Trust Theory
Cristiano Castelfranchivisibility
…
description
13 pages
link
2 files
Sign up for access to the world's latest research
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Abstract
We present the index of the book & the draft of one crucial Chapter of the book in its non final and correct version For the correct and complete version see our book "Trust Theory. A socio-cognitive and computational model" Wiley, 2010
Related papers
Socio-cognitive model of trust
2005
Humans have learned to cooperate in many ways and in many environments, on different tasks, and for achieving different and several goals. Collaboration and cooperation in their more general sense (and, in particular, negotiation, exchange, help, delegation, adoption, and so on) are important characteristics-or better, the most foundational aspects-of human societies (Tuomela, 1995).
A general definition of trust
In this paper a definition and conceptual analysis of trust is given in terms of trustworthiness. Its focus will be as wide as possible, and will not be restricted to any particular type of trust. The aim is to show the key parameters that enable us to investigate and understand trust, thereby facilitating the development of systems, institutions and technologies to support, model or mimic trust. The paper will also show the strong connection between trust and trustworthiness, showing how the subjectivity of trust reveals itself in attitudes toward others' trustworthiness; to trust someone/something is to believe that he/she/it is trustworthy. Both trust and trustworthiness are context-dependent, but the relevant contexts are different depending on whether one is trusting or trustworthy. Finally, the paper will discuss some of the complex issues connected with the alignment of trust with trustworthiness.
Trust and Technology
In this book our effort is to model and rationalize the trust notion trying to catch all the different and varying aspects of this broad concept. In fact, there are today many studies, models, simulations and experiments trying to integrate trust in the technological infrastructures: The most advanced disciplines in Human-) are forced to cope with trust. But why does trust seem to be so important in the advanced technological contexts? Is it necessary to involve such a complex, fuzzy and human related concept? Is it not sufficient to consider just more technical and simply applicable notions like security? To give a satisfactory response to these questions we have to evaluate which kind of network infrastructures are taken into consideration in the new communication and interaction scenarios, which kind of peculiar features should have the artificial agents we have to cope with, which kind of computing is going to invade (pervade) the future physical environments? In fact, trust becomes fundamental in the open multi-agent systems where the agents (which could be both human beings and artificial agents owned by other human stakeholders) can (more or less freely) enter and leave the system. The evolution of the interaction and communication technological paradigms toward human style, is, on the one hand, a really difficult task to realize, but, on the other hand, it potentially increases the people accessing to (and fruitful in using) the new technologies. In fact, in the history of their evolution humans have learned to cooperate in many ways and environments; on different tasks; and to achieve different goals. They have intentionally realized (or they were spontaneously emerging) diverse cooperative constructs (purely interactional, technical-legal, organizational, socio-cognitive, etc.) for establishing trust among them. It is now necessary to remodel the trust concept in the new current and future scenarios (new channels and infrastructures of communication; new artificial entities, new environments) and the efforts in the previously cited scientific fields (HCI, MAS, DAI, NCS) are trying to give positive answers to these main requirements. Trust Theory: A Socio-Cognitive and Computational Model Cristiano Castelfranchi and Rino Falcone
Social Trust: A Cognitive Approach
Trust and Deception in Virtual Societies, 2001
After arguing about the crucial importance of trust for Agents and MAS, we provide a definition of trust both as a mental state and as a social attitude and relation. We present the mental ingredients of trust: its specific beliefs and goals, with special attention to evaluations and expectations. We show the relation between trust and the mental background of delegation. We explain why trust is a bet, and implies some risks, and analyse the more complex forms of social trust, based on a theory of mind and in particular on morality, reputation and disposition, and authority (three party trust). We explain why promises, contracts, authorities can increase our trust by modifying our mental representations. We present a principled quantification of trust, based on its cognitive ingredients, and use this "degree of trust" as the basis for a rational decision to delegate or not to another agent. We explain when trust is rational, and why it is not an irrational decision by definition. We also criticise the economic and game-theoretic view of trust for underestimating the importance of cognitive ingredients of trust and for reducing it to subjective probability and risk. The paper is intended to contribute both to the conceptual analysis and to the practical use of trust in social theory and MAS. * This research has been developed within the agreement between CNR and Provincia Autonoma di Trento, research project on "Applicazioni avanzate di informatica". We would like to thanks Maria Miceli, YaoHua Tan, Maj Bonniver Tuomela, Giovan Francesco Lanzara, and Walter Thoen for useful comments and discussions.
Trust: Decision problem or performance
Established theories on trust-development explain trust as a product of information, portraying individuals as passive receptacles of trust-related information. These theories however tend to ignore the goals of the trustor. In short existing contribution on trusts tend to portray the development of trust as a decision problem. In this paper we suggest an alternative perspective of trust and trust formation, where we see trust as an intentional goal initiated performance. People trust other people in order to attain social ends. In order to effectively communicate trust people seek to influence their own subjective experience of trust. We refer to this second perspective as the Trust-as-Performance-perspective. The assumptions underlying the two perspectives are compared and contrasted. We show how the two perspectives can be seen as corresponding to different models of trust development. The final section discusses the possible relationships between the perspectives and how the two perspectives together may improve our understanding of trust in relationships.
Trust: A Collision of Paradigms
Financial Cryptography, 2002
The technological challenges of securing networks are great, as recently witnessed in widespread denial of service and virus attacks. The human reaction to these attacks may be either a loss of trust or a willingness to tolerate increasing risk having weathered one assault. Examining human and computer interaction with a focus on evaluations the human response to loss of trust is a key part of the search for more secure networks. The success of current efforts to design appropriate security mechanisms depends as much on an understanding of human extensions of trust to computers as it does on an understanding of underlying mathematics. However, the former has not been sufficiently examined. In this work we survey the findings in social psychology and philosophy with respect to trust. We introduce three hypotheses that remain unanswered with respect to the manner in which humans react to computers. We discuss potential design revisions in light of findings from other disciplines. Then we conclude by noting that research which empowers users in order to be their own security manager may be based on a fundamentally flawed view of human-computer interaction. We close by encouraging designers of computer security systems to examine the humans, which these systems are intended to empower, and recommend that any security system be built on the basis of understanding of human trust provided by the social sciences.
The meanings of trust
1996
612) 624-4074 ____________________ Our trust conceptualizations have benefited from discussions with Ellen Berscheid and Larry Cummings of the University of Minnesota. The authors also thank three anonymous reviewers from the Organizational Behavior division of the 1996 meeting of the Academy of Management for their comments on an earlier version of this paper.
The Psychology (and Economics) of Trust
Social and Personality Psychology Compass, 2009
Interpersonal trust is a mental construct with implications for social functioning and economic behavior. We review contemporary theories of trust from behavioral economics and social psychology. Neoclassical economic theory considers trust in strangers to be irrational, but observed behavior reveals widespread trust and trustworthiness. Theories of social preferences and adherence to social norms have been proposed to rationalize trust. Psychological approaches investigate trusting behavior in terms of an underlying disposition, intergroup processes, and cognitive expectations. The breadth of these approaches illustrates the multi-faceted nature of trusting behavior. The determinants of trust are related to the relevant characteristics of the individual, the situation, and their interaction. 'The entire fabric of our day-today living, of our social world, rests on trust-buying gasoline, paying taxes, going to the dentist, flying to a convention-almost all our decisions involve trusting someone else'. (Rotter, 1970, p. 443) Psychologists from diverse backgrounds have identified interpersonal trust as a social construct of far-reaching significance. The noted developmentalist Erik Erikson (1950), called trust 'the first task of the ego' (p. 221). He argued that the ability to trust others is necessary to form relationships and function in the social world. There are inherent risks in any social situation and we often trust that others will not attack us, tell our secrets, or break our hearts. Without the capacity to trust, we would be unable to accept these ordinary uncertainties. Contemporary research examines trust in many domains; such as close relationships (Simpson, 2007), political beliefs (Brewer, 2004), community involvement (Rahn & Transue, 1998), and organizational behavior (Kramer, 2007). The consequences of trust are observed in the workplace, in intimate relationships, and in society at large. The question of why we trust also has implications in economic exchanges. Nobel laureate Kenneth Arrow (1974) called trust 'a lubricant for social systems' (p. 23). A consumer trusts that the car he buys is not a lemon; an employer trusts that his new assistant will be dependable; and an investor trusts that corporate accountants are not cooking the books. The micro-level trust that exists between individuals contributes to the economic development of nations (Fukuyama, 1995; Zak & Knack, 2000). Trust allows a society to invest more capital and conduct transactions with greater efficiency. In recent years, many behavioral (or psychological) economists have begun to examine trust and its consequences. Their work extends traditional theories of economic behavior with theoretical constructs and experimental methods adopted from psychology. According to a widely accepted definition, trust is 'a psychological state comprising the intention to accept vulnerability based upon the positive expectations of the intentions or behavior of another' (Rousseau, Sitkin, Burt, & Camerer, 1998, p. 395). There are numerous variations on this definition, but the critical elements of any definition of trust Social and Personality Psychology Compass 3/
A Conceptual Framework for Trust Models
Trust, Privacy and Security in Digital Business, 2012
During the last twenty years, a huge amount of trust and reputation models have been proposed, each of them with their own particularities and targeting different domains. While much effort has been made in defining ever-increasing complex models, little attention has been paid to abstract away the particularities of these models into a common set of easily understandable concepts. We propose a conceptual framework for computational trust models that will be used for analyzing their features and for comparing heterogeneous and relevant trust models.
A Formal Model of Trust Based on Outcomes
Academy of management …, 1998
The study of trust has occupied scholars from a number of disciplines, who have made limited attempts at integrating the different streams. One reason for lack of integration is that no clear definition of trust exists. In this article we grapple with this issue by going back to first principles to derive a mathematically precise and statistically rigorous definition of trust. In giving a rigorous meaning of trust, we also capture the key elements of the concept as highlighted by various disciplinary lenses. Our definition of trust, although rationally based, is consistent with many of the findings of earlier behavioral and sociological research. Its contribution is in adding precision and richness to our understanding of how trust is created and maintained in various social and economic interactions.
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
TRUST THEORY
Trust Theory: A Socio-Cognitive and Computational Model Cristiano Castelfranchi and Rino Falcone
© 2010 John Wiley & Sons Ltd. ISBN: 978-0-470-02875-9
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Wiley Series in Agent Technology
Series Editor: Michael Wooldridge, University of Liverpool, UK
The ‘Wiley Series in Agent Technology’ is a series of comprehensive practical guides and
cutting-edge research titles on new developments in agent technologies. The series focuses
on all aspects of developing agent-based applications, drawing from the Internet, telecom-
munications, and Artificial Intelligence communities with a strong applications/technologies
focus.
The books will provide timely, accurate and reliable information about the state of the art
to researchers and developers in the Telecommunications and Computing sectors.
Titles in the series:
Padgham/Winikoff: Developing Intelligent Agent Systems 0-470-86120-7 (June 2004)
Bellifemine/Caire/Greenwood: Developing Multi-Agent Systems with JADE 0-470-05747-5
(February 2007)
Bordini/Hübner/Wooldrige: Programming Multi-Agent Systems in AgentSpeak using Jason
0-470-02900-5 (October 2007)
Nishida: Conversational Informatics: An Engineering Approach 0-470-02699-5 (November
2007)
Jokinen: Constructive Dialogue Modelling: Speech Interaction and Rational Agents 0-470-
06026-3 (April 2009)
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
TRUST THEORY
A SOCIO-COGNITIVE AND
COMPUTATIONAL MODEL
Cristiano Castelfranchi
Institute of Cognitive Sciences and Technologies (ISTC) of the Italian National Research
Council (CNR), Italy
Rino Falcone
Institute of Cognitive Sciences and Technologies (ISTC) of the Italian National Research
Council (CNR), Italy
A John Wiley and Sons, Ltd., Publication
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
This edition first published 2010
C 2010 John Wiley & Sons Ltd.,
Registered office
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom
For details of our global editorial offices, for customer services and for information about how to apply for permission to
reuse the copyright material in this book please see our website at www.wiley.com.
The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright,
Designs and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form
or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright,
Designs and Patents Act 1988, without the prior permission of the publisher.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in
electronic books.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and
product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective
owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to
provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that
the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required,
the services of a competent professional should be sought.
Library of Congress Cataloging-in-Publication Data
Castelfranchi, Cristiano.
Trust theory : a socio-cognitive and computational model / Cristiano Castelfranchi, Rino Falcone.
p. cm.
Includes index.
ISBN 978-0-470-02875-9 (cloth)
1. Trust. 2. Trust–Simulation methods. 3. Artificial intelligence–Psychological aspects. 4. Cognitive science.
I. Falcone, Rino. II. Title.
BF575.T7C37 2010
302’.1–dc22
2009040166
A catalogue record for this book is available from the British Library.
ISBN 9780470028759 (H/B)
Typeset in 10/12pt Times by Aptara Inc., New Delhi, India
Printed and Bound in Singapore by Markono.
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
This book is dedicated
to Mario, Ketty, Eugenio, and Maria, our roots,
to Rosanna and Ivana, our life companions;
to Yurij, Vania and Giulio, our realized dreams;
to the many collegues who consciously or unconsciuously contributed to the
ideas included in it;
to our Country, better to the Idea of Country, of Collective, of Institutional
Entity, in which as such it is possible develop socially centred hopes,
ambitions, dreams, so contributing to the dignity and future for any individual.
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Contents
Foreword xv
Introduction 1
1 Definitions of Trust: From Conceptual Components to the General Core 7
1.1 A Content Analysis 8
1.2 Missed Components and Obscure Links 12
1.3 Intentional Action and Lack of Controllability: Relying on What is
Beyond Our Power 15
1.4 Two Intertwined Notions of Trust: Trust as Attitude vs. Trust as Act 17
1.5 A Critique of Some Significant Definitions of Trust 19
1.5.1 Gambetta: Is Trust Only About Predictability? 19
1.5.2 Mayer, Davis, & Schoorman: Is Trust Only Willingness,
for Any Kind of Vulnerability? 19
1.5.3 McKnight: The Black Boxes of Trust 21
1.5.4 Marsh: Is a Mere Expectation Enough for Modeling Trust? 21
1.5.5 Yamagishi: Mixing up the Act of Trusting and the Act of
Cooperating 22
1.5.6 Trust as Based on Reciprocity 26
1.5.7 Hardin: Trust as Encapsulated Interest 26
1.5.8 Rousseau: What Kind of Intention is ‘Trust’? 30
References 31
2 Socio-Cognitive Model of Trust: Basic Ingredients 35
2.1 A Five-Part Relation and a Layered Model 36
2.1.1 A Layered Notion 36
2.1.2 Goal State and Side Effects 38
2.2 Trust as Mental Attitude: a Belief-Based and Goal-Based Model 38
2.2.1 Trust as Positive Evaluation 39
2.2.2 The ‘Motivational’ Side of Trust 44
2.2.3 The Crucial Notion of ‘Goal’ 45
2.2.4 Trust Versus Trustworthiness 47
2.2.5 Two Main Components: Competence Versus Predictability 47
2.2.6 Trustworthiness (and trust) as Multidimensional
Evaluative Profiles 49
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
viii Contents
2.2.7 The Inherently Attributional Nature of Trust 50
2.2.8 Trust, Positive Evaluation and Positive Expectation 52
2.3 Expectations: Their Nature and Cognitive Anatomy 54
2.3.1 Epistemic Goals and Activity 54
2.3.2 Content Goals 55
2.3.3 The Quantitative Aspects of Mental Attitudes 56
2.3.4 The Implicit Counterpart of Expectations 58
2.3.5 Emotional Response to Expectation is Specific: the Strength
of Disappointment 58
2.3.6 Trust is not Reducible to a Positive Expectation 60
2.4 ‘No Danger’: Negative or Passive or Defensive Trust 60
2.5 Weakening the Belief-Base: Implicit Beliefs, Acceptances, and Trust
by-Default 62
2.6 From Disposition to Action 64
2.6.1 Trust That and Trust in 66
2.6.2 Trust Pre-disposition and Disposition: From Potential to
Actual Trust 67
2.6.3 The Decision and Act of Trust Implies the Decision to Rely on 69
2.7 Can we Decide to Trust? 72
2.8 Risk, Investment and Bet 73
2.8.1 ‘Risk’ Definition and Ontology 74
2.8.2 What Kinds of Taken Risks Characterize Trust Decisions? 76
2.9 Trust and Delegation 77
2.9.1 Trust in Different Forms of Delegation 79
2.9.2 Trust in Open Delegation Versus Trust in Closed Delegation 80
2.10 The Other Parts of the Relation: the Delegated Task and the Context 82
2.10.1 Why Does X Trust Y? 82
2.10.2 The Role of the Context/Environment in Trust 83
2.11 Genuine Social Trust: Trust and Adoption 84
2.11.1 Concern 88
2.11.2 How Expectations Generate (Entitled) Prescriptions:
Towards ‘Betrayal’ 88
2.11.3 Super-Trust or Tutorial Trust 89
2.12 Resuming the Model 91
References 92
3 Socio-Cognitive Model of Trust: Quantitative Aspects 95
3.1 Degrees of Trust: a Principled Quantification of Trust 95
3.2 Relationships between Trust in Beliefs and Trust in Action
and Delegation 97
3.3 A Belief-Based Degree of Trust 98
3.4 To Trust or Not to Trust: Degrees of Trust and Decision to Trust 101
3.5 Positive Trust is not Enough: a Variable Threshold for Risk
Acceptance/Avoidance 107
3.6 Generalizing the Trust Decision to a Set of Agents 111
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Contents ix
3.7 When Trust is Too Few or Too Much 112
3.7.1 Rational Trust 112
3.7.2 Over-Confidence and Over-Diffidence 112
3.8 Conclusions 114
References 115
4 The Negative Side: Lack of Trust, Implicit Trust, Mistrust,
Doubts and Diffidence 117
4.1 From Lack of Trust to Diffidence: Not Simply a Matter of Degree 117
4.1.1 Mistrust as a Negative Evaluation 118
4.2 Lack of Trust 119
4.3 The Complete Picture 120
4.4 In Sum 121
4.5 Trust and Fear 122
