New Advances in Causation, Agency and Moral Responsibility [1 ed.] 9781443873567, 9781443866255

Citation preview

New Advances in Causation, Agency and Moral Responsibility

New Advances in Causation, Agency and Moral Responsibility

Edited by

Fabio Bacchini, Stefano Caputo and Massimo Dell'Utri

New Advances in Causation, Agency and Moral Responsibility, Edited by Fabio Bacchini, Stefano Caputo and Massimo Dell'Utri This book first published 2014 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2014 by Fabio Bacchini, Stefano Caputo, Massimo Dell'Utri and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-6625-3, ISBN (13): 978-1-4438-6625-5

TABLE OF CONTENTS

Introduction ............................................................................................... vii Chapter One ................................................................................................. 1 Causation and the Interventionist Vector of Explanatory Depth Fernanda Samaniego Chapter Two .............................................................................................. 18 Explanatory Depth, Inference to the Best Explanation and Proportionality Alexandre Marcellesi Chapter Three ............................................................................................ 32 Complex Causation in the Race Debate Ludovica Lorusso Chapter Four .............................................................................................. 49 Causal Sufficiency and Interferences Lorenzo Azzano Chapter Five .............................................................................................. 63 Absolute Coincidences and Salmon's Interactive Fork Model Alessandra Melas Chapter Six ................................................................................................ 79 The Method of Contrast and the Perception of Causality in Audition Elvira Di Bona Chapter Seven............................................................................................ 94 Actual-Sequence Freedom Carolina Sartorio Chapter Eight ........................................................................................... 106 Causes and Blame: A Pragmatic Analysis Daniele Santoro and Marcello Di Paola

vi

Table of Contents

Chapter Nine............................................................................................ 121 Responsibility Regardless of Causation Federico L.G. Faroldi Chapter Ten ............................................................................................. 144 Skepticism and Control Sofia Bonicalzi Chapter Eleven ........................................................................................ 165 Understanding Strength of Will Michael Brent Chapter Twelve ........................................................................................ 179 Responsibility and Narrative Agency Irene Bucelli Chapter Thirteen ...................................................................................... 194 Empathy and Concern for Others Sarah Songhorian Chapter Fourteen ..................................................................................... 214 Truth and Moral Responsibility P. Roger Turner Chapter Fifteen ........................................................................................ 231 “I was not aware I was doing that” Franziska Müller Abstracts .................................................................................................. 246 Contributors ............................................................................................. 253 Index of Names........................................................................................ 256

INTRODUCTION

This volume brings together a number of previously unpublished essays that aim to advance our philosophical understanding of specific aspects of causality, agency and moral responsibility. The first group of contributions focuses on causality. One of the most recent and influential proposals in accounting for causal explanations is the interventionist theory developed by James Woodward, and the first two essays deal with the notion of ‘explanatory depth’ developed by Woodward and Hitchcock within interventionism. Fernanda Samaniego argues that interventionism may lead to counterintuitive results when it is used in an attempt to define a notion of explanatory depth. In particular, if we apply the interventionist criteria of explanatory depth to cases with multiple correlated causes, such as the case of the causal relationship between cholesterol and atherosclerosis (also known as arteriosclerotic vascular disease), simpler explanations may turn out to be deeper than more detailed ones. Samaniego claims that we do have good reasons to rescue interventionism from this difficulty, and provides a new interventionist definition of explanatory depth that avoids it. Alexandre Marcellesi goes further in maintaining that explanatory depth, as developed by Woodward and Hitchcock, is not an adequate account of the factors that make one causal explanation better than another. He claims that explanatory depth, for one thing, is inconsistent with the plausible and perhaps inescapable view that causal explanations are better when they refer to causes that are proportional to their effects— that is, causes that are both necessary and sufficient for their effects. He goes on to argue that explanatory depth is not even an explanatory notion; in fact, after reminding us that explanatory depth should be one explanatory virtue of causal explanations, to be distinguished from ordinary theoretical virtues, such as predictive power, he shows that it is actually no more than predictive power in disguise. Marcellesi assumes that causal explanations and causal predictions are two at-leastconceptually distinct things—an assumption on which philosophers of science are perhaps required to focus their attention anew by the interventionist theory and its troubles. In the context of the debate about the use of racial categories in

viii

Introduction

biomedicine, Ludovica Lorusso can be read as stressing the same point: since causal predictions are not the same thing as causal explanations, we should not confuse the predictive use—both causal and noncausal—of self-identified races in biomedical research with their putative causal explanatory role. Lorusso argues against the presumption that the only way of causally explaining different risks of developing a complex disease between different self-identified racial groups consists in postulating causally relevant genetic differences between the groups. In fact, there is no biological evidence that such causally relevant genetic differences exist. Interestingly, self-identified races are claimed to be not only possibly correlated to non-genetic causes of an increased risk of developing a complex disease—which would make races useful proxies in biomedicine even if there are no genetic causes of an increased risk that are correlated to them—but even possible causes of those non-genetic causes—what would make races non-genetically causally relevant to the increased risk. There are two papers which deal with the metaphysics of causal relations rather than with the epistemology of causal explanations. Lorenzo Azzano examines what it is for something to be an interference in a causal process. After determining that an interference is required not to directly alter either the cause or the effect of the causal process it alters, he concludes that interferences can only alter causal processes that admit intermediate steps. Thus showing that all causation is sufficient causation is actually “tremendous work” consisting of sustaining some very robust metaphysical claims about either simultaneity, transitivity or the micro/macro relationship. For instance, one way of defending causal sufficiency is holding that all breakable causation just is macrocausation that can be exhaustively reduced to two-event discrete unbreakable microcausation. Another interesting notion concerning causal interaction besides interference is that of coincidence. In her paper, Alessandra Melas deals with absolute coincidences and shows how a common cause model—i.e. Salmon's interactive fork model—can account for them. Seeing absolute coincidences as x-shaped interactive forks—in Salmon’s terms—reconciles this particular aspect of chance with the Principle of Causality. An interesting question that has become popular in the philosophy of perception is that of whether we can perceive some higher-order relation as causation. Various positions have emerged concerning our alleged capacity to visually perceive causal relations, but little space has been devoted to answering the same question for other sense modalities. Elvira Di Bona takes into account a few versions of the phenomenal contrast method developed by Susanna Siegel in order to isolate one version

New Advances in Causation, Agency and Moral Responsibility

ix

entitling us to affirm that, after all, we can auditorily perceive a causal relation. The main difficulty, here, is putting together the epistemological and the phenomenological perspectives. Causation, agency and moral responsibility are of course deeply intertwined notions, and a large proportion of the volume is taken up by papers that attempt