistemically rational heuristics when observation and historical evidence is available

Is there any method for generating problems-hypotheses in a logically sound way, even without experience, data or observation of regularities and patterns? Consider the following story, drawn from a personal interview to a world-class scientist and entrepreneur.8

The Olive Water case

I am a chemist by training I founded five firms in 25 years I am also fond

of natural food and environment-friendly agriculture I was thinking of buying a piece of land in Tuscany to spend some holiday time at, cultivate