4.6 Implicit and by Default Forms of Trust 122
4.6.1 Social by-Default Trust 124
4.7 Insufficient Trust 125
4.8 Trust on Credit: The Game of Ignorance 126
4.8.1 Control and Uncertainty 126
4.8.2 Conditional Trust 127
4.8.3 To Give or Not to Give Credit 127
4.8.4 Distrust as Not Giving Credit 129
References 131
5 The Affective and Intuitive Forms of Trust: The Confidence We Inspire 133
5.1 Two Forms of ‘Evaluation’ 134
5.2 The Dual Nature of Valence: Cognitive Evaluations Versus
Intuitive Appraisal 134
5.3 Evaluations 135
5.3.1 Evaluations and Emotions 136
5.4 Appraisal 137
5.5 Relationships Between Appraisal and Evaluation 138
5.6 Trust as Feeling 140
5.7 Trust Disposition as an Emotion and Trust Action as an Impulse 141
5.8 Basing Trust on the Emotions of the Other 142
5.9 The Possible Affective Base of ‘Generalized Trust’ and
‘Trust Atmosphere’ 143
5.10 Layers and Paths 143
5.11 Conclusions About Trust and Emotions 144
References 145
6 Dynamics of Trust 147
6.1 Mental Ingredients in Trust Dynamics 148
6.2 Experience as an Interpretation Process: Causal Attribution for Trust 150
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
x Contents
6.3 Changing the Trustee’s Trustworthiness 154
6.3.1 The Case of Weak Delegation 154
6.3.2 The Case of Strong Delegation 158
6.3.3 Anticipated Effects: A Planned Dynamics 161
6.4 The Dynamics of Reciprocal Trust and Distrust 164
6.5 The Diffusion of Trust: Authority, Example, Contagion, Web of Trust 168
6.5.1 Since Z Trusts Y, Also X Trusts Y 168
6.5.2 Since X Trusts Y, (by Analogy) Z Trusts W 173
6.5.3 Calculated Influence 173
6.6 Trust Through Transfer and Generalization 174
6.6.1 Classes of Tasks and Classes of Agents 175
6.6.2 Matching Agents’ Features and Tasks’ Properties 175
6.6.3 Formal Analysis 177
6.6.4 Generalizing to Different Tasks and Agents 178
6.6.5 Classes of Agents and Tasks 182
6.7 The Relativity of Trust: Reasons for Trust Crisis 184
6.8 Concluding Remarks 188
References 189
7 Trust, Control and Autonomy: A Dialectic Relationship 191
7.1 Trust and Control: A Complex Relationship 191
7.1.1 To Trust or to Control? Two Opposite Notions 192
7.1.2 What Control is 192
7.1.3 Control Replaces Trust and Trust Makes Control Superflous? 195
7.1.4 Trust Notions: Strict (Antagonist of Control) and Broad
(Including Control) 196
7.1.5 Relying on Control and Bonds Requires Additional Trust:
Three Party Trust 198
7.1.6 How Control Increases and Complements Trust 200
7.1.7 Two Kinds of Control 201
7.1.8 Filling the Gap between Doing/Action and Achieving/Results 203
7.1.9 The Dynamics 204
7.1.10 Control Kills Trust 205
7.1.11 Resuming the Relationships between Trust and Control 206
7.2 Adjusting Autonomy and Delegation on the Basis of Trust in Y 206
7.2.1 The Notion of Autonomy in Collaboration 209
7.2.2 Delegation/Adoption Theory 209
7.2.3 The Adjustment of Delegation/Adoption 213
7.2.4 Channels for the Bilateral Adjustments 222
7.2.5 Protocols for Control Adjustments 223
7.2.6 From Delegation Adjustment to Autonomy Adjustment 225
7.2.7 Adjusting Meta-Autonomy and Realization-Autonomy of the Trustee 225
7.2.8 Adjusting Autonomy by Modyfing Control 226
7.2.9 When to Adjust the Autonomy of the Agents 227
7.3 Conclusions 230
References 232
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Contents xi
8 The Economic Reductionism and Trust (Ir)rationality 235
8.1 Irrational Basis for Trust? 236
8.1.1 Is Trust a Belief in the Other’s Irrationality? 236
8.2 Is Trust an ‘Optimistic’ and Irrational Attitude and Decision? 239
8.2.1 The Rose-Tinted Glasses of Trust 239
8.2.2 Risk Perception 246
8.3 Is Trust Just the Subjective Probability of the Favorable Event? 247
8.3.1 Is Trust Only about Predictability? A Very Bad Service but
a Sure One 247
8.3.2 Probability Collapses Trust ‘that’ and ‘in’ 248
8.3.3 Probability Collapses Internal and External
(Attributions of) Trust 248
8.3.4 Probability Misses the Active View of Trust 250
8.3.5 Probability or Plausibility? 250
8.3.6 Probability Reduction Exposes to Eliminative Behavior:
Against Williamson 250
8.3.7 Probability Mixes up Various Kinds of Beliefs, Evaluations,
Expectations about the Trustee and Their Mind 252
8.4 Trust in Game Theory: from Opportunism to Reciprocity 254
8.4.1 Limiting Trust to the Danger of Opportunistic Behavior 255
8.4.2 ‘To Trust’ is not ‘to Cooperate’ 255
8.5 Trust Game: A Procuste’s Bed for Trust Theory 256
8.6 Does Trust Presuppose Reciprocity? 258
8.7 The Varieties of Trust Responsiveness 260
8.8 Trusting as Signaling 260
8.9 Concluding Remarks 261
References 261
9 The Glue of Society 265
9.1 Why Trust is the ‘Glue of Society’ 265
9.2 Trust and Social Order 266
9.2.1 Trust Routinization 268
9.3 How the Action of Trust Acquires the Social Function of Creating Trust 268
9.4 From Micro to Macro: a Web of Trust 270
9.4.1 Local Repercussions 270
9.4.2 Trans-Local Repercussions 271
9.5 Trust and Contracts 272
9.5.1 Do Contracts Replace Trust? 272
9.5.2 Increasing Trust: from Intentions to Contracts 272
9.5.3 Negotiation and Pacts: Trust as Premise and Consequence 275
9.6 Is Trust Based on Norms? 275
9.6.1 Does Trust Create Trust and does There Exist a Norm of
Reciprocating Trust? 277
9.7 Trust: The Catalyst of Institutions 278
9.7.1 The Radical Trust Crisis: Institutional Deconstruction 279
References 279
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
xii Contents
10 On the Trustee’s Side: Trust As Relational Capital 281
10.1 Trust and Relational Capital 282
10.2 Cognitive Model of Being Trusted 284
10.2.1 Objective and Subjective Dependence 285
10.2.2 Dependence and Negotiation Power 289
10.2.3 Trust Role in Dependence Networks 292
10.3 Dynamics of Relational Capital 297
10.3.1 Increasing, Decreasing and Transferring 297
10.3.2 Strategic Behavior of the Trustee 300
10.4 From Trust Relational Capital to Reputational Capital 301
10.5 Conclusions 302
References 302
11 A Fuzzy Implementation for the Socio-Cognitive Approach to Trust 305
11.1 Using a Fuzzy Approach 306
11.2 Scenarios 306
11.3 Belief Sources 307
11.4 Building Belief Sources 307
11.4.1 A Note on Self-Trust 309
11.5 Implementation with Nested FCMs 310
11.6 Converging and Diverging Belief Sources 311
11.7 Homogeneous and Heterogeneous Sources 312
11.8 Modeling Beliefs and Sources 312
11.9 Overview of the Implementation 313
11.9.1 A Note on Fuzzy Values 315
11.10 Description of the Model 316
11.11 Running the Model 316
11.12 Experimental Setting 317
11.12.1 Routine Visit Scenario 317
11.12.2 Emergency Visit Scenario 319
11.12.3 Trustfulness and Decision 320
11.12.4 Experimental Discussion 321
11.12.5 Evaluating the Behavior of the FCMs 322
11.12.6 Personality Factors 322
11.13 Learning Mechanisms 323
11.13.1 Implicit Revision 324
11.13.2 Explicit Revision 324
11.13.3 A Taxonomy of Possible Revisions 325
11.14 Contract Nets for Evaluating Agent Trustworthiness 326
11.14.1 Experimental Setting 326
11.14.2 Delegation Strategies 327
11.14.3 The Contract Net Structure 328
11.14.4 Performing a Task 329
11.14.5 FCMs for Trust 329
11.14.6 Experiments Description 330
11.14.7 Using Partial Knowledge: the Strength of a Cognitive Analysis 333
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Contents xiii
11.14.8 Results Discussion 339
11.14.9 Comparison with Other Existing Models and Conclusions 341
References 342
12 Trust and Technology 343
12.1 Main Difference Between Security and Trust 344
12.2 Trust Models and Technology 345
12.2.1 Logical Approaches 346
12.2.2 Computational Approach 347
12.2.3 Different Kinds of Sources 347
12.2.4 Centralized Reputation Mechanisms 348
12.2.5 Decentralized Reputation Mechanisms 349
12.2.6 Different Kinds of Metrics 350
12.2.7 Other Models and Approaches to Trust in the
Computational Framework 351
12.3 Concluding Remarks 354
References 354
13 Concluding Remarks and Pointers 359
13.1 Against Reductionism 359
13.2 Neuro-Trust and the Need for a Theoretical Model 360
13.3 Trust, Institutions, Politics (Some Pills of Reflection) 361
13.3.1 For Italy (All’Italia) 362
References 363
Index 365
For a schematic view of the main terms introduced and analyzed in this book see the Trust,
Theory and Technology site at http://www.istc.cnr.it/T3/.
P1: OTE/OTE/SPH P2: OTE
FM JWBK436-Castelfranchi January 21, 2010 16:18 Printer Name: Yet to Come
Foreword
I turn up to give a lecture at 9 am on a Monday morning, trusting that my students will attend;
and they in turn reluctantly drag themselves out of bed to attend, trusting that I will be there
to give the lecture. When my wife tells me that she will collect our children from school, I
expect to see the children at home that night safe and sound. Every month, I spend money,
trusting that, on the last Thursday of the month, my employer will deposit my salary in my
bank account; and I trust my bank to safeguard this money, investing my savings prudently.
Sometimes, of course, my trust is misplaced. Students don’t turn up to lectures; my bank
makes loans to people who have no chance of repaying them, and as a consequence they go
bankrupt, taking my savings with them. But despite such disappointments, our lives revolve
around trust: we could hardly imagine society functioning without it.
The rise of autonomous, computer-based agents as a technology gives trust an interesting
new dimension. Of course, one issue is that we may not be comfortable trusting a computer
program to handle our precious savings. But when software agents interact with people an
entirely new concern arises: why or how should a computer program trust ‘us’? How can we
design computer programs that are safe from exploitation by un-trustworthy people? How can
we design software agents that can understand how trust works in human societies, and live up
to human expectations of trust? And what kind of models of trust make sense when software
agents interact with ‘other’ software agents?
These considerations have led to attempts by cognitive scientists, computer scientists,
psychologists, and others, to develop models of trust, and to implement these tentative models
of trust in computer programs. The present book is the first comprehensive overview of the
nascent field of modeling trust and computational models of trust. It discusses trust and the
allied concept of reputation from a range of different backgrounds. It will be essential reading
for anybody who wants to understand the issues associated with building computer systems
that work with people in sensitive situations, and in particular for researchers in multi-agent
systems, who will deploy and build on the techniques and concepts presented herein. The
journey to understand trust from a scientific, technological, and computational perspective
may only just have begun, but this book represents a critical milestone on that journey.
Michael Wooldridge
Chapter 2. Socio-Cognitive Model of Trust: Basic Ingredients
Trust means different things, but they are systematically related one with the other. In particular we analyze three crucial concepts that have been recognized and distinguished in the scientific literature. Trust is:
• A mere mental attitude (prediction and evaluation) towards an other agent; a simple (pre)disposition;
Or even just a feeling, an affective disposition where those ‘beliefs’ are just implicit (see Chapter 8).
this mental attitude is in fact an opinion, a judgment, a preventive evaluation about specific and relevant ‘virtues’ needed for relying on the potential trustee, but that might remain separated from the actual exercise of trust.
• A decision to rely upon the other, i.e. an intention to delegate and trust, which makes the trustor "vulnerable" [Mayer et al, 1995];
this is again a mental attitude, but it is the result of a complex comparison and match among the preventive evaluations of the different potential trustees, about the risks and the costs, and about the applicability of these evaluations to the actual environments and context.
• A behavior, i.e. the intentional act of (en)trusting, and the consequent overt and practical relation between the trustor and the trustee;
this is the consequent act, behavior, of the trustor, generally coordinated and coherent with the previous decision; and the public ‘announcement’ and social relation.
Trust is in general all these things together.
For understanding this concept in its real meaning we have to analyze it, and showing the complex relations among different terms and constituents.
2.1 A five-part relation and a layered model
In our view trust is a relational construct between:
- An Agent X (the trustor) (we will name Xania, a woman). This agent X is necessarily an “intentional entity” (see intentional stance [Dennett, 1989]); let’s say a “cognitive agent”, that is, an agent with mental ingredients (beliefs, goals, intentions, and so on). Trust must be (also) described in cognitive terms, as X’s specific mental attitudes towards other agents and about given tasks.
- An Addressee Y (the trustee) that is an “agent” in the broader sense of this term [Castelfranchi. 1998; 2000a]: an entity able to cause some effect (outcome) in the world; the outcome X is waiting for. When Y is an intentional agent: we will name it Yody, and he is a man.
- Such a “causal” process (the act, or performance) and its result; that is, an act a of Y possibly producing the outcome p; which is positive or desirable because includes (or corresponds to) the content of a goal of X (GoalX(g)=gX), the specific goal for which X is trusting Y. We call this act: Y’s task: t. t is the couple (a, p), with g included in p or in some cases p=g.
- gX =GoalX(g) is in fact a crucial element of the trust concept and relation, frequently omitted.
- a context (C) or situation or environment where X takes into account Y (thus affecting X’s evaluation and choice of Y) and/or where Y is supposed to act and to realize the task (thus affecting the possibility of success).
In other words, in our model Trust is a five-part relation (at least):
TRUST (X Y C t=(a, p), gX) (2.1)
that can be read as X trusts (in) Y in context C for performing action a (executing task t) and realizing the result p (that includes or corresponds to her goal GoalX(g)=gX).
A deep analysis of each component is needed, and a theory of their relationships and variations.
2.1.1 A layered notion
As we said, we consider and analyze Trust as a composed and “layered” notion, where the various more or less complex meanings are not just in a ‘family resemblance’ relation, but are embedded one into the other, and it is important the explicit theory of the relations between those layers and of the transition from one to the other; since the most simple form can be there without the richer one, but not vice versa.
As we said, in our theory there is a double link between these forms of trust (from the dispositional one to the active one): a conceptual link and a process/causal link (see §1.4).
<INSERT Figure 2.1 here>
Figure2.1:
Trust stages and layers
Trust as a layered notion means:
- in its basic sense just a mental (cognitive and affective
We will put aside here the affective nature of Trust; trust as an intuitive disposition, as a feeling. Here we consider Trust as an explicit (and usually grounded) judgment. We dedicate the entire Chapter 8 to affective trust dispositions and impulses. Consider however, that those two ‘forms’ of Trust can coexist: there may be an explicit evaluation joined with (eliciting or based on) a feeling [Castelfranchi,2000b].) attitude and disposition towards Y (beliefs: evaluations and expectations); this already is a social relation;
- in its broader and richer use, a decision and intention based on that disposition;
- and then the act of relying upon Y’s expected behavior;
- and the consequent overt social interaction and relation with Y.
The trust attitude/disposition is a determinant and precursor of the decision, which is a determinant and a precondition of the action; however, the disposition and the intention remain as necessary mental constituents during the next steps.
Let us introduce an example for better clarify the different concepts: Xania is the new director of a firm; Mary and Yody are employees, possible candidates for the role of Xania's secretary. Now, consider these three different aspects (and components) of Xania's trust towards them:
1) Xania, evaluates Mary on the basis of her personal experience with Mary -if any-, of Mary’s CV and of what the others say about her; on such a basis she forms her own opinion and evaluation about Mary: how much she considers Mary trustworthy as a secretary? Will this trust be enough for choosing Mary and decide to bet on her?
2) Now Xania has also considered Yody’s CV, reputation, etc.; she knows that there are no other candidates and decides of counting on Yody as her secretary; i.e. she has the intention of delegating to Yody this job/task when she will reorganize her staff.
3) One week later in fact Xania nominates Yody as her secretary and uses him, thus actually she trusts him for this job: the trust relationship is established.
In situation (1) Xania has to formulate some kind of trust that is in fact an evaluation about Mary as a secretary. This trust might be sufficient or not in that situation. Since she knows that there is also another candidate she cannot decide to choose Mary (she has to wait for this) but she can just express a mere evaluation on her, and perhaps this evaluation is not good enough for being chosen (also before next candidates’ evaluations).
In situation (2) Xania arrives to a decision that is based on trust and is the decision to trust. It is also possible that -in Xania’s opinion- neither Mary nor Yody are suitable persons with respect the secretary’s role and she might take the decision of not trusting them, and of considering different hypotheses: i) to send a new call for secretary or ii) to choose the less worst between them for a brief period, looking for new candidates in parallel; or iii) to stay a period without any secretary; and so on.
In situation (3) Xania expresses her trust through an (official) delegation, in fact an act of communication (or more in general, through an observable external behaviour). In general, this kind of relationship has not necessary to be official, it is simply known to the two agents (as we will see in the following, in special cases, it could be also unknown by the trustee).
In the case of autonomous agents who have no external constraints conditioning their freedom of making reliance
Absence of external constraints is an ideal condition: in fact, also in case of autonomous agents some constraints are always present. In the case of the previous example, a constraint is given from the impossibility to evaluate all the potential candidates available in the world., we can say that a sufficient value
In fact, trust is also a quantitative notion: it is constituted by different ingredients to which is possible/necessary to attribute a value. of core trust is a necessary but not sufficient condition for a positive decision to trust (see Chapter 3); vice versa, a freely chosen delegation [Castelfranchi&Falcone,1998] implies the decision to trust, and the decision to trust implies a sufficient value of core trust. We will specify which beliefs and which goals characterize the trust of the Trustor (X) in another agent (the Trustee, Y) about Y's behavior/action relevant for a given result p corresponding to (or included in) the goal of X, gX. Given the overlap between trust and (mental) reliance/delegation, we need also to clarify their relationships.
Before to start with the analysis of the basic kernel, let us analyze the relationship between p and gX.
2.1.2 Goal state and side effects
As we said: gXp (the set of the results of Y’s action a contains not only X’s goal but also a set of side effects). In other words, sometimes the achievement of the goal gX is conditioned (given the specific trustee and his own features, the contextual situation, and so on) by the realization of other results in the world that could compromise other interests or goals of X.
I can trust my dentist as for being able and solve my problem with my tooth, but – while going there – I know (expect) that although he will be careful and honest I will feel some pain and will spend a lot of money. Or, I know that if I trust John – which is actually reliable and able for what I need – however he will become too familiar, he will take liberties with me, and maybe I do not like this.
So the analysis and the knowledge of these side effects is really relevant for deciding of trusting agents, and often a merely qualitative analysis is not sufficient (see Chapter 3).
2.2 Trust as mental attitude: A Belief-based and Goal-based model
Our main claim is that: only a cognitive agent can trust another agent; only an agent endowed with goals and beliefs.
First, one trusts another only relatively to a goal, i.e. for something s/he wants to achieve, that s/he desires or needs. If I don’t potentially have goals, I cannot really decide, nor care about something (“welfare”): I cannot subjectively trust somebody. These goals could also be maintenance goals (of a situation, a state), not necessarily achievement goals.
Second, trust itself basically consists of (implicit or explicit) beliefs.
The root of trust is a mental state, a complex mental attitude of an agent X towards another agent Y, in context C, about the behavior/action a relevant for the result (goal) gX.
Since Y’s action is useful to X, and X is relying on it, this means that X is delegating to Y some action/goal in her own mental plan. This is the strict relation between trust and reliance and delegation: Trust is the mental counter-part of reliance and delegation.
Given this strict relation and the foundational role of reliance and delegation (see §2.6) we need to define delegation and its levels; and to clarify also differences between reliance, delegation, and trust.
This mental attitude is based on and consists of beliefs (or in general of doxastic representations
See § 2.2.8 on Trust and Acceptance.), about the trustee and his behavior. And in fact they may be wrong. X can have false (although well grounded and subjectively justified; rational) beliefs about Y’s qualities, skills, and behavior. In short, the main beliefs are:
i) X believes that Y is able and well disposed (willing) to do the needed action;
ii) X believes that in fact Y will appropriately do the action, as she wishes.
iii) X believes that Y is not dangerous; then she will be safe in the relation with Y, and can make herself less defended and more vulnerable.
The first (and the third) family of beliefs are ‘evaluations’ about Y: to trust Y means to have a good evaluation of him. Trust implies some appraisal.
The second (and the third) family of beliefs are ‘expectations’, that is (quite firm) predictions about Y’s behavior, relevant for X’s goal: X both wishes and forecasts a given action a of Y, and exclude bad actions; she feels safe.
The basic nucleus of Trust – as a mental disposition towards Y – is a positive expectation based on a positive evaluation; plus the idea that X might need Y’s action.
Let us carefully consider these various roles of the beliefs in trust mental state, and these two facets of trust: as valuation, as expectation.
2.2.1 Trust as Positive Evaluation
An explicit positive evaluation is a judgment, a belief about the goodness of Y for the achievement of a certain goal. “Y is good” actually means “Y is good for...” [Miceli&Castelfranchi, 2000]. Sometimes we do not specify for what Y is good, just because it is included in the very concept of Y (“This is a good knife/pen/car/... mechanic/doctor/father/...”), or because it is clear in the context of the evaluation (“On which mountain we can do some climbing?” “That mountain is good!”). These are direct explicit evaluations: “Y is good, is OK, is apt, able, useful”, and so on.
The abstract content of these predicates is that: given the goal gX, Y is able to realize it or to help X in realizing it (a tool). In other terms, Y has the power of realizing gX.
The evaluation of Y is about the internal powers for gX (internal resources, capabilities, competences, intelligence, willingness, etc.), but for relying on Y for gX also external conditions and the control on the external conditions might be necessary: Y may also have (or not) the external powers, the external conditions and resources for realizing gX; Y is both ‘able’ and ‘in condition’ for realizing gX. For example, in J.J. Meyer’s logic [Meyer, 1992] Y ‘CanDo’ when both is ‘Able’ and also ‘In Condition’.
However, trust is not only a belief that Y “is good for t”, it is also a set of beliefs about the needed qualities or ‘virtues’ of Y. Trust, as evaluation, is in fact a model of Y's qualities and defects (which define his trustworthiness dimensions).
Like for any kind of explicit evaluation also with Trust we are not satisfied by the mere belief that Y is OK, is "good for", has the power to achieve goal gX, to execute the action/task a/t (delegated to him). We try to understand why Y is good for this (while Z is not); to have a theory of what makes Y able.
While we put the other dimension (the evaluation of the conditions and external resources) in the ‘environmental trust’ (see §2.5.1 and Chapter 3). In other terms, we try to know which kind of characteristics are useful or required for effectively performing t or achieving gX. And many of them are just hidden, internal (and mental): kripta (§2.2.5).
Qualities and Standards
This applies in general to the Evaluation theory [Miceli&Castelfranchi, 2000].
Given that Y is "good for" t (for example, this knife Y is good for cutting the bread), to which features of Y should this to be ascribed? In case of the knife: To its plastic or wooden handle? To its sharpening? To its color? To its being serrated and long? And so on. In several cases this implies a true causal model (although naive and commonsensical) of t, of Y, and of what effects it produces. In the example of the knife, it is clear that the plastic handle or the color are irrelevant, while a good knife for bread needs to be long, serrated, and sharpened.
Those features (F) to whom the "goodness of Y for..." is ascribed are Y's qualities (Q).
Defects (D) are those features of Y's to whom is attributed the lack of power, the inadequacy or dangerousness of Y for a/t.
<INSERT Figure 2.2 here>
Figure2.2:
From Features (F) to Qualities (Q) through Evaluation: in particular (Feature1 and Feature3) are evaluated as good for the task t while (Feature4) is good for the task t’
Notice that a feature of Y that is a ‘quality’ relative to task t, can be a ‘defect’ relative to another task t’, and vice versa.
Let us also remark how this theory of 'qualities' and 'defects' is cognitive and pragmatically very crucial and effective. In fact, while buying in a store a new knife we are not allowed to bring there a piece of bread and experimentally evaluate which knife is "good for". Thus, how can we choose a "good" knife without trying it? Just because we have a theory of the needed qualities, of what makes a knife a good knife for bread. We just look for these characteristics; we compare that knife with these 'standards' for a good bread-knife.
Standards are in fact just qualities generalized to the class of the object; the ideal properties that such a kind of object O must possess in order to be good as an O.
Qualities (and standards) are necessary not only for recognizing and choosing a good instance of object O; that is they are signs of its being good, reliable for t (trustworthy) (see § 2.2.5 on Signs); but they:
- Are fundamental also for generalizing trust from one t to another (“are for t" needed more or less the same qualities than for t' ?”), or from one agent to another: “has Z the relevant qualities remarked in Y?” (see Chapter 5); and thus they
- Are also fundamental for predictions based on general models and classes.
X trusts Y - that is, she has a positive evaluation of him for t (and she generates positive expectations on him) also because she ascribes to him certain qualities.
Then, we can say that if (see Figure 2.3):
SetQ is the set of qualities (powers) needed for a/t, then for p (gX);
In case Y possesses all the elements in SetQ then X could trust Y for achieving gX.
For any Y, the more are the elements in SetQ owned by Y the more trustworthy is Y for a/t.
The problem is the X’s beliefs about these facts.
<INSERT Figure 2.3 here>
Figure2.3:
Given a Set of Abstract Qualities for a task, how to select the right Agent for that task
So we can say that Trust is constituted of what we call 'implicit and indirect evaluations', that is of specific features (like "sharpening; long; serrated", or for a person like "tall, bold, agile, intelligent, ....") which only apparently are just descriptive predicates and usually are evaluative predicates. These are the specific and analytic beliefs on which the global positive or negative judgment is grounded.
Reason-based Trust is in fact a theory about Y, and about its qualities and defects that make him more or less apt and reliable for a/t; on this base we make predictions and build expectations about Y's performance and its outcome; and we explain the reasons of success or failure; like for any theory.
This also means that the feedback on trust due to the observed outcome of Y's performance (see Chapter 5), not only can be attributed to Y, that is to internal factors (and not to external accidents or conditions), but in some cases can be specifically ascribed to a sub-feature of this causal model of Y's power on t. And one can revise this specific piece of evaluation and trust: "Y is not so sharp as assumed"; "Y is not so agile as supposed", and so on.
Trust and Powers
An evaluation is a judgment about the possible powers of Y. In our abstract vocabulary Y can (has the power of) perform a given action or realize a given goal. It is relatively to this that Y is good for. Correspondently, qualities are just powers of (to be strong, to be smart) or conditions for powers of (to be entitled, to be prepared).
This is about not only personal powers (be strong, intelligent, skilled, etc.) but also social powers: be influent, prestigious, being friend of Z (and thus be able to obtain from Z something), being sexually appealing, etc.; and also institutional powers, the new capacities that a given institutional role gives to its players: like nominating somebody, or proclaiming, or officially signing, or to marry, etc.
See [Searle, 1995] ; As for the relationships between personal and institutional powers, see [Castelfranchi, 2003]
So there is a complex relation between powers and trust. On the one side powers - and in particular perceived and appreciated powers - make us trustworthy for the others (they can need us and wish to rely on us), and this is why our social image and the signals that we give of our qualities are so important. As we will see trust (being trusted in a given community, and in particular being better evaluated than others) is a capital (see Chapter 6).
On the other side, this is a capital precisely because increases our powers; provides us new powers. Since we (being perceived as trustworthy) are appreciated and demanded, we can enter in exchange relations, we can receive cooperation, we can influence other people, etc. This means that we enlarge very much our practical powers (the possibility to achieve - thanks to the other's cooperation - our goals) and social powers: powers on the others. Also for being invested by institutional roles (and receive new special powers) we need some competence and reliability, or better we need some trust from the others on this. In sum, powers give trustworthiness and trust from the others; and trust gives powers.
The Recursive Nature of Trust as Mental Representation
One might claim that in strict sense trust just is or implies the positive expectation, which in its turn is based on positive evaluations; but that (i) the evaluations are not trust or (ii) they are not implied by trust and they are not strictly necessary: One might trust Y without specific and explicit beliefs on his skills or reliability, just on the base of experience and learning.
We thank Fabio Paglieri for this observation. This is also one of Andrew Jones’ criticisms to our model [Jones, 2002]; see Ch. xx for our reply.
However, this is not fully true.
As for (ii), Y’s trustworthiness or the trust in Y always imply some (perhaps vague) attribution of internal qualities (such that the task could be achieved/realized); some ‘kripta’ which makes Y apt and reliable (§2.2.5).
As for (i), the evaluations of Y, when used as the bases and reasons for a trust expectation and decision, are subsumed and rewritten as aspects and forms of ‘trust’. Given that on the base of those features (qualities) X trusts Y, then X trusts those features of Y. “I trust (in) his competence”, “I trust (in) his benevolence”, “I trust (in) his morality”, “I trust (in) his punctuality”, and so on. (See also note 1 in Introduction: CONTROLLARE e LEVARE Dictionary definition).
Let us grounding this on the general theory of goals and actions, on which actually trust should be grounded. Given her goal gX (X’s aim or end) X searches for some possible action a (usually of X) able to achieve gX (where gX is included in the outcomes p, the post-conditions of a). Given an action a useful for gX, this action in order to be successfully performed requires some condition C to be true. If C holds the subject can perform a; if C does not hold it becomes the new goal g1X of X, subordinated and instrumental to gX: a sub-goal. This obviously is the essential abstract principle of planning.
Now, given that X has the goal of realizing gX, and that she is dependent on Y and needs an action aY of Y, she has the sub-goal that Y successfully performs aY. However, certain conditions are needed for both (i) Y successfully performing aY; (ii) X can decide to count on this and counts on this.
Y's valuation is precisely the epistemic activity aimed at ascertain those conditions; and the same hold for X's predictions about Y doing or not aY. As we have underlined, evaluations are about goals (something is or isn't "good for"), and predictions in trust are in fact not simple 'predictions' (beliefs about the future) but more rich 'expectations' (involving goals). Actually, since X wishes to achieve gX through Y's action, and has the goal that Y be able, in condition, and predictable in performing aY, all these necessary conditions (also those 'internal' to Y) are new (sub)goals for X: X wishes that Y is skilled enough, competent enough, not hostile or dangerous, willing and reliable, and so on. Relative to those sub-goals she evaluates Y and has (or not) trust in Y. She trusts Y for being competent, for being persistent, etc.
So trust in Y as for action aY for goal gX, (at least implicitly - but frequently explicitly) entails sub-trust supporting the broad trust about the action; in a recursive way. Since any new goal about Y might have its sub-conditions and needed sub-qualities on which X relies; thus potential sub-trusts are generated. For example, X can trust Y for being really willing to cooperate with her, because she knows and trusts in Y's friendship, or because she knows and trusts in Y's generosity and empathy, or because Y is morally obliged and she knows and trusts in Y's morality.
This is not only the real use of the word and the commonsense on trust; it is a logical and principled use, as we have just explained.
2.2.2. The crucial notion of “Goal”
For our non-reductive theory of trust (not simply epistemic but motivational and pragmatic) it is crucial to make clear the central notion of the ‘motivational’ dimension: “Goal”.
Trust attitude is not just a (grounded) belief, a prediction; is not just subjective probability of an event, because this belief structure is motivated, is charged of value, is anchored to a goal of X.
In trust X is interested, concerned; the event is a ‘favorable’ one; X’s “welfare” is involved. An “expectation” is not a “prediction” or “forecast”; when X trusts somebody this is implies a positive evaluation of him. Also affects and emotions can be involved, as the real base of a trust disposition or as complement of the judgment and decision (for example, X will not be just surprised but she will be disappointed or even feel betrayed); but, in fact, no affective reactions or emotions are possible without involved goals. Under all these crucial aspects a motivational component is presupposed in X, and makes clear “expectations”, “evaluations”, “concern” and “interest”, “affects”, etc.
Actually, when we say all these phenomena, we explain them in terms of important motives, needs, projects, desires, preferences, objectives, ... for the realization of which the agent is evaluating other agents and the possibility to rely on (exploit) them. The abstract category we use for all these motivational terms and categories is “goal”. However, it must be very clear what a “Goal” is in cognitive psychology, on the basis of the cybernetic definition and of the following psychological models.
“Goal”: what is this?
“Goal” is a perfect term but not in the most superficial and typical English use, where it usually refers to a pursued external objective to be actively reached; some step in my active plan, driving my action.
Consider for example that one of Andrew Jones’ central objections to our model of trust is precisely due to such a peculiar use of the notion of “goal” [Jones, A., 2002]: “While it is true to say that a goal-component of this sort is often present, this is by no means always so. For example, x might trust y to pay his (y’s) taxes… , even though it is not a goal of x that y pays. Also, x might trust y when y asserts that p, even though x does not have it as a goal to find out whether p is the case.” (p. 229). On the contrary, the general notion of “Goal” precisely covers also those cases. Sure that X has the goal that the other guys pay their taxes! Only in this sense he “trusts them as for paying their taxes”. Obviously, this does not mean that X is personally doing something in order to produce this result in the world; she is not actively pursuing that goal. But not all our goals are or must be personally pursued. X is wishing, desiring, prescribing, expecting, …. that the others will pay taxes. That’s why, in case, she will be not just surprised but frustrated and upset, and will blame them, and so on. Analogously, “x does not have it as a goal to find out whether p is the case”, sure! However, x has the goal (wish, prescription, ….) that y says the truth; precisely in this sense “x .. trusts y when y asserts that p”. (See also Ch. Xx § xx)
The right general notion of “goal” and of “goal-directed or driven behavior” has been provided by cybernetics and control-theory many years ago (Rosenbleuth Wiener it “Purposive Systems”), and has been imported in psychology and Cognitive Science in the '60s with the TOTE model by Miller Galanter & Pribram [Miller, Galanter and Pribram, 1960].
A “Goal” is the mental representation that 'evaluates' the world (we evaluate the world against it); if the world does not match with, we can be activated for changing the world and realize that state; but, before this, we examine if it is the case or not: is the goal self-realizing or impossible? Or: is up to us to realize it? Do we have the appropriate actions for this; are there the necessary conditions for acting successfully? Are there more important goals to be preferred? After all these tests the Goal may become our pursued objective. But, it already is a “Goal” in all the other conditions and previous steps.
It is a “Goal” even before that we believe or know that it is realized or not; before we decide that it depends on us to (attempt to) realize it; that we can and should act. Then it can become an 'intention' (if chosen, if I have a plan for it, if I'm ready to; if I have decided to act for realizing it) and be 'pursued'. (Castelfranchi and Paglieri, 2007)
In sum, a Goal is a Goal even before or without being pursued: happiness is due to goal-realization and sufferance is due to goal frustration, but not necessarily to our active successes or failures: we are crying because our mother has died, or happy because, without asking, doing or expecting nothing, she gave us a kiss.
Given this - not vague, common sense, or reductive - notion, we can make clear that:
- In Trust as 'disposition', 'attitude', evaluation of Y (before deciding to delegate or not, before delegating) Goals are not yet or necessarily pursued: I evaluate 'potential' partners or goods or services relative to a possible goal of mine, or relative to a goal of mine that I did not decided yet if I want to achieved or pursue.
Saying that X trusts Y relatively to a given (possible) goal of her, does not necessarily mean that she is actively pursuing a goal: on the one side, she can be just evaluating a potential delegation; on the other side, she can be completely passive, just waiting and expecting.
- In Trust as decision and action clearly that goal has become not only 'active' but 'pursued': I want to realize it. However, it is pursued in a strange way: thanks to the action of another agent; parasitically. It is indirectly “pursued”; perhaps I'm doing nothing, just expecting and waiting for realizing it thanks to Y's behavior.
- When I do actively and personally pursue a Goal, this is a Goal in the reductive sense, and I trust myself (self-trust, self-confidence; feeling able and competent, etc.).
This is the right, broad, scientific notion of “goal” needed for the theory of Trust.
2.2.3 Trust versus Trustworthiness
It is also fundamental to make clear the relationships between trustworthiness and trust; they are frequently mixed up; in many cases we tend to use trust (T) instead of trustworthiness (TW).
TW is a property of Y (but in relation with a potential evaluator/partner X); while T is a property of X (but in relation to Y).
On the one hand, there is an objective TW of Y (what he is actually able and willing to do in standard conditions; his actual reliability on a more or less specific task, and so on). It is just relatively to this reliability that X's T can be misplaced and X's evaluation (belief) can be wrong.
On the other hand, there is a perceived, or evaluated, or subjective TW of Y for X: XTWY.
Now, the relation between XTWY and X’s T in Y: Trust(X Y t)
We consider p=gX and do not consider here the context C. is not simple. They cannot be identified. XTWY is one of the bases of T, but the latter cannot be reduced to the former.
T is a direct function of XTWY, but not only of it. T is also a function of a factor of X's personality and general trusty disposition (see [McKnight and Chervany, 2001]). Moreover, T depends on the favorable or unfavorable ascription of the plausibility gap (lack of evidences); so the perceived XTWY is only one component of the positive evaluation (T attitude).
Of course, XTWY is even more insufficient for defining and determining T’s potential decision and T’s actual decision and act (see below and Chapter 3).
TW is multidimensional; thus also XTWY is a multidimensional evaluation and profile of Y, and cannot be collapsed in just one number or measure. The same holds for Trust. We will in fact present a multi-dimensional model of T (and TW) (§ 2.2.4).
2.2.4 Two main components: Competence versus Predictability
This (positive) evaluation of Y has different aspects and is about different qualities of Y. The most important dimensions for trust, that is, of Y’s trustworthiness, are the following ones.
Competence
Competence is the set of qualities that makes Y able for t; Y’s internal powers: skills, know how, expertise, knowledge, self-esteem and self-confidence,
In rational beings (who decide to act on the base of what they believe about the possibility of achieving the goal) there is a strange paradox of power: It is not enough ‘to be able to’; in order to really be able, having the power of, the agent must also believe (be aware) of having the ‘power of’, otherwise he will renounce, he will nor exploit his skills or resources. [Castelfrancvhi, 2003] and so on.
When X trusts Y for t, she ascribes to Y some competence.
Competence – as we claimed - cannot be put aside (like in many models and definitions; see Chapter 1) and cannot be separated from trust in Y’s reliability.
First of all, it is an important issue for rejecting the fully normative foundation of trust – pursued by many authors (like [Elster, 1979]; [Hertzberg, 1988]; [Jones, 2002]) -, which cannot cover in a simple way Y’s skills and competence.
Although this can be made simpler precisely by our theory of ‘standards’ as the needed, expected, and thus ‘prescribed’ qualities that a given instance of class O must possess in order to be a good or regular O. However, in any case, one has to distinguish ‘normative’ from ‘moral’.
Moreover, they are not fully independent dimensions. Even Y’s cooperative (adoptive) attitude towards X may require some skill and competence. For example, Y may be more or less competent and able in understanding X’s needs or in comprehending X’s interests (even beyond X’s own understanding); it is not just a matter of ‘concern’ or of good will. For example, Y’s competence, ability, expertise, can be the base for his self-confidence and sense of mastering, and this can be crucial in Y’s willingness and intention to adopt X’s goal, in Y’s persistence in this intention, which are crucial aspects of Y’s ‘reliability’. And so on.
Predictability and Willingness
The second fundamental dimension is not about Y’s potential and abstract capability of doing t, but about his actual behavior; the fact that Y is reliable, predictable, one can count on him; he not only is able to do, but will actually do the needed action.
Applied to a cognitive Y, this means that Y is willing (really has the intention to do a for gX) and persistent [Castelfranchi and Falcone, 1998a], [Falcone and Castelfranchi, 2001b]. Also in this case we have to consider on the one hand, the abstract predictability and willingness of Y not related with other elements and, on the other hand, its relevant correlations with the specific tasks and the different trustors.
These (Competence and Willingness) are the prototypical components of trust as an attitude towards Y. They will be enriched and supported by other beliefs depending on different kind of delegation and different kind of agents; however they are the real cognitive kernel of trust. As we will see later even the goal can be varied (in negative expectation and in aversive forms of trust), but not these beliefs.
Those (evaluative) beliefs are not enough; other important beliefs are necessary, especially for moving towards the decision to trust, and the intention and action of trusting.
Using Meyer, van Linder, van der Hoek et al.’s logics ([Meyer, 1992], [vanLinder, 1996]), and introducing some "ad hoc" predicate (like WillDo)
This is a simplification. Before being a belief that “Y will do” this is a belief about a potential delegation: “Y (in case) would do” “Y would be able and willing to …”, “X might rely on Y”. (See § 2.3.2) we can summarize and simplify the mental ingredients of trust as follows:
<INSERT Figure 2.4 here>
Figure 2.4:
Mental Ingredients of the “core trust”
In Figure 2.4 we simplify and summarize the kernel of trust as potential evaluation and expectation for a potential delegation.
Notice that Competence, Willingness (Predictability), and also Safety are three necessary components and Dimensions of Trust. This doesn't mean that in order to trust Y (and possibly and eventually to decide to trust him) X should necessarily have a good evaluation of Y's competence and of Y's willingness. As it will be clear later, after introducing the degree of believing as the base for the degree of trust, trust is not a yes/no object; only trust decision eventually is a yes/no choice, and clearly needs some threshold. So X's trust in Y (evaluation of trustworthiness) must be just sufficient (and frequently just in a comparative way) for risking on him. Perhaps competence is perceived as rather low, but all together the positive evaluation and expectation is enough. Of course there might also be a specific threshold for a given dimension: “no less that this”; and in this case X must focus on this and have an explicit perception or evaluation of it.
For example, we assume that we have a threshold of risk acceptance: although convenient we may refuse a choice which involves more than a certain risk. What we claim is just that (explicitly or implicitly) these dimensions about Y’s ability and know how, about his predictability and reliability, about his safety, are there, in the very disposition to trust and entrust Y.
2.2.5 Trustworthiness (and Trust) as multidimensional evaluative profiles
As we have seen, both while explaining the theory of qualities, and when analyzing the basic constituents or dimensions of trust evaluation (competence and willingness), which can be further decomposed and supported: Trustworthiness is not just a simple mono-dimensional quality. It is the resultant of several dimensions. We can for example consider two rather independent dimensions under the notion of competence: the skills or abilities of Y versus his know how (knowledge of recipes, techniques, plans for)
One can have a perfect knowledge about “how to do” something, but not be able to, since he lacks the necessary skills and abilities; vice versa, one might in principle be able to successfully perform a given activity, but lack the necessary “know how”: the instructions or recipes about how to do that.; and others rather independent dimensions around the willingness in social trust: Y’s concern and certainty of adoption versus his persistence in intending (Figure 2.5).
<INSERT Figure 2.5 here>
Figure 2.5:
Dimensions of Trust Evaluation and their sub-components
2.2.6 The inherently attributional nature of Trust
The very attitude, act, and relation of Trust of X in Y (for a given performance) implies X's causal internal attribution of the possibility of success. Trust is not simply a prediction (although it implies a prediction). In particular, it is not a prediction of a given behavior based on some observed frequency, on some estimated probability, or on some regularity and norm. It requires the grounding of such a prediction (and hope) on an internal attribution to Y.
This is why one trusts in Y, and trusts Y. Also the "trust that (event)" (that something will happily happen) cognitively and procedurally implies a "Trust in" something or somebody on which one relies, that is, the assumption that this entity will produce the desired event.
Trusting (in) Y presupposes a possibly primitive and vague "causal mental model" [Johnson-Laird, 1983] such that produces the expected result. Even when trust is non-social, when for example we decide to trust a weak chair for sitting on it, we assume that its material, or its mechanics, or its structure will resist under our weight. Trust presupposes at least a design stance (to use Dennet’s terminology
Although we believe that our theory of ‘functioning’ versus function [Miceli and Castelfranchi, 1983] be more clear. [Dennet, 1989]).
This is the deepest meaning of the competence ascribed to Y, of the internal power of appropriately executing the task; this is the real meaning of the predicate Able used for its decomposition. It is different from the other component. In order to perform the task Y must be both Able and in condition but while Able is an internal attribution, in condition can be external and contextual.
Moreover, this is also the deep meaning of the other component: willingness, intention, disposition, motivation of Y that make that prediction grounded.
In other words, in social trust the necessarily internal causal attribution of the Trust requires an intentional stance towards Y [Dennett, 1989]. Trust is a disposition (and an affect) and a social decision only possible in agents able to assume an 'intentional stance', that is to ascribe a mind and mental stuff to the other agent. The prediction of the other's behavior is based on some theory of mind or on some projection/ascription of internal factors (including moods and emotions) playing a causal role in the activation and control of the behavior.
The fact is that in humans those qualities are internal in a psychic sense; thus – by definition – they are unobservable. They can just be abduced by external signs. They are ‘kripta’ ([Bacharach and Gambetta, 19xx], [Bacharach and Gambetta, 1999]); but it is precisely in ‘kripta’ that we trust and on ‘kripta’ that we count on when we ‘trust’ somebody; not the ‘manifesta’ (visible signs). Trust is a hermeneutic and semiotic act. We trust the signs only for being reliable, for informing us about the ‘kripta’, but at the end it is of /on/for the ‘kripta’ that we trust.
‘Manifesta’, signs, are not only the direct observable behaviors or markers of Y, but also all the other sources of trust (of trust beliefs) – like experience, reputation, certificates, category, role, and so on. All of them are direct or indirect manifestations of how Y is or ‘works’, of what inside him makes him trustworthy.
Our favorite example of weak delegation (§2.6.1) makes our claim clear: I'm running to take the bus, the driver and the people cannot see me, but I see that there is people at the bus stop apparently ‘waiting for the bus’ (notice that this is just an mental ascription), on such a base - attributing them the intention to take the bus and thus the intention to stop the bus - I rely on them, I trust them for this.
While a simple prediction that somebody standing at the bus stop might probably rise the arm would be based on simple observed frequency and probability, we deny that Trust in Y as for stopping the bus is just this. It is based on the interpretation of Y as ‘waiting for the bus’ and on the ascription of the intention that will cause the action on which one relies. This is the - never clarified - gap between mere subjective probability and trust.
In Chapter 3 and 4 we will criticize the reduction of Trust to subjective probability, arguing about the importance of explicitly distinguishing internal and external components of the prediction; and also explaining how crucial and inherent for real trust are the bases (beliefs) supporting the prediction. We do agree with Williamson (see Chapter 4): if trust is just a euphemism for 'subjective probability' we do not need it, we already have a strong theory of it, and a new vague and merely evocative term is just confusing cosmetics. On the contrary, we believe that trust is a specific, well-defined, mental and social construct.
On the basis of this 'internal attribution' (or ‘design’ or ‘intentional’ stance) foundation of Trust we will be able to account for several things.
For example in §2.7.2 we argue that there is internal versus external trust and explain why it is important to differentiate them. For example, distinguishing on the web the trust of the user in the possible partners from her trust in the infrastructure or in the institution (norms, etc.). We also show that they have different dynamics; etc.
However, it is important to make clear that a given trust is internal or external only 'relative' to the assumed entity. While changing the target, again the distinction applies and again a form of internality is needed. If we consider for example our Trust in the technical infrastructure (that relative to the partner Y, was an 'external' condition), we are now necessarily doing some 'causal internal attribution' relative to its (good) working. In general: if I trust an entity E (of any kind) I'm always ascribing to E some internal properties, some virtues, on which the possible success depends, and I depend on these ‘virtues’.
This 'internal attribution' foundation of Trust explains why it is trivial and false that the failure of Y necessarily produces a decrement of X's trust in Y, and a success of Y should necessarily increase or cannot reduce X's trust in Y. The effect of the failure or of the success again depends on its 'attribution': How much and for which aspect is it ascribed to Y? For which aspect is it ascribed to external (from Y) circumstance? Only internal attribution to Y affects Trust in Y, since trust holds upon this; while an external attribution to C (say the environment, the infrastructure, etc.) obviously affects the trust in C (of course, it is really important also the relation and correlation between Y and C) (see Chapter 5 for more details).
Mistrust and diffidence (Chapter 9) are real negative forms of trust. They too entail a real internal causal attribution of the inadequate or bad behavior of Y. It is for some internal ‘virtue’ that Y is poor or harmful; there is not simply something lacking such that I do not trust Y (enough); but, positively, I attribute to Y some ‘defect’; I think something bad of him.
Trust as an external goal on Y
When X trusts Y, an external goal is put on Y [Castelfranchi, 2000c]. Moreover, Y is assumed to respond to this impinging goal:
(i) either, with an internalization of it; that is by an internal goal, copied by the external one; by ‘goal adoption’ (or goal-adhesion) (§ 2.8); of course, this is possible only if Y is an intentional agent;
(ii) or with some internal mediator of the external function; some ‘mechanism’, some ‘functioning’ satisfying/performing that function.
2.2.7 Trust, Positive Evaluation and Positive Expectation
Trust is not reducible to a Positive Evaluation
That Trust is a positive evaluation is also confirmed by the fact that expressing trust towards Y is a way of appreciating Y. To express trust in Y is an indirect/implicit positive evaluation act (even a compliment) towards him; and this can be highly appreciated, and is one of the reasons of reciprocation (Chapter 4).
However (as we will see) trust cannot be reduced to a positive evaluation. This for two main reasons. First of all, because there is much more that evaluations in trust mental attitude: there are also other kinds of beliefs, for example expectations. Second, a positive valuation about Y is not per se trust in Y, or a trust attitude towards Y. It is only a possibility and a potential for trust. Only relatively to a possible dependence of X on Y, and to a possible delegation/counting of X on Y, that evaluative beliefs become a trust attitude.
Given X's goal that Y bring it about that gXp, as a means for X's achieving gX, the beliefs about Y's qualities for this acquire the color of trust.
Good evaluations, positive beliefs, good reputation, are just potential trust. Trust also implies a (potential or actual) decision of ‘counting on’ Y, of risking something while betting on it. It cannot just be a positive evaluation where X is not concerned and involved. X can evaluate that Y has such a good quality (X likes it), or has done or even will do such a good action, but this doesn’t mean that X trusts Y (for that quality or action): X has to consider the possibility to rely on this action and quality for achieving some goal of her in the future. Trust is about the future and about potentially exploiting something (that’s why X ‘likes’ Y). (See § 2.3.2)
To be more precise, a positive evaluation of Y is not per se’ “trust”, even as simple trust attitude or disposition, because it must be considered within a possible frame. It is a matter of the ‘gestalt’ nature of complex mental states. The side of a square is a linear segment; but: is a segment the side of a square? Not per se’; only if considered, imagined, within that figure, as a component of a larger configuration that changes its meaning/role. Analogously: trust is based on and implies a positive evaluation, and when there is not yet a decision/intention it just consists in this, but only if viewed in a perspective of the potential larger mental state. Given X’s positive beliefs (evaluation) about of Y (as for something), if it would be the case X might decide to rely on Y. In this sense those evaluations are a pre-disposition to trust, a trust attitude towards Y.
The same holds for the ‘prediction’ about Y’s behavior. It is not yet trust. It is trust only as a possible “positive expectation”, that is in relation to a goal of X and in the perspective of a possible reliance on it (See below).
Decomposing a “gestalt” is not reducing it to its components.
So the correct representation of the trust “core” would be the insertion of the basic square (Fig. 2.4) within the broad potential phenomenon.
Notice that this can just be the view of an ‘observer’: I see in X a trust attitude, predisposition, potential, towards Y. Actually in X’s mind there is just a good evaluation of Y.
We can arrive to a true/full trust disposition of X towards Y if this though as been formulated in X’s mind. X not only has a positive evaluation of Y, but she has explicitly considered this as a potential, a base for a possible (non excluded) delegation to Y: “If I want I could count on Y as for...”, “If it will be the case I might rely on Y, since ....”. (For a clear distinction – on such a basis - between mere potential attitude and a real ‘disposition’ see later § 2.3.2)
This is the psychological relationship between a mere positive evaluation of Y and a positive evaluation as a trust component or basic trust attitude.
Trust as Positive Expectation
On the basis of her positive beliefs about Y’s powers and willingness (or actualization) X formulates a prediction about Y’s behavior and outcomes. This is why a lot of scholars define trust in term of an expectation. However, an expectation is not simply a prediction (or a strong prediction).
X is also concerned (it is not just a neutral predictor), is interested in the performance of a, she is waiting for (expects) some positive result; (at least) one of the outcomes of a (is considered as or) is a goal of X. As we have remarked, an aspect absolutely necessary but frequently ignored (or at least leaved implicit) is that of the goal, the need, relatively to whom and for the achievement of whom the trustor counts upon the trustee. Without this X’s goal there is no trust: the trust concept does not exist.
So trust is a positive expectation. Where a positive expectation is the combination of a goal and of a belief about the future (prediction). X in fact both believes that Y will do the action a and desires/wishes/plans so. And she both believes and wants that the goal gX will be realized (thanks to Y).
The fact that when X trusts Y, X has a positive expectation, explains why there is an important relationship between trust and hope, since also hope implies some positive expectation (although weaker and passive: it does not necessarily depend on X, X cannot do anything else to induce the desired behavior); and why trust can be ‘disappointed’. Moreover, X is “expecting”, that is, waiting and checking for something. She is concerned with her prediction; and has the goal to know whether the expected event will be really happen [Castelfranchi, 2005] [Miceli and Castelfranchi, 2002]; [Lorini and Castelfranchi, 2003]. DOVE e’ UN PEZZO LUNGO SULLE EXPECTATIONS????
Trust is not reducible to a positive expectation
Is Trust reducible to a positive expectation? For example, to the estimated subjective probability of a favorable event?
Trust as belief-structure is not just an 'expectation' (positive/favorable).
Let us put aside the fact that -as we said- Trust (at least implicitly) is trust in an agent; it is an expectation grounded on an ‘internal attribution’. Even not considering "trust that Y will ...." or "trust in Y", but just "trust that p" (for example, "I trust that tomorrow it will be sunny") there is something more than a simple positive expectation. X is not only positively predicting that p, but is 'counting on' that, it is actively concerned; X has something to do or to achieve, such that p is a useful condition for that. Moreover, such an expectation is rather sure: the perceived favorable chances are greater that the adverse ones, or the uncertainty (the plausible cases) is assumed as favorable. This is one of the differences between trust and hope; the difference between "I trust that tomorrow will be sunny" and "I hope that tomorrow will be sunny". In the second one, I'm less certain, and just "would like so"; in the first one, I am more sure about this, and that is why I (am ready to) count on this.
In fact, even non-social trust cannot be simply reduced to a favorable prediction. This is even clearer for the strict notion of ‘social trust’ (‘genuine’ trust) (§ 2.7): which is based on the expectation of adoption, not just on the prediction of a favorable behavior of Y.
2.2.8 “No danger”: Negative or Passive or Defensive Trust
As we said, in addition of Competence and Willingness, there is a third dimension in evaluating the trustworthiness of Y: Y should be perceived as not threatening, as harmless.
Either Y is benevolent towards X (for similarity, co-interest, sympathy, friendship, etc.), or there are strong internal (moral) or external (vigilance, sanctions) reasons for not harming X. This very important dimension looks missing in the definitions considered by [Castaldo, 2002]; however, it is present in others, for example from social psychology or philosophy
See for example [Hart, 1988]: trust enables us to assume “benign or at least non-hostile intentions on the part of partners in social interactions”.
Perhaps, the most ‘primitive’ and original component (nucleus) of trust (especially of implicit and affective trust) is precisely the belief or feeling: “no harm here/from…” and thus to feel safe, no alarm, no hostility, being ‘open to’.., well disposed. This is why trust usually implies no suspect, no arousal and alarm; being accessible and non-defended or diffident; and thus being relaxed. The idea of "no danger" is equivalent to “the goals of mine will not be frustrated by Y”; which - applied to animated entities (animals, humans, groups, and anthropomorphic entities) - is specified as the idea that "Y has no the goal of harming me".
We call this elementary form of trust: a-form (Negative or Passive or Defensive Trust). In a sense “feeling safe” can be the basic nucleus of Trust and be alone; seemingly without any additional component. However, looking more carefully we can identify the other core components. Clearly positive evaluations and expectations (beliefs) are there. If I don’t’ worry and do not suspect any harm from you, this means that I evaluate you positively (good for me; not to be avoided; at least harmless), since not being harmed is a goal of mine. Moreover, this feeling/belief is an expectation about you: I do not expect damage from you; which is a passive, weak form of positive expectation. Perhaps I do not expect that you actively realize an achievement goal of mine; but I at least expect that you do not compromise a maintenance goal of mine: to continue to have what I have.
It is rather strange that this basic meaning of ‘trusting’ and this component is not captured in those definitions (except indirectly, for example, with the term ‘confidence’
The English term ‘confident’/’confidence’ seems mainly capture this nucleus of Trust. , or marginally). This is for us the most ‘primitive’ and basic nucleus of trust, even before relying on Y for the active realization of a goal of X; just passively relying on Y as for not being hostile or dangerous, non harming X.
Of course there is a stronger, richer, and more complete form of trust (b-form: that we call ‘active’, ‘positive’, ‘achievement’ trust) not only due to the idea/feeling (expectation) that the other will not harm me (lack of negative expectations); but including specific positive expectations: the idea/feeling that the other will 'adopt' (some of) my achievement goals, will be helpful; that Y's attitude is 'benevolent' in the sense that not only is not hostile, noxious or indifferent, but that he can be disposed to adopt and realize my goal (or at least that Y can be useful for achieving my goals). I can count on Y, and make myself dependent on Y for realizing (some of my) goals.
In terms of the theory of 'interference' (the basic notion founding social relations and action [Castelfranchi, 1998], a-form is the assumption or feeling that "there can/will not be negative interferences" from/by Y's side; while b-form is the assumption or the feeling that "there can/will be positive interferences from/by Y" (Where 'by' means 'on purpose': goal-oriented or intentional).
To be 'full' and complete trust should contains both ideas (a-form) and (b-form); but this is not always true. Sometimes it is more elementary and just limited to (a-form); sometimes it is mainly (b-form).
In a sense – as we have said – a-form is always present, at least implicitly; also when there is the richer and more specific attitude of b-form.
In fact, when applied to the same set of goals (b-form) implies (a-form):
if Y will be useful or even helpful for the achievement of the goal gX, he is not a problem, a threat for the goal gX.
To favor goal gX implies to not harm goal gX, since the achievement of goal gX implies the non frustration of goal gX.
However, this implication does not mean that when X Trusts Y (as capable and responsible for certain goals) X is always and fully relaxed and has nothing to worry from Y. In fact: what about possible other goals of X?
Except when applied to the same sub-set of goals or when ‘generalized’ (i.e., applied to all X’s possible goals) b-form in fact doesn't necessary imply a-form. If trust is relative to a sub-set of X’s goals, it is perfectly possible that X trusts Y (in the sense of b-form) for certain goals, but X could worry about others goals of her; or, vice versa, that X trusts Y (in the sense of a-form) which is unwarlike e not threatening, but X cannot trusts (in the sense of b-form) him as able and helpful for some goal of her. Thus, b-form and a-form doesn’t necessarily co-occur, except for the same sub-set of goals.
2.2.9 Weakening the belief-base: Implicit Beliefs, Acceptances, and Trust by-default
To make things simpler, we assume in our model that trust is composed by and based on ‘beliefs’. However, this is an antonomasia: trust is based on doxastic attitudes: beliefs, knowledge, but also just acceptances (in our vocabulary: assumptions). Beliefs are assumed to be true in the world; to match with the (current, future, previous) world, if/when tested; or at least they are produced with this function and perspective. But we also have different and weaker doxastic attitudes on mental representations; or better different functions and uses of them.
For example, a very important function for the theory of purposive behavior (and for the theory of Trust) is the use of doxastic representations as conditions for actions and decision. In order to decide to act and to act (and for choosing an action) some conditions must be true, or better: they must be assumed (but not necessarily verified or proved). These are assumptions. We can use beliefs as assumptions; but they can also be not real beliefs.
We can base our actions or reasoning on simple or mere ‘assumptions’ (non-belief assumptions), which have not to be tested or are not destined to be tested. They are just -implicitly and automatically or explicitly- ‘given for granted’ or used ‘as if’. Only the success of the practical action based on them, will provide an unconscious and indirect feedback about their ‘truth’; will ‘confirm’ and indirectly ‘verify’ them. It is important to distinguish between mere-assumptions and beliefs because one cannot decide to believe something while one can decide to assume something [Cohen, 1992] [Engel, 1998].
This is very relevant also for trust because sometimes we trust Y not on the basis of real - more or less certain - beliefs (based on experience or on inference), but just assuming something about Y, and acting ‘as if’. It is even possible to explicitly ‘decide’ to trust Y. We do not have sufficient evidences; current evidences do not provide us enough certainty (‘degree of trust’) for trusting Y, but we can overcome this situation not by waiting or searching for additional information and evidences, but just taking our decision: “I have decided to trust you, (although...)”. This does not only mean that I have decided to rely on you (trust as act), but it can also mean that I’m assuming something about you (expectations, evaluations), and that I’m testing you out: precisely those assumptions will be confirmed or invalidated by your behavior (see § 2.4).
Trust beliefs obviously can be just implicit; not explicitly represented and considered in X's mind. They can be just presupposed as logical conditions or logical entailment of some (explicit) belief.
Suppose, for example, that X has such an evaluation about Y: " Y is a very good medical doctor" and that this evaluation comes from Z's recommendation (and X trusts very much Z), or from her direct practical experience. In this evaluation X implicitly assumes that Y is well prepared (competent on the subject), and also technically able to apply this doctrine, and also reliable in interaction, he takes care of you and of your problems. All these evaluations, or these possible pieces of X’s trust in Y (the fact that X trusts Y for being prepared, for taking care of, etc.; the fact that X trusts in Y's expertise, attention, etc.), are just implicit beliefs; not necessarily explicitly derived and written in X’s mind (just 'potential' [Levesque, 1984; Castefranchi, 1996; 1997]), and/or not explicitly focused and taken into account.
It is also important to remark that there are many forms of trust which are not based on such an explicit and reason-based (argumentative) process we have presented in previous sections. They are rather automatic; not real 'decisions' or 'deliberations'. These are forms of Trust 'choices' and 'acts' based on some 'default rules' (positive evaluations are just implicit). The rule is: "Except you have specific signals and specific reasons for do not trusting Y and rely on him, trust him".
So the lack of distrust is the condition for trusting more than the explicit presence of trust evaluations (see Table 2.1).
Table 2.1:
By-default Trust
IF
Not (Believe X Not (Trustworthy Y t C))
THEN
Trust (X Y C t gX) will be “naturally” over the threshold for delegating
These form of automatic or by-default trust are very important not only for characterizing generalized dispositions of people or affective trust attitudes but also in other domains.
For example, trust in our own natural information sources (our memory, our eyes, our reasoning,..) and frequently even in social information sources is suspect-less, is automatic, by-default, and doesn't need additional justification and meta-beliefs. [Castelfranchi, 1997].
Moreover, to 'believe' is in general not a real 'decision', but for sure is the result of some computation process based on sources, evidences, and data. But - in a sense - to 'come to believe' is an act of Trust:
Trust in the belief (you rely on it and implicitly assume that it is valid, true), and
Implicitly trust in the sources and bases of the belief, and
(Just procedurally and fully implicitly) even trust in the machinery or 'procedure' which outputs the belief.
2.3 From Disposition to Action
As we said, trust cannot be just limited to a (positive) evaluation, an esteem of Y, and to a potential disposition to relying on him. This potential can become an act. On the basis of such a valuation and expectation, X can decide to entrust Y within a given ‘task’, that is to achieve a given goal thank to Y’s competent action. ‘To trust’ is also a decision and an action.
The decision to trust is the decision to depend on another guy for achieving our own goals; the free intention to rely on the other, to entrust the other for our welfare. However, to pass from a mere potential evaluation to a reliance disposition, that is, to the beliefs supporting the decision and the act to rely upon Y, the kernel ingredients we just identified are not enough.
At least a third belief (a part from that on being safe) is necessary for this: a Dependence Belief.
In order to trust Y and delegate to him, X believes that either X needs him, X depends on him (strong dependence, or at least that it is better to X to rely than do not rely on Y (weak dependence).
In other terms, when X trusts someone:
- X has an active goal (not just a potential one; see §2.3.1), and
- X is personally and not neutrally ‘evaluating’ Y; moreover,
- X is in a strategic situation [Deutsch, 1985]: X believes that there is ‘interference’ [Castelfranchi, 1998] and that her rewards, the results of her projects, depend on the actions of another agent Y.
Indicating better we have to say that:
Strong dependence [Sichman et al., 1994], is when X is not able at all to achieve her goal; she lacks skills or (internal or external) resources, while Y is able and in condition to realize her goal. Y’s action is a necessity for X.
Weak dependence [Jennings, 1993], is when X would be able to realize her goal; however, she prefers to delegate to Y, to depend on Y. This actually means that X is strongly dependent on Y and needs Y for a broader outcome, which includes her original goal plus some additional advantage (like less effort, higher quality, etc.). This is why she prefers and decides to delegate to Y. In other words, X is reformulating her goal (which includes the original one: G0), and, relative to this new goal, she is strongly dependent on Y. Then she formulates the instrumental sub-goal (plan) about “Y doing the action a”, and – of course – also for this goal she strictly depends on Y.
These Dependence beliefs (plus the goal gX) characterize X’s "trusting Y" or "trust in Y"
We are stressing now the internal attribution of trust and putting aside for the moment the external circumstances of the action (opportunities, obstacles, etc.). We will better analyze this important distinction in 2.8 about social trust, and in Chapter 3 on Decision. in delegation.
However, another crucial belief arises in X’s mental state -supported and implied by the previous ones- the Fulfillment Belief: X believes that gX will be achieved and p will be true (thanks to Y in this case). This is the "trust that" gX (case in which gX =p).
That is, X’s trust in Y about task t, for goal that p, implies also some trust that p will be realized (thanks to Y).
Is § 2.3.1 we will be more precise and complete on the relationships between “Trust that” and “Trust in”.
When X decides to trust, X has also the new goal that Y performs a, and X relies on Y's a in her plan (delegation) (about this decision see Chapter 3). In other words, on the basis of those beliefs about Y, X "leans against", "count on", "depends upon", "relies on"; X practically "trusts" Y. Where, notice, "to trust" does not only means those basic beliefs (the core: see Figure 2.4) but also the decision (the broad mental state) and the act of delegating.
To be more explicit: on the basis of those beliefs about Y, X decides of not renouncing to gX, not personally bringing it about, not searching for alternatives to Y, and to pursue gX through Y.
This decision is the second crucial component of the mental state of trust: let us call this part reliance trust (we called the first part core trust) and the whole picture mental state of trust and the delegation behavior.
Also using again Meyer, van Linder, van der Hoek et al.’s logics ([Meyer, 1992], [vanLinder, 1996]), we can summarize and simplify the mental ingredients of trust as in Figure 2.6.
<INSERT Figure 2.6 here>
Figure 2.6:
Mental State of the Decision to Trust
Of course, there is a coherence relation between these two aspects of trust (core and reliance): the decision of betting and wagering on Y is grounded on and justified by these beliefs. More than this: the degree or strength (see Chapter 3) of trust must be sufficient to decide to rely and bet on Y ([Marsh, 1994], [Snijders, 1996]). The trustful beliefs about Y (core) are the presupposition of the act of trusting Y.
2.3.1 Trust that and Trust in
As we saw both analyzing the fulfillment belief and the attributional nature of trust, there are two notions of trust which are orthogonal to the other fundamental distinction between trust as mere attitude (beliefs) and trust as decision and action. We refer to the distinction between Trust that and Trust in.
Trust that is our trust in a given desired state of the world, event, outcome; while Trust in is the trust in a given agent for doing something: in sentences like “I trust John (for that/doing that)”, “I have/feel/put trust in John (for that/doing that)”, “I entrust John within that”, etc.). Necessary and systematic relationships have emerged between these two forms of trust: they imply each other.
On the one side, Trust in Y (both as an evaluation and potential expectation and decision, and as an actual decision and action) necessarily entails the Trust that the action will be performed, and –thus– the trust that p will be true, the goal gX will be realized. In other words, the “Trust in Y” is just the trust that Y will “bring it about that p”.
Even more analytically: the trust in Y’s qualities: competence, willingness, etc. entails the trust in Y as for doing a, which entails that a will be correctly performed, which entails that the goal will be realized (p will be true).
But, on the other side, as we said, any trust that a given event will happen, as trust, is more that a mere quite firm and positive expectation, not only because X counts on this (she is betting on this for achieving something) but because it is based on the idea of some (although vague) active process realizing that result.
Any Trust that presupposes some Trust in some natural or social agent. Any "Trust that p will be true" (not just hope, not simply an expectation) presupposes a Trust that some Y will bring it about that p.
Our favorite example is Lula (the president of Brazil) interviewed before the vote with other candidates. While other candidates were saying "I wish" "I hope" or even "I'm sure to win", Lula’s response was: "I trust to win". The difference with the other sentences in not only about the degree of certainty ("hope" and "wish" are weaker or less specified) or about the positive expectation ("sure" has a high degree but might be also negative); the difference is that while saying "I trust that the result will be this" Lula is implicitly saying: "I trust people, the voters" "I trust my party", "I trust myself as for being able to convince and attract".
If I conceptualize the relation as ‘trust’, I implicitly assume that the result depends on some entity or process (some agency) and the trust is “in” such an entity, which (I trust that) will bring about that p. So we can introduce a new operator TRUST-That, with just two explicit arguments (X and p) and show how it can be translated in the former TRUST operator:
TRUST-that (X p) implies TRUST (X Y C t=(a,p)) (2.2)
where X believes that there is any active Y (maybe he is unknown to X) that will bring it about that p through a behaviour/action/task (maybe they are unknown to X too). In any case, X trusting the final achievement p, trusts also the agent performing the task.
We claim that this is true even with natural agents, in sentences like: "I trust that it will rain"; there is something more than in “I’m sure that...”, “I hope that....”; there is an implicit trust in some vague causal process that will produce that "it will rain".
On the other hand, as seen above:
TRUST (X Y C t=(a,p)) implies TRUST-that (X p) (2.3)
The relation between “Trust in” Y and in his action and Trust that p is just one and the same relation (or better reflects the same basic action-theory relation) between the Intention to do a given action and the Intention that a given result holds: when I have the Intention that something hold, this necessarily implies that I believe that it depends on me, I have to act, and that I Intend to do something; vice versa, if I have the Intention to do an action, for sure is in order to achieve a given goal/result; thus I also Intend that this result be realized. When all this is not up to me, but it is delegated to another agent Y, we get all these forms of Trust that (outcome and performance) and the prerequisites of the trust in the agent, his virtues, and the virtues of his action.
In sum, we can say that:
TRUST (X Y C t=(a,p)) <==> TRUST-that (X p) (2.4)
where <==> means “implies and is implied by”.
There is a bidirectional relationship between Trust-that and Trust-in.
2.3.2 Trust pre-disposition and disposition: from potential to actual trust
Frequently in Chapter 1 and in this chapter (and also later) for the sake of simplicity, we have been identifying and collapsing the notion of Trust ‘attitude’ and ‘disposition’. However, one might distinguish between the mere beliefs about Y (evaluations and expectations) and a true ‘disposition’, which is something more than an evaluation, but it is also something less than the actual decision and act. It is something in between; and preliminary to the actual decision.
Trust disposition
We will not consider the other notion of ‘disposition’ relevant for Trust. The idea of a personality trait, or of mood, which make us generally open, well disposed, trustful towards the others; and increase the probability that we trust Y. This is the notion traditional in social psychology, and used – for example – in McKnight’s model [McKnight and Chervany, 2001]. is the potential decision to trust, or better the decision to possibly (en)trust Y. Not only X evaluates Y, but she also perceives this evaluation as sufficient for (if/when needed) trusting Y and relying on him. "If it would be the case/ when there will be the opportunity... I will trust Y", “One/I might trust Y”. X is disposed to trust Y (if/when it will be the case).
In Trust disposition in strict sense the expectations are also conditional or potential (not actual); X has not the actual goal that Y does a given action and realizes a given result. X only has the prediction that “if she would have the goal that p, Y will/would realize it; she might successfully rely on Y”. X is not actually concerned, actually waiting for something (expecting); she is only potentially expecting this, because she only potentially has the goal. Only this potential reliance actually makes the mere evaluation a possible expectation, a trust attitude (2.2.6).
This specific mental attitude (Trust disposition) is in fact very important also for both selecting and ordinating the preferences of the trustor; they contribute to decide in which context and environment she has to situate herself in the (near or far) future. If Mary can decide to live in different environments where different goals of her will be supported (she believes so) by different (artificial, human, institutional, and so on) trustees, she will decide to live in one in which her more relevant (with higher priority) potential goals will be better supported.
So we can say that Trust disposition is in fact a real very important regulator of agents’ behavior and goal selection.
Even the decision to trust can be conditional or hypothetical; I have already decided that, I have the future directed Intention: “(if it will be the case) to address myself to Y, to trust, to rely on Y”. I have already decided, but, actually, I am not trusting him (as decision in action and act).
So, Figure 2.1, on trust stages and layers, should be even more articulated: in a potential trust attitude ((pre)disposition) versus a richer trust attitude (disposition) contained in a trust decision.
We have to sophisticate a bit our analysis of the cognitive attitude of trust, by explaining that such a nucleus evolves in fact from its preliminary stage (before the decision) to its inclusion in the decision.
It is important to realize that the disposition kernel of the decision, intention, and action of (en)trusting includes or presupposes the first kind and nucleus of trust that we have just characterized (evaluation, prediction,..) but is broader, or better it is the actualization of it.
We pass from a potential evaluation/expectation to an actual one. There is a difference between the mere preliminary and potential judgment “One can trust Y”, “Y is trustworthy”, and the executive prediction that Y will actually (in that circumstance) do as expected and will realize the goal.
X passes from the (a and b) beliefs (2.5 and 2.6) to the additional and derived belief (2.7):
BelX <CanY (a)>p (2.5)
that means: X believes that Y is able and in condition to do a and the result of this action would be p true in the world (which is a positive evaluation of Y and of the context); and
BelX <(q DoY(a))>p (2.6)
that means: X believes that there is a condition q able to activate the performance a of Y; if it will be the case, Y will do a
Where q is something like: “if X will need his help”; “if X will make the request”, “if it will happen that….”, and so on. (a prediction and just a ‘potential’ expectation, since not necessarily X, while evaluating Y relatively to the goal resulting from a, do currently have such a goal [Miceli&Castelfranchi,2000].
BelX <Will-DoY (a)>p (2.7)
that means: X believes that Y will do the action and will achieve p (which is in fact combined with the active goal that DoY(a) and thus is a real expectation), and also contains the expectation of a and of its desired outcome p (the goal X is relying on Y for).
We call the formulas (2.5) and (2.6) the potential evaluation and mental attitude towards Y: trust pre-disposition; and the mental attitude towards Y and (2.7) in the decision to rely on him: trust disposition.
2.3.3 The decision and act of trust implies the decision to rely on
Let us now come back to the relations between trust and reliance. Consider Holton’s very nice example of the drama course [Holton, 1994] (p. 63-64): “If you have ever taken a drama course, you have probably played this game. You are blindfolded. You stand in the middle of a circle formed by the others. They turn you round till you lose your bearings. And then, with your arms by your sides and your legs straight, you let yourself fall. You let yourself fall because the others will catch you. Or at least that is what they told you they would do. You do not know that they will. You let yourself fall because you trust them to catch you”.
We would like just to add to Holton analysis a more subtle distinction.
To decide to let you fall down is not the same of deciding to trust them. You can decide to let you fall down even if you do not trust them at all; you believe that they want to play a trick and to make fun of you, and you are ready to protect yourself at the last moment. If you decide to trust you not only decide to let you fall, but you decide to count on them, to act assuming that they will catch you. You decide to do your part of the plan relying on them doing their part of the plan. (Moreover – in ‘genuine’ social trust - you would count on them because you count on their motivations and their social-adoptive attitude towards you, and – following Holton – assuming a ‘participant stance’ towards them (p.66) as persons treating you as a person).
Deciding to attempt, to try and see, is not deciding to rely/count on, and this is necessary in ‘deciding to trust’; although also to decide to rely on is not enough for deciding to trust.
For many authors ‘trust’ is only social (and in a deep sense of ‘social’, §2.7); and they try to disentangle ‘trust’ from ‘reliance’ just on such a base. See again [Holton, 1994] (p. 68): “I have reason for simple reliance on an object if I need something done and reliance on it is my best bet for getting it done; likewise for simple reliance on a person. But in cases which involve not just simple reliance but trust, my reasons can be more complicated. Just because trust involves moving to a participant stance, I can have further reasons to trust, since that move can itself be something I value. Suppose we are rock climbing together. I have a choice between taking your hand, or taking the rope. I might think each equally reliable; but I can have a reason for taking your hand that I do not have for taking the rope. In taking your hand, I trust you; in so doing our relationship moves a little further forward.(..)”
The text continues like this: “.... This can itself be something I value. We need not imagine that you would be hurt if I chose the rope over your hand; you might be perfectly understanding of the needs of the neophyte climber. But our relationship would not progress.” This is a nice issue: why the act of trusting Y creates or improves a positive relationship with him? We examine this issue in Chapter 5.
In that Holton seems very closed to [Baier, 1986], which claims that trust must be distinguished from mere reliance, because it is a special kind of reliance: reliance on a person’s goodwill towards me. (p.234)
Baier’s claim is based on the example of some safety and reliance assured by threats and intimidation on Y. If I count on Y’s fear of me or of my bodyguards, or on their protection, I do not really ‘trust’ Y. We disagree on this, because we claim that there are different kinds and layers of social trust; the one based on ‘good will’ or benevolence is only a sub-case of the one based on goal-adoption towards me for whatever reason (even avoiding sanctions or revenge) (See § 2.8).
We agree that trust must be distinguished from mere reliance, but in our view, the real distinction is not directly based on ‘sociality’: intentional stance, or the richer ‘participant stance’, good will, or moral stuff. There is a preliminary distinction, before arriving to the special form of ‘genuine’ trust (§ 2.8).
In natural language, I can ‘trust’ even the rope or the rock, but this is more than just ‘relying on’ it or deciding to grasp it.
Trust is (conscious and free, deliberated) reliance based on a judgment, on an evaluation of Y’s virtues, and some explicit or entailed prediction/expectation: “How could you trust that rock? It was clearly so friable!” “No, I have tested it before; I evaluated it and I was convinced that it would have supported me!”.
What is Reliance?
As showed in §1.3, in any (intentional, external) action a there is one part of the causal process triggered by the action and necessary for producing the result/goal of the action and defining it which is beyond the direct executive control of the Agent (Ag) of a. In performing a, Ag is making reliance on these processes and this is true in both cases:
- if Ag knows this, models this in his mind, and expects this;
- if Ag doesn't understand the process, is not aware of it, or at least doesn't explicitly represent it in his plan (although he would be able to do so).
As we said (§3.1), in the first case reliance becomes delegation. "Delegation" would be the subjective and chosen reliance. Counting upon: conceiving in Ag’s individual mind a multi-agent plan including (planning, expecting) the action of another autonomous agent.
In the second case we have pure Reliance.
In Delegation (at least) one part of the delegator’s subjective plan for successfully accomplishing the intentional act a and achieve its goal, is 'allocated' to another agent either natural (like the sun in bronzing; or a coffee for being awake) or social (like a waiter for bringing food).
Let us clarify the concept: Ag is making reliance upon Y/P (where Y is another agent and P is a process) when: there are actions (or inactions) in Ag’s plan which are based on/rely upon Y/P, which depend on it for their efficacy (in other words: that process P due to Agent Y creates some conditions for the performance or for the efficacy of those actions), and Ag decides to perform those actions or directly performs them, Ag invests on Y/P (costs), Ag risks, Ag is relying on the fact that P will actually happen.
P (due to Y) is a necessary process/condition for the achievement of Ag’s goal, but it is not sufficient: Ag has to do (or abstain from doing) something, and thus Ag has to decide something: whether counting on Y/P or not, whether investing on it; Ag has to take her own decision of exploiting it or not.
‘Delegation’ requires some trust, and trust as free decision and action is about delegation. This also means that trust implies that X has not complete power and control over the agent/process Y, he is relying and counting upon. Trust is a case of limited power, of ‘dependence’.
When Y is an autonomous cognitive agent this perceived degree of freedom and autonomy consists in its ‘choice’: Y can decide to do or not to do the expected action. With this kind of agents (social Trust) we in fact trust Y for deciding, be willing, to do -against possible conflicting goals at the very moment and in the circumstance of the performance- what Y ‘has to’ do (for us); we trust (in) Y’s motivation, decision, and intention.
This feature of trust strictly derives from our founding trust on reliance on a non-directly controlled process and agent, on the perception of this ‘non-complete control', and risk; on the distinction between trust ‘in’ Y, and global trust (internal attribution); on the idea of ‘delegation’: of deciding to counting upon such a process/agent. If I do not decide to depend on this, I do not care of its non-controllability.
Reliance is a much broader phenomenon than trust. It even covers cases where the agent is unaware of the needed mediation. Let us consider the various cases and degrees before trust.
a) X does not understand or know to rely on a given agent or process. However, the positive result -due also to Y’s action- reinforces and reproduces X’s behavior and his reliance on Y
While walking actually I’m implicitly and unconsciously relying on the floor; until I do not have some bad surprise.. We can call this ‘confidence’.
b) X is aware of the contribution of Y, but he doesn’t decide of relying on Y; it is just so. It is, for example, when I just become aware of my confidence and reliance in a given support; I realize that it is only thanks to Y (that physical support? The obscure work of that guy?) that my activity was possible and effective.
c) X decides to rely on Y (not necessarily because he trusts Y; even without trusting Y; for example it is obliged to).
d) X decides to count on Y, but Y is not an autonomous agent; Y doesn’t decide ‘to do’ what X needs (for example, I rely on the fact that –after this– she will be tired; or I decide to rely/bet on the fact that tomorrow it will be sunny).
e) X decides to rely on Y because he trusts Y (autonomous agent), but X does not rely on Y’s adoption of his goal (not ‘genuine’ trust).
The 'act' of trust is not reducible to reliance; "to trust" (as act) implies 'counting on', which implies 'to rely on', but is more than this.
'Counting on' is not just relying, it is first of all a (originally) conscious reliance; the agent knows to rely on a given process and entity. Moreover, this reliance is not simply the discover of a state of fact, of a static given situation; it is a decision and the result of a decision, or at least a course of events, something that X expects that will happen while 'doing' something. X is doing something (or deciding to do something) and she expects that this process will bring to a good/desired result thanks to the action of another entity, that will create some necessary condition for the successful realization of the act or of the goal.
Counting on means to have in mind a multi-agent plan, where the action (the contribution) of Y is enclosed; and where X has to do her share, at least deciding to counting on, or deciding to do nothing (which is an action) and delegating and waiting for the result, or at least expecting for.
In other words: Counting on it is not just to ‘delegate’, but is to do my share since and until I assume that Y will do his own share. Delegating is (deciding to) allocate/assign an action to Y in order – then – to count on this. They are two complementary moves and attitudes; two faces of the same complex relation.
Trust (as act) of course is not just counting on, it is counting on based on a good evaluation (of Y’s capacity and predictability) and on a good expectation. Moreover, to count on may be weaker – as degree of certainty – than to trust, or less ‘free’; trust in an autonomous decision based on an internal evaluation or feeling. One might ‘count on’ something even when pushed, obliged to do so, without really trusting it, and without a free decision.
Suppose that you don’t trust at all a drunk guy as a driver, but you are forced by his gun to let he drive your car. If X’s choice was free she wouldn’t count on Y, precisely because she does not trust Y (enough).
Delegation is the free act of counting on; and precisely for this it normally presupposes some trust. So, Trust includes 'counting on' (but is not just reducible to it) which includes 'relying on'.
<INSERT Figure 2.7 here>
Figure 2.7:
Relationships among Reliance, Counting on, Delegation, and Trust
2.4 Can we decide to Trust ?
In the next chapter we will model the decision process and how trust enters into it. However, in relation to our layered notion of trust, and to a belief-based account of trust, a crucial preliminary question arises. Can one decide to Trust ? For example, Annette Baier in her work on “Trust and Antitrust” [Baier, 1986] claimed that we can never decide to trust (p. 235). We disagree on this.
First of all, the question is not well addressed without immediately distinguishing between Trust as belief-structure (attitude/disposition) and Trust as decision and action.
(i) As for Trust-act, the answer is ‘yes’: I can decide to trust somebody, and this even in case that the background belief-trust wouldn't be sufficient. I can in fact perfectly say “I have decided to trust him (to entrust him), although I do not really trust him”. I may in fact have other reasons and other values such that (even with a too high perceived risk) I accept that (‘as if’) Y be reliable and I decide to delegate and relying on Y and to be vulnerable by Y. For example, I might do so in order to see whether I am right or wrong, whether Y is really untrustworthy; or in order to show my good-faith and good-will; and I am disposed to pay such a price for having this proof or for exhibiting my virtues.
Consider that in this case (trust-act with not sufficient belief-trust) the trustor’s mental elements will be conditioned by this act: for example, X will have, after the act and caused by it, the goal that Y is able and willing to achieve the task.
(ii) As for Trust as beliefs about Y (evaluations and expectations), like for any other belief I cannot decide about.
However, - as we have already seen in §2.2.8- those assumptions might not necessarily and always be true/full beliefs, but just acceptances. I act on the presumption that, since I do not really know/assume that it is true; I do not know, but it might be.
This is another meaning of the expression: “I have decided to trust him”: “I have decided to give for credible, to give credit”; and on such a base I have decided to rely on him.
The two decisions are not fully independent. On the one side, obviously, if I can decide to believe (assume) that you are trustworthy, then – on such a base – I can decide to trust you, as action of reliance. But there is also a strange relation in the other sense. Following Festinger’s model of ‘cognitive dissonance’ reduction [Festinger, 1957], after a decision automatically and unconsciously we adjust the strength and the value of our beliefs on which the decision is based, in order to feel consonant and coherent (probably the function is to make our intentions more stable). So we increase the value of the beliefs favorable to our preference and choice, and make weaker the contrasting beliefs (focused on costs, risks, and the value of alternative choices). In term of trust, after the decision to trust Y (as intention and action) we will adjust the attitude/disposition beliefs (evaluations and expectations about Y). So, the decision to believe obviously affects the decision to trust (counting on), but also the decision to trust may affect (feed back on) the beliefs.
Taking into account explicitly and consciously this effect in our decision-making would be irrational, since Y’s trustworthiness is not enhanced at all; only our subjective perception of it is enhanced. We can be (after the choice) less anxious, but not more safe.
We can take into account this expected effect in our decision, not as an increased reliability of Y, but as a good outcome of the decision, to be evaluated. This is quite different from the other prediction about Y’s increased trustworthiness due to our action to trust him, as an actual effect on him (see Chapter 5).
2.5 Risk, Investment and Bet
Any act of trusting and counting on implies some bet and some risk. Trust is there precisely because the world is uncertain and risky [Luhmann, 1979]. In fact, X is basing her behavior on uncertain expectations and predictions; and is making herself dependent of Y, and thus exposed to be vulnerable by Y. Also because Y is not fully under X’s control; especially when he is an autonomous agent, with his own mind and interest.
Trust is “a device for coping with the freedom of others” [Gambetta, 1988] (p.219), or better with their “autonomy” X might eventually be disappointed, deceived and betrayed by Y: her beliefs may be wrong. At the same time X bets something on Y.
First, X renounced to (search for) possible alternatives (for example, other partners) and X might have lost her opportunity: thus X is risking on Y the utility of her goal gX (and of her whole plan).
Second, X had some cost in evaluating Y, in waiting for its actions, etc. and X wasted her own time and resources.
Third, perhaps X had some cost to induce Y to do what X wants or to have him at her disposal (for example, X has paid for Y or for his service); now this investment is a real bet [Deutsch, 1985] on Y.
Thus, to be precise we can say that:
With the decision to trust Y (in the sense of relying on Y), X makes herself both more dependent and more vulnerable, and is more exposed to risks and harms by Y.
When X trusts Y there are three risks:
a) the risk of failure, the frustration of gX (missed gains) (possibly for ever, and possibly of the entire plan containing gX);
Moreover there might be not only the frustration of gX, the missed gain, but there might be additional damages as effect of failure, negative side effects: the risks in case of failure are not the simple counterpart of gains in case of success.
b) the risk of wasting efforts and investments (losses);
c) the risk of unexpected harms (frustration of others goals and/or interests of X).
As for the first risk (case a), it is important the increment of X's dependence from Y.
Two typical cases are the dependence from time resources and trusted agents resources. May be that after Y’s failure there is no time for achieving gX (it has a specific time expiration); may be that initially X might have alternatives to Y (rely on Z or W) after her choice (and perhaps because of this choice) Z and W might be no more at her disposal (for example they might be busy); this means that X's alternatives means (partners) for gX are reduced and then X 's dependence on Y has increased [Sichman et al., 1994].
Given those (in part perceived) risks and thus the explicit or implicit additional goals of avoiding these harms, X becomes – relative to these additional goals – more “dependent” on Y, since actually it is up to Y (after X decision to trust him and relying on him) do not cause those harms to X.
As for becoming more vulnerable (case c), since X expects some help from Y (as for goal gX) X feels well disposed towards Y. The (implicit) idea that there is no danger from Y (as for gX), reduces X’s diffidence and vigilance; X feels confident towards Y, and this generalizes beyond gX. This makes X – less suspicious and careful - more accessible and undefended.
This is also due to a bit of transitivity in Positive Trust from one goal to other: if X trusts Y for gX, X can be a bit prone to trust Y as for a different goal g’X (where g’X is different from gX) even if we have to consider all the limits of the transitivity applied to the trust concept (see Chapter 5).
In sum, Y – after and because of X’s trust (attitude + act) in him – can harm X in several ways:
- By frustrating X’s goal (gX) for which X relies on him. This also implies frustrating X’s expectations and hope: that is disappointing X. And this will impact on X’s self-esteem (as evaluator and/or decision maker). Y can also damage X’s general attitude and feeling towards the world; and so on [Miceli&Castelfranchi, 1997]. Moreover, X may, not only, be surprised, frustrated, disappointed, but X can feel resentment (and even indignation) for moral violations, for been betrayed (see Figure 2.8).
<INSERT Figure 2.8 here>
Figure 2.8:
Potential Effects for the Failure of the Different Trust Elements
- By frustrating other goals of X that she did not protect and defend from Y’s possible attack, being relaxed and non-diffident. This will imply analogous consequent frustrations.
It was necessary to immediately mention uncertainty and risk, in connection with the notions of reliance and expectations. However, we will deeply develop these issues (Chapter 3), after introducing the degree of certainty of beliefs, and the degree of trust in decision-making.
2.6 Trust and Delegation
What delegation is
As we said, in Delegation the delegating agent (X) needs or likes an action of the delegated agent (Y) and includes it in her own plan: X relies, counts on Y. X plans to achieve gX through Y. So, she is formulating in her mind not a single-agent but a multi-agent plan and Y has an allocated share in this plan: Y's delegated task is either a state-goal or an action-goal [Castelfranchi, 1998] (see Figure 2.9).
<INSERT Figure 2.9 here>
Figure 2.9:
The potential for Delegation
To do this X has some trust both in Y’s ability and in Y’s predictability, and X should abstain from doing and from delegating to others the same task.
We have classified delegation in three main categories: weak, mild and strong delegation.
(i) In weak delegation there is no influence from X to Y, no agreement: generally, Y is not aware of the fact that X is exploiting his action.
As an example of weak and passive but already social delegation, which is the simplest form of social delegation, consider a hunter who is waiting and is ready to shoot an arrow at a bird flying towards its nest. In his plan the hunter includes an action of the bird: to fly in a specific direction; in fact, this is why he is not pointing at the bird but at where the bird will be in a second. He is delegating to the bird an action in his plan; and the bird is unconsciously and (of course) unintentionally collaborating with the hunter's plan.
(ii) In a bit stronger form of delegation (mild delegation) X is herself eliciting, inducing the desired behavior of Y to exploit it. Depending on the reactive or deliberative character of Y, the induction is just based on some stimulus or is based on beliefs and complex types of influence.
(iii) Strong delegation is based on Y's awareness of X's intention to exploit his action; normally it is based on Y's adopting X's goal (for any reason: love, reciprocation, common interest, etc.), possibly after some negotiation (request, offer, etc.) concluded by some agreement and social commitment.
The act of Delegation
Notice that weak delegation is just a mental operation or action, and a mental representation. X’s external action is just waiting for or abstaining from doing the delegated action or doing her own part of the plan. On the contrary, in (ii) and (iii) Delegation is an external action of X on Y, which affects Y and induces him to do the allocated task. Here to Delegate means to bring it about that Y brings it about that p. If EX(p) represents the operator “to bring it about that”, indicating with X the subject of the action and p the state resulting of X’s action, we have the situation shown in Table 2.2.
Table 2.2:
Role of X in the different Kind of Delegation
Weak Delegation
X is exploiting EY(p)
Mild Delegation and
Strong Delegation
EX(EY(p))
2.6.1 Trust in different forms of delegation
Although we claim that trust is the mental counter-part of delegation, i.e. that it is a structured set of mental attitudes characterizing the mind of a delegating agent/trustor, however -as we saw- there are important differences, and some independence, between trust and delegation. Trust and delegation are not the same thing.
Delegation necessarily is an action (at least mental) and the result of a decision, while trust can be just a potential, a mental attitude. The external, observable behavior of delegating either consists of the action of provoking the desired behavior, of convincing and negotiating, of charging and empowering, or just consists of the action of doing nothing (omission) waiting for and exploiting the behavior of the other. Indeed, we will use trust and reliance only to denote the mental state preparing and underlying delegation (trust will be both: the small nucleus and the whole)
In our previous works we used "reliance" as a synonym of "delegation", denoting the action of relying on; here we decide to use "reliance" for the (part of the) mental state, and only "delegation" for the action of relying and trusting..
There may be trust without delegation: either the level of trust is not sufficient to delegate; or the level of trust would be sufficient but there are other reasons preventing delegation (for example prohibitions); or trust is just potential, a predisposition: “X will, would, might rely on Y, if/when...”, but it is not (yet) the case. So, trust is normally necessary for delegation, but it is not sufficient: delegation requires a richer decision.
There may be delegation (or better just ‘counting on’) without trust: these are exceptional cases in which either the delegating agent is not free (coercive delegation
Consider the example of the drunk driver, in note 27.) or she has no information and no alternative to delegating, so that she must just make a trial (blind delegation). So, all trust decisions and acts imply an act of delegation, but not any act of delegation is an act of trust.
Moreover, the decision to delegate has no degrees: either X delegates or X does not delegate. Indeed trust has degrees: X trusts Y more or less relatively to a. And there is a threshold under which trust is not enough for delegating.
Trust in Weak Delegation
While considering the possible temporal gap between the decision to trust and the delegation we have to consider some other interesting mental elements. (The temporal gap ranges between 0 and ¥; 0 means that we have delegation at the same time of decision to trust; ¥ means that delegation remains just a potential action). In particular we have in all the cases (weak, mild and strong delegation) an X's intention-that Y will achieve the task [Grosz and Kraus, 1996]. In every case this intention is composed by different intentions:
<INSERT Figure 2.10 here>
Figure 2.10:
Mental Ingredients for Weak Delegation
In weak delegation, we have 3 additional intentions (I1, I2, and I3 in Figure 2.10) respectively the intention that Y achieves p by the action a; the intention to not to do (or do not delegate to others) that action; and the intention to not hinder that Y’s action with other interfering actions.
Trust in Mild Delegation
In mild delegation in addition to I1, I2, and I3 there is another intention (I4), that is, X's intention to influence Y in order that Y will achieve t (Figure 2.11).
<INSERT Figure 2.11 here>
Figure 2.11:
Mental Ingredients for Mild Delegation
Trust in Strong Delegation
<INSERT Figure 2.12 here>
Figure 2.12:
Mental Ingredients for Strong Delegation
In strong delegation the X’s intended action is an explicit request (followed by an acceptance) to Y about p (see Figure 2.12).
Consider that in mild and strong delegation the intentions are already present in the decisional phase and they are the result of an evaluation. For example, X has to evaluate if the delegation will be successful or not in the case of influence, request, etc.
2.6.2 Trust in Open Delegation versus Trust in Closed Delegation
A very important distinction is also that between open and closed delegation.
In closed delegation the task assigned to Y is fully specified. It is just a sequence of actions to be performed. The task is a merely executive task. The extreme case of this is the classical tayloristic industrial organization of work, where to the worker is explicitly forbidden to think about the delegated task (the conditions for realizing the task are constraining any potential free initiative in the job), and he has just to perform a repetitive and mechanic movement.
In (completely) open delegation on the contrary the assigned task to Y is “to bring it about that p”, to achieve (in some way) a given result. Y has to find and chose ‘how’ to realize his ‘mission’. He has to use his local and timely information, his knowledge and experience, his reasoning, and so on.
Of course, our evaluation and expectations about Y are very different in the two cases; and the trust that we have in Y refers to different things. In closed delegation we trust Y for being obedient, precise, and skilled; we trust him for ‘executing’ (with no personal contribution – if not minimal sensory-motor adaptation) the task, which is completely specified. In open delegation we trust Y not only for his practical skills, but also for his understanding of the real sense of the task, for his problem solving ability, for his competent (and even creative) solution.
We might even trust Y for violating our request and specific expectation in order to over-help us. Over-help [Castelfranchi and Falcone, 1998b] is when Y does more than requested (for example, he satisfies not only the required goal but also other goals of X in someway linked with it).
A special form of over-help is critical help: when Y is not able or in condition to do as requested, or understands that X’s request is simply wrong (for her goal), or he has a better solution for her problem, and violates her specific request but in order to satisfy her higher goal. This is the best form of real collaboration (even if risky in someway); Y is really helpful towards X’s goals; he is not a stupid executor.
Sometimes, we deeply trust Y for over-help or for critical-help. We confidently expect that –in case– he will do more than requested or will even violate our request in order to realize the adopted goal of us (or to guarantee our ‘interests’). This is close to the most advanced trust: tutorial trust (§ 2.9).
2.7 The others parts of the relation: the delegated task and the context
2.7.1 Why does X trust Y ?
It is not enough to stress that X trusts Y to do a (an action), or as for doing something.
First of all, as we said, Trust is not simply a (firm) prediction, but is a positive expectation. In other words X is interested in the performance of a, she expects some positive result; one of the outcomes of a is (considered as or is) a goal of X. As we have remarked, an aspect absolutely necessary but frequently ignored (or at least leaved implicit) is that of the goal, the need, relatively to whom and for the achievement of whom the trustor counts upon the trustee. This is implicit when some “positive” result/outcome or the “welfare” or “interest” are mentioned in the definition, or some “dependence” or “reliance”, or the “vulnerability” are invoked. This is also why, when X trusts Y for something, he cannot only be surprised, but it can be ‘disappointed’ (and even betrayed) by Y or by his performance.
Moreover, as we have anticipated, the task allocated to Y –especially in social trust- is delegated to Y, since Y is an ‘autonomous’ agent, and in any case X is relying on a process which is beyond his power and control. Trust implies some (perceived) lack of controllability.
We call ‘task’ (t) the delegated action a to be performed and the goal state p (corresponding or including gX) to be realized by Y (in both cases in fact Y has to ‘see it that ...’, to ‘bring it about that ...’) (see § 2.6.1 on closed versus open delegation), because it is something allocated to Y within a multi-agent plan; something to be done; something on which X counts on; and something that frequently Y has an obligation to do (due to a promise, a role, etc.).
The theory of tasks is important (see trust generalization in Chapter 5).
2.7.2 The Role of the Context/Environment in Trust
Trust is a context dependent phenomenon and notion. This means that X trusts Y for t on the basis of a specific context; just changing the context (for the same t and the same Y) X’s attitude and decision might be different.
Consider X’s trust attitude towards the same agent Y for the same task t when:
- he (Y) is in two completely different contexts (maybe with different environmental and/or social conditions);
- she (X) is in two completely different contexts (maybe with different environmental and/or social conditions).
In fact, one should perhaps be more subtle, and clearly distinguish these two kinds of Context:
the Context of X’s evaluation and decision (affecting her mind) while feeling trust for Y and deciding to trust him or not (evaluation context); and
the Context of Y’s performance of a (execution context).
They are not one and the same context. The execution context affects Y’s objective trustworthiness; his possibility to really achieve the goal in a good way; and – as perceived by X (XTWY) – affects X’s expectation.
But the evaluation context is the social and mental environment of X’s decision. This can affect:
- X’s mood and basic social disposition;
- X’s information and sources;
- The beliefs activated and taken into account by X;
- X’s risk perception and acceptance;
- X’s evaluation of the execution context; and so on.
Moreover, the evaluation and decision of X also depends on this complex environmental trust: X’s trust in the environment where a will be executed, which can be more or less interfering or harmful; in the supporting infrastructure (execution tools, coordination and communication tools, etc.); in the institutional context (authorities, norms, and so on); in the generalized atmosphere and social values; and so on.
Environmental trust (external attribution) and trust “in” Y (internal attribution) must be combined for a decision; and they are also non-independent one from the other.
Not only the trust in Y as for t is context dependent, but if the context (environment) in the mental model of X plays an active causal role, X has also to trust the context, as favorable or not too much adverse or even hostile (to Y or to t). But –of course– Y’s capacity and reliability may vary with the more or less adverse nature of the context: it might decrease or even increase. On this we have developed a specific section (see Chapter 5). This is also very important in trust dynamics, since it is not true that a failure of Y necessarily will decrease Y’s perceived trustworthiness for X; it depends on the causal attribution of the failure. The same holds for success.
Another important way in which the context is relevant for trust, is that there can be different trusts about Y in different social contexts, related to the same task: for example, Y is a medical doctor, and he is very well reputed among the clients, but not at all among his colleagues. Or there can be different trusts in different social contexts because different tasks are relevant in those contexts. For example, Y is well reputed guy within his university (as teacher and researcher), but has a very bad reputation in his apartment building (as an antisocial, not very polite and clean, noise guy).
Trust can migrate from one task to another, from one trustor to another, from one trustee to another (see Chapter 5), and also from one social context to another. It depends on the connections between the two contexts: are they part one of the other? Are they connected in a social network? Do they share people, values, tasks, etc.? So, trust is not only a context dependent and sensible phenomenon but is a context-dynamic phenomenon.
Moreover, not only trust is context-dependent but it can be also conditional: A special event (e) could be considered by X, in a given context and with respect a specific trustee, as crucial for trusting Y (see Table 2.3).
Table 2.3:
Conditional Trust
IF
Event (e) = true
THEN
Trust (X Y C t gX) will be over the threshold for delegating
Consider our example of bus stop, in weak delegation. After Y raised his harm to stop the bus the driver is more sure that that boy at the bus stop will take the bus. In our view, this is not just simple ‘conditional’ probability (after the first event, or given condition C, the probability of the second event is greater or smaller). In real trust –given its attributional nature– the first event can be interpreted by X as a signal. For example, a given act or attitude or sentence of Y can be a sign for X of Y’s capacity or of his internal disposition, which makes more reliable his doing t.
2.8 Genuine Social trust: Trust and Adoption
As we saw trust is not only a ‘social’ attitude. It can be directed towards an artifact or unanimated process. Someone would prefer another term, say confidence, but this is just a (reasonable) technical convention, not the real use and meaning of these words.
Moreover, also ‘confidence’ is very close to ‘trust’ as non-technical meaning, it just seems to contain some reliance, and be quite social. It also seems more based just on learning and experience. However, it is true that the most theoretically and practically relevant and the most typical notion of trust is the social one.
Social trust means trust towards another autonomous agent perceived (conceived) as such. That is, towards a purposive, self-governed system, not only with its own resources and causal-processes, but with its own internal control and choices. This is why social trust is there at all; or better, why social interaction requires trust.
The trustor cannot and does not fully ‘control’ (monitor and guide) Y’s activity towards the goal (X’s expected result). X passes to Y (part of) the control needed for realizing the expected result. X relies precisely on this. On the one side, this precisely is one of the main advantages of delegating (task assignment and counting on): delegating also the control and governance of the activity (even if X was able to perform it herself). But, on the other side, this is precisely the specific risk of social trust: not just possible external (environmental) interferences, but ‘internal’ interferences (due to the Y’s nature and functioning).
Y is selecting the right action, employing resources, planning, persisting, executing; he might be defective on this. Moreover, since he has his own control system for purposive behavior, usually he has this for his own internal goals. Now, those individual goals may interfere; taking precedence, being in conflict and prevailing, etc.
If X decides to trust Y, to count on him, X expects (predicts and wishes) that X’s goal – adopted by Y – will prevail on Y’s autonomous goals, and will be pursued. This is the typical bet of social trust.
There is a peculiar relation between (social) trust and autonomy: we trust in autonomous systems and this is our specific (social) risk: possible indifference, or hostility, or change of mind, or profiting and exploitation, up to a ‘betrayal’, which presupposes a specific or general, explicit or implicit, ‘pact’.
Since social trust is towards another autonomous agent, considered as an autonomous agent, with its attitudes, motivations (including the social ones), and some freedom of choice, it requires an intentional stance towards a social entity (with its own intentional stance towards us).
However, this is not yet enough for capturing the most typical social notion of trust; what many authors (like Baier, Hardin, Holton, Tuomela,) would like to call genuine trust.
Genuine (social) trust, the basic, natural form of social trust, is based on Y’s adoptive attitude. That is, X trusts Y’s adoption of her interest/goal, and counts on this.
Y is perceived as taking into account X’s goals/interests; and possibly giving priority to them (in case of conflicts). This is true trust in a social agent “as a social agent”.
Social goal-adoption, is the idea that another agent takes into account in his mind -in order to satisfy them- my goals (needs, desires, interests, projects, etc.); he ‘adopts’ them as goals of himself, since he is an ‘autonomous agent’, i.e. self-driven and self-motivated (but not necessarily ‘selfish’!), and is not an hetero-directed agent, and can only act in view, be driven by, some internal purposive representation [Conte and Castelfranchi, 1995]. So -if such an (internally represented) goal will be preferred to others- he will be regulated by my goal; for some motive he will act in order to realize my goal.
A very important case of goal-adoption (relevant for trust theory) is goal–adhesion, where X wants and expects that Y adopts her goal, communicates (implicitly or explicitly) this expectation or request to Y; Y knows that X has such an expectation and adopts X’s goal not unilaterally and spontaneously, but also because X wants so. Thus not only Y adopts X’s goal that p, but he also adopts X’s goal: “that Y adopts her goal p”. In social trust frequently Y’s adoption (cooperation) is precisely due to X’s expectation and trust in Y’s adoption; and X relies on this response and adhesion.
We agree with Hardin ([Hardin, 2002]; Chapter1) that there is a restrict notion of social trust which is based on the expectation of adoption (or even adhesion), not just on the prediction of a favorable behavior of Y. When X trusts Y in strict social sense and counts on him, she expects that Y will adopt her goal and this goal will prevail – in case of conflict with other active goals. That is, X not only expects an adoptive goal by Y but an adoptive decision and intention. A simple regularity based prediction or an expectation simply based on some role or norm prescribing behavior to Y, are not enough – we agree with Hardin – for characterizing what he call “trust in strong sense”, the “central nature of trust”, what we call ‘genuine social trust’.
However, in our view, Hardin is not able to realize the broad theory of goal-adoption, and provides us –with his notion of encapsulated interests- a restricted and reductive view of it.
The various authors searching for a socially focused and more strict notion of Trust go in this direction, but using a non general and not well defined notions, like: benevolence, good-will, other-regarding attitude, benignity [Hart, 1988], altruism, social-preferences, reciprocity, participant stance [Holton, 1994].
And even the strange and unacceptable notion proposed by Deutch [Deutsch, 1985] (we discuss on this in Chapter 1 and Chapter 4) and repeated several times (for example, Bernard Barber “to place the others interest before their own”) where in order to trust Y one should assume that he is altruistic or even irrational.
What really X has to believes about Y is that:
i) Y has some motive for adopting X’s goal (for doing that action for X; for taking care of X’s interest); and that he will actually adopt the goal.
ii) Not only Y will adopt X’s goal (that is, he will formulate in his mind the goal of X, because it is the goal of X) but also that this goal will become an intention, so that Y will actually do as desired.
If (i) and (ii) are both true we can say that the adopted goal will prevail against other possible active goals of Y, included non-adopted goals (selfish).
More precisely we can claim that the motives X ascribes to Y while adopting X’s goal are assumed to prevail on the other possible motives (goals) of Y. Thus, what X is really relying on in genuine trust, are Y’s motives for an adoptive intention.
The fact that a genuine social trust is based/relies on Y’s adoption should not be misinterpreted. One should not confuse goal-adoption with specific motives for adopting. Claiming that X counts on Y’s adoptive intention is not to claim that she counts on Y’s altruism, benevolence, good will, social preferences, respect, reciprocity, or moral norms. This are just specific sub-cases of the reasons and motives Y is supposed to adopt X’s goal for. X might count on Y’s willingness to be well reputed (for future exchanges), or on his desire to receive gratitude or approval, or of avoiding blame or sanctions, or for his own approval, etc.. In other words: Y can be fully self-interested.
To realize this it is necessary to keep in mind that the usual structures of goals are means-end chains: not all goals are final goals; they can be instrumental goals, simple means for higher goals. Thus, on the top of an adoptive and adopted goal there can be other goals, which motivate the goal-adoption. For example, I can do something for you, just in order to receive what I want for me, what you promise to me.
“It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest. We address ourselves, not to their humanity but to their self-love, and never talk to them of our own necessities but of their advantages” [Smith, 1776]; however, when I order the brewer to send me a box of beer and I send the money, I definitely trust him as for giving me the beer.
As seen in §1.5.7 we have three kinds of social goal-adoption [Conte e Castelfranchi, 1995]: Instrumental, Cooperative and Terminal.
X can trust Y, and trusts that Y will do as expected, for any kind of adoption, also (or better, usually) instrumental (with both external or internal incentives). Trust in Y doesn’t presuppose that Y is ‘generous’ or that he will make ‘sacrifices’ for X; he can strictly be selfish.
Now, we can formulate in a more reasonable way Deutch’s claim and definition, without giving the impression of trust as counting on Y’s altruism or even irrationality.
Y can perfectly be self-motivated or interested (autonomous, guided by his own goals) and can even be selfish or egoistic; what matters is that the intention to adopt X’s goal (and thus the adopted goal and the consequent intention to do a) will prevail on other non-adoptive, private (and perhaps selfish) goals of Y. But this only means that:
Y’s (selfish) motives for adopting X’s goal will prevail on Y’s (selfish) motives for not doing so and giving precedence to other goals.
So, X can count on Y doing as expected, in X’s interest (and perhaps for Y’s interest).
Trustworthiness is a social ‘virtue’ but not necessarily an altruistic one.
This makes also clear that not all ‘genuine’ trust is ‘normative’ (based on norms) (for example, the generous impulse of helping somebody which is in serious danger is not motivated by the respect of a moral/social norm, even if this behavior (later) is socially/morally approved).
Moreover, not all ‘normative’ trust is ‘genuine’. We can trust somebody for doing (or not doing a) just because we know that he has to do so (for a specific law or role), independently on his realizing or not and adopting or not our goal. For example, I trust a policeman for blocking and arresting some guy which was aggressing me, not because he has to respond to my desire, but just because he is a policeman in front of a crime (he can even ignore me).
It is also important do not mix up ‘genuine’ adoption-based trust with trust in ‘strong delegation’: delegation based on X’s request and Y’s acceptance. ‘Genuine’ trust can also be there in weak and in mild delegation/reliance: when Y ignores X’s reliance and acts on own account, or when Y’s behavior is elicited by X (but without Y’s understanding). In fact, Y might have spontaneous reasons for adopting X’s interests (and X might count on and exploit this), or X might elicit in Y adoptive motives and attitudes by manipulating Y, without Y knowing that X is expecting and counting upon his adoption.
In sum, in genuine trust X just counts upon the fact that Y will understand and care of her (delegated) goal, Y will adopt her goal and possibly prefer it against conflicting goal (for example selfish ones), and this for whatever reason: from selfish advantages to altruism, from duty and obligations to cooperation, from love to identification, and so on.
Also May Tuomela [Tuomela, 2003] introduces and define an interesting notion of ‘genuine’ social Trust. But in our view this notion is too limited and specific. We disagree her constraint that there is genuine trust only when it is symmetrical and reciprocal (for us counterintuitive and restrictive). In addition, her conditions (to be respected, the fact that the other will care of my rights, etc.) look quite peculiar of specific -important- social relationships where for sure there is ‘genuine’ trust, but which exclude other typical situations of trust (like child-mother) that must be covered.
Paradoxically, sometimes we trust Y precisely for his selfishness, which makes him trustworthy and reliable for that task/mission.
2.8.1 Concern
A very important notion in goal-adoption is the notion of concern. How much the goal of X is important for Y; how much Y is concerned with/by X’s interest. That is, which is for Y the value of X’s goal gX, or better of X achieving her goal. This value is determined by:
(i) the reasons (higher motivations) that Y has for adopting X’s goal, and their value for him; how much and why Y cares of X’s welfare;
(ii) X’s opinion about the subjective value of gX for Y.
It is precisely on this base that the adopted goal will prevail or not against possible costs, against other private conflicting goals of Y, and thus will possibly become/produce an adoptive intention of Y; and will also -as intention- persist against possible new interferences and temptations.
It is precisely on Y’s concern for X’s goal (not be confused with benevolence, good will, benignity, and so on) that X relies while betting on Y’s adoptive intention and persistence. She also has some ‘theory’ about the reasons Y should be concerned with her welfare and to adopt her goal.
2.8.2 How Expectations generate (entitled) Prescriptions: Towards ‘betrayal’
It is characteristic of the most typical/genuine forms of social trust that -in case of failure- X is not only surprised and disappointed [Miceli and Castefranchi, 2002; Castelfranchi and Giardini, 2003], but feels betrayed. Where does this affective reaction come from? On which beliefs and goals (present in the trust attitude) is it based?
Social expectation can be entitled, can be based on Y’s ‘commitment’ and thus obligation towards X [Castelfranchi, 1995]. What X expects from Y can be ‘due’. The violation of this kind of expectations involves not only disappointment but stronger and social emotions, like anger, indignation, etc. In particular it is different if this entitlement, this duty of Y towards X comes from legal norms or from interpersonal relations and merely social norms, like in a promise or like in friendship where fairness and adoption are presupposed.
In these forms of ‘genuine’ trust, where the expectation of Y’s adopting/caring of my needs, requests, wishes (goals), is based on an assumption of a moral duty towards me, if Y disappoints this expectation I feel betrayed by Y in my trust and reliance in him.
This commitment – and the consequent moral duty, social norm – it is not necessarily established in an explicit way; for example by a promise. Not only -as we said- it can be presupposed in the very relationship between us: friends, same family, same group, shared identity (which oblige to some solidarity). It can be established by tacit consent, implicit behavioral communication [Castelfranchi, 2006; Tummolini and Castelfranchi, 2006 ]. See Table 2.4 for an example.
Table 2.4:
Example of Tacit Consent
IF
X decides to count on Y, AND
Y is aware of such an expectation, AND
X is aware that Y is aware, and Y knows this, AND
Y would be able and in condition of rejecting such a delegation, to refuse to ‘help’ X, and to inform X about this (a very relevant information for her), and Y knows that X knows this; AND
Y says nothing; doesn’t provide any sign of his refusal
THEN
Y ‘tacitly consent’:
Y takes a commitment, a tacit obligation to do as expected, to not disappoint X; AND
X gets a soft right towards Y: she is entitled to ask and claim for Y’s action, and to complain and protest for his not doing as ‘committed’
More in general, this mechanism is responsible for the tendency of shared social expectations (expectations about the behavior of the other agents, which are common knowledge) to become prescriptions: not only I predict that you will do something, but I wish so; I want that you behave in such a way (expectation). Moreover, I know that you know (etc.), and you did not disconfirm this (etc.), so you get some obligation of not violating my expectations. And I want that you do as expected also in force of this reason: because you have a duty and you know (recognize) that I want that you do that for such a reason. So my expectation becomes a true prescription [Castelfranchi and Giardini, 2003]. This is how common expectations become social ‘conventions’ and ‘norms’.
In sum, also on the bases of such tacit “promises” and interpersonal norms, or of those obligations implicit in the relationship, X can feel betrayed by Y, since she was trusting Y on such a specific base.
2.9 Super-Trust or Tutorial trust
There are very extreme forms of trust, where X ‘puts herself in Y’s hands’ in a radical sense; in the sense that she believes and accepts that Y will care of her welfare better than her, beyond what she intends, asks, desires. One case is Over-Trust: trust in Y’s over-help. As we saw, we might confidently expect that -in case- Y will do more that requested or will even violate our request in order to realize the adopted goal of us. However, we can even go beyond this. In Over-help Y is supposed to take care of our actual goals, desires, of better pursuing what we want; but there are forms of trust where we accept that Y goes against our current desires and objectives, while pursuing our (non understood) interests. This is Super-Trust or Tutorial Trust: trust in the ‘tutorial’ role of Y towards me. I feel so confident in Y that I am convinced that Y is pursuing my good, and is helpful, even when Y is acting against my current goals and I do not understand on what he is taking care of me.
In other words, I assume that Y is taking care of my wellness, of the best for me, of my (possibly not so clear to me) interests, not just of my actual and explicit goals and desires (and may be against them). He makes that for my good.
This presupposes that I feel/believe that I ignore part of my interests, of what is good for me (now or for the future), and I assume that, on the contrary, Y is able to understand better than me what is good for me (my interests) and cares of this, and wants -even against me- to protect my interests or oblige me to realize them.
We have modeled [Conte & Castelfranchi, 1995] this kind of social relationship between Y and X (when Y claims to know better than X what is better for X, and care of this, and try to influence X to do what is better for X); and we have label this tutorial relation of Y towards X. It exists also when X doesn't recognize or even contests it (like between parents and adolescents, or between psychiatrists and patients, etc.).
However, sometimes this is just an arrogant and arbitrary claim, hiding Y's power and advantages, or 'paternalism'.
In Super-Trust, X presumes a tutorial attitude and relation from Y, and relies on this, since he feels/believes that Y's is really capable to understand and will care of what is better for X.
2.10 Resuming the model
Let us resume in a schematic and synthetic way how in our cognitive model of trust the different elements, playing a role in the trust concept, are composed and ordered for producing the trusting behavior of an agent. As we have seen a main role is played by the goal of the trustor that has to be achieved through the trustee (without this motivational component there is no trust). In fact, in addition to the goal, it is also necessary that the trustor believes to be (strongly or weakly) dependent from the trustee himself see Figure 2.13.
<INSERT Figure 2.13 here>
Figure 2.13:
Dependence Belief and Goal: First step towards the decision of trusting
On the basis of the goal, of her (potential) dependence beliefs
To be true, the Dependence belief already implies some belief about Y’s skills or resources, that are useful for X’s goal., of her beliefs about the trustee attributes (internal trust), of her beliefs about the context in which the trustee performance will come, the trustor (potentially) arrives to the decision to trust or not (Figure 2.14).
<INSERT Figure 2.14 here>
Figure 2.14:
The complex set of beliefs converging towards the decision of trusting
As explained in § 2.2.1., all this possible beliefs are not simply external bases and supports of X’s trust in Y (reduced to the Willingness and Competence and Dependence beliefs, and to the Decision and Act), but they are possible internal sub-components and forms of trust, in a recursive trust-structure. The frame looks quite complicated and complex, but, in fact, it is only a potential frame: not all these sub-components (for example, the beliefs about X’s morality, or fear of authority, or self-esteem) are necessarily and already there or explicitly represented.
Moreover, as we will see in detail in the next Chapter 3, a relevant role is played by the quantification of the different elements: the weight of the beliefs, the value of the goal, the potential utilities resulting from a delegation and so on (see Figure 2.15).
<INSERT Figure 2.15 here>
Figure 2.15:
The role of the quantification in the complex scenario of Decision to Trust
REFERENCE
[Baier, 1986] Annette Baier “Trust and Antitrust”, Ethics 96 (1986), pp.231-60
[Bacharach and Gambetta, 2001] M. Bacharach, D. Gambetta, Trust as type detection, in: C. Castelfranchi, Y.-H. Tan (Eds.), Trust and Deception in Virtual Societies, Kluwer Academic Publishing, Dordrecht, The Netherlands, 2001, pp. 1 –26.
Cook (ed.) Trust and Social Structure, New York: Russel Sage Foundation, forthcoming.
[Castaldo, 2002] Castaldo, S. (2002). Fiducia e relazioni di mercato. Bologna: Il Mulino.
[Castelfranchi, 1995] Castelfranchi, C. Social Commitment: from individual intentions to groups and organizations. In ICMAS'95 First International Conference on Multi-Agent Systems, AAAI-MIT Press, 1995, 41-49 (versione preliminare in AAAI Workshop on "AI and Theory of Group and Organization", Washington, D. C., maggio 1993).
[Castelfranchi, 1996] Castelfranchi, C.(1996). Reasons: Belief Support and Goal Dynamics. Mathware & Soft Computing, 3. pp. 233-47.
[Castelfranchi, 1997] Castelfranchi, C. (1997) Representation and integration of multiple knowledge sources: issues and questions. In Cantoni, Di Gesu', Setti e Tegolo (eds), Human & Machine Perception: Information Fusion. Plenum Press.
[Castelfranchi, 1998] Castelfranchi, C. (1998). Modelling Social Action for AI Agents. Artificial Intelligence, 103, 1998, 157-182.
[Castelfranchi, 1998] Castelfranchi, C. Towards an Agent Ontology: Autonomy, Delegation, Adaptivity. AI*IA Notizie. 11, 3, 1998; Special issue on “Autonomous Intelligent Agents”, Italian Association for Artificial Intelligence, Roma, 45-50.
[Castelfranchi, 2000a] Cristiano Castelfranchi: Again on Agents' Autonomy: A Homage to Alan Turing. ATAL 2000: 339-342
[Castelfranchi, 2000b] Castelfranchi, C. (2000). Affective Appraisal vs Cognitive Evaluation in Social Emotions and Interactions. In A. Paiva (ed.) Affective Interactions. Towards a New Generation of Computer Interfaces. Heidelbergh, Springer, LNAI 1814, 76-106.
[Castelfranchi, 2000c] Castelfranchi, C. Through the Agents’ Minds: Cognitive Mediators of Social Action. Mind and Society. Torino, Rosembergh, 2000 pp.109-140.
[Castelfranchi, 2003] Castelfranchi, C. The Micro-Macro Constitution of Power, ProtoSociology, An International Journal of Interdisciplinary Research Double Vol. 18-19, 2003 Understanding the Social II – Philosophy of Sociality, Edited by Raimo Tuomela, Gerhard Preyer, and Georg Peter.
[Castelfranchi, 2005] C. Castelfranchi: Mind as an Anticipatory Device: For a Theory of Expectations. In “ Brain, vision, and artificial intelligence” Proceedings of the First international symposium, BVAI 2005, Naples, Italy, October 19-21, 2005 . Lecture notes in computer science ISSN 0302-9743 vol. 3704, pp. 258-276
[Castelfranchi, 2006] C. Castelfranchi: SILENT AGENTS: From Observation to Tacit Communication. IBERAMIA-SBIA 2006: 98-107
[Castelfranchi and Falcone, 1998a] Castelfranchi C., Falcone R., (1998) Principles of trust for MAS: cognitive anatomy, social importance, and quantification, Proceedings of the International Conference on Multi-Agent Systems (ICMAS'98), Paris, July, pp.72-79.
[Castelfranchi and Falcone, 1998b] Castelfranchi, C., Falcone, R., (1998) Towards a Theory of Delegation for Agent-based Systems, Robotics and Autonomous Systems, Special issue on Multi-Agent Rationality, Elsevier Editor, Vol 24, Nos 3-4, pp.141-157.
[Castelfranchi and Giardini, 2003] Castelfranchi, C., Giardini, F., Lorini, E., Tummolini, L. (2003) The Prescriptive Destiny of Predictive Attitudes: from Expectations to Norms via Conventions, in R. Alterman, D. Kirsh (Eds) Proceedings of the 25th Annual Meeting of the Cognitive Science Society, Boston, MA.
[Castelfranchi and Lorini, 2003] Castelfranchi, C., Lorini, E. (2003). Cognitive Anatomy and Functions of Expectations. In Proceedings of IJCAI’03 Workshop on Cognitive Modeling of Agents and Multi-Agent Interactions, Acapulco, Mexico, August 9-11, 2003.
[Castelfranchi and Paglier, 2007] C. Castelfranchi, F. Paglieri (2007). The role of beliefs in goal dynamics: Prolegomena to a constructive theory of intentions. Synthese, 155, 237-263.
[Cohen, 1992] Cohen, J., 1992, An Essay on Belief and Acceptance, Oxford University Press, Oxford
[Conte and Castelfr, 1995]
[Dennett, 1989], D.C. Dennett, The Intentional Stance. Cambridge, MA: MIT Press, 1989
[Deutsch, 1985] Deutsch, M. (1985). The resolution of conflict: Constructive and destructive processes. New Haven, CT: Yale University Press.
[Elster, 1979] Elster Jon. Ulysses and the Sirens. Cambridge: Cambridge University Press. 1979.
[Engel, 1998] Engel, P., 1998, ‘Believing, Holding True, and Accepting’, Philosophical Explorations 1, 140-151
[Falcone and Castelfranchi, 2001b] Falcone R., Castelfranchi C., (2001). Social Trust: A Cognitive Approach, in Trust and Deception in Virtual Societies by Castelfranchi C. and Yao-Hua Tan (eds), Kluwer Academic Publishers, pp. 55-90.
[Festinger, 1957] Festinger, L., (1957). A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.
[Gambetta, 1988] Gambetta Diego. “Can we trust trust?” In Trust: Making and Breaking Cooperative Relations, edited by D. Gambetta. Oxford Blackwell. 1988.
[Grosz and Kraus, 1996] Grosz, B. Kraus, S. 1996. Collaborative plans for complex group action Artificial Intelligence, 86, 269-358.
[Hardin, 2002] Hardin R., Trust and Trustworthiness, New York: Russel Sage Foundation, 2002.
[Hart, 1988] Hart, K., (1988), Kinship, Contract and Trust: Economic Organization of Migrants in an African City Slum, In Trust: Making and Breaking Cooperative Relations, edited by Diego Gambetta. Oxford: Blackwell.
[Hertzberg, 1988] Hertzberg L., (1988). On the Attitude of Trust. Inquiry 31 (3):307 – 322.
[Holton, 1994], R. Holton Deciding to Trust, Coming to Believe. Australian Journal of Philosophy 72 (1994),pp. 63-76.
[Jennings, 1993] Jennings N.R. 1993. Commitments and conventions: The foundation of coordination in multi-agent systems. The Knowledge Engineering Review, 3, 223-50.
[Jones, 2002] Jones A.J., “On the concept of trust”, Decision Support Systems, 2002, Volume 33, Issue 3: pp. 225 - 232 - Special issue: Formal modeling and electronic commerce.
[Johnson-Laird, 1983] Johnson-Laird, P.N. (1983). Mental models: towards a cognitive science of language, inference and consciousness. Cambridge, UK: Cambridge University Press.
[Levesque, 1984] Levesque, H. J. (1984). A logic of implicit and explicit belief. In Proceedings of the Fourth National Conference on Artificial Intelligence (AAAI-84), pages 198-202, Austin, TX.
[Luhmann, 1979] Luhmann N., (1979), Trust and Power, Wiley, New York.
[Marsh, 1994], Marsh, S.P., (1994), Formalising Trust as a computational concept. PhD thesis, University of Stirling. Available at: http://www.nr.no/abie/papers/TR133.pdf.
[Mayer et al, 1995] Mayer, R.C., Davis, J.H., & Schoorman, F.D. (1995). An integrative model of organizational trust. Academy of Management Review, 20(3), 709-734.
[McKnight and Chervany, 2001] McKnight, D. H. and Chervany, N. L.Trust and Distrust Definitions: One Bite at a Time. In Trust In Cyber-societies, Volume 2246 of Lecture Notes in Computer Science, pages 27-54, Springer, 2001.
[Meyer, 1992] Meyer, J.J. Ch., van der Hoek W., (1992), A modal logic for nonmonotonic reasoning. In W. van der Hoek, J.J. Ch. Meyer, Y. H. Tan and C. Witteveen, editors, Non-Monotonic Reasoning and Partial Semantics, pages 37-77. Ellis Horwood, Chichester, 1992.
[Miceli and Castelfranchi,1983] Miceli, M. Castelfranchi, C. e Parisi, D. Verso una etnosemantica delle funzioni e dei funzionamenti: 'Rotto', 'guasto', 'non funziona', 'malato'. Quaderni di semantica, 8 (IV), 1983, 179-208.
[Miceli&Castelfranchi, 1997] Miceli, M. e Castelfranchi C. (1997). Basic principles of psychic suffering: A prelimirary account. Theory & Psychology, 7, 769-798.
[Miceli&Castelfranchi, 2000] Miceli, M. & Castelfranchi, C. (2000). The role of evaluation in cognition and social interaction. In K. Dautenhahn (Ed.), Human cognition and agent technology (pp. 225-261). Amsterdam: Benjamins.
[Miceli e Castelfranchi, 2002] Miceli, M. e Castelfranchi, C. (2002). The mind and the future: The (negative) power of expectations. Theory & Psychology. Vol.12 (3): 335-.366
[Miller, Galanter & Pribram, 1960 ] Miller, GA, Galanter, E., & Pribram, KA (1960). Plans and the structure of behavior. New York: Holt, Rhinehart, & Winston.
[Searle, 1995] John R. Searle, The Construction of Social Reality, Free Press: N.Y., 1995
[Sichman et al., 1994] Sichman, J, R. Conte, C. Castelfranchi, Y. Demazeau. A social reasoning mechanism based on dependence networks. In Proceedings of the 11th ECAI, 1994.
[Smith, 1776] Smith, A. (1776), An Inquiry into the Nature and Causes of the Wealth of Nations, London: Methuen & Co., Ltd.
[Snijders, 1996] Snijders C., Keren G., (1996), Determinants of Trust, Proceedings of the workshop in honor of Amnon Rapoport, University of North Carolina at Chapel Hill, USA, 6-7 August.
[Tuomela, 2003] Tuomela, M. (2003). A Collective's Rational Trust in a Collective's Action. In Understanding the Social II: Philosophy of Sociality, Protosociology. 18-19, 87-126.
[Tummolini and Castelfranchi, 2006] Luca Tummolini, Cristiano Castelfranchi: Trace Signals: The Meanings of Stigmergy. E4MAS 2006: 141-156
[vanLinder, 1996] van Linder, B., (1996) Modal Logics for Rational Agents, PhD thesis, Department of Computing Science, University of Utrecht,.
55
References (2)
- On the Trustee's Side: Trust As Relational Capital 281 10.1 Trust and Relational Capital 282 10.2 Cognitive Model of Being Trusted 284 10.2.1 Objective and Subjective Dependence 285 10.2.2 Dependence and Negotiation Power 289 10.2.3 Trust Role in Dependence Networks 292 10.3 Dynamics of Relational Capital 297 10.3.1 Increasing, Decreasing and Transferring 297 10.3.2 Strategic Behavior of the Trustee 300 10.4 From Trust Relational Capital to Reputational Capital 301 10.5 Conclusions 302 References 302 11 A Fuzzy Implementation for the Socio-Cognitive Approach to Trust 305 11.1 Using a Fuzzy Approach 306 11.2 Scenarios 306 11.3 Belief Sources 307 11.4 Building Belief Sources 307 11.4.1 A Note on Self-Trust 309 11.5 Implementation with Nested FCMs 310 11.6 Converging and Diverging Belief Sources 311 11.7 Homogeneous and Heterogeneous Sources 312 11.8 Modeling Beliefs and Sources 312 11.
- 9 Overview of the Implementation 313 11.9.1 A Note on Fuzzy Values 315 11.10 Description of the Model 316 11.11 Running the Model 316 11.12 Experimental Setting 317 11.12.1 Routine Visit Scenario 317 11.12.2 Emergency Visit Scenario 319 11.12.3 Trustfulness and Decision 320 11.12.4 Experimental Discussion 321 11.12.5 Evaluating the Behavior of the FCMs 322 11.12.6 Personality Factors 322 11.13 Learning Mechanisms 323 11.13.1 Implicit Revision 324 11.13.2 Explicit Revision 324 11.13.3 A Taxonomy of Possible Revisions 325 11.14 Contract Nets for Evaluating Agent Trustworthiness 326 11.14.1 Experimental Setting 326 11.14.2 Delegation Strategies 327 11.14.3 The Contract Net Structure 328 11.14.4 Performing a Task 329 11.14.5 FCMs for Trust 329 11.14.6 Experiments Description 330 11.14.7 Using Partial Knowledge: the Strength of a Cognitive Analysis 333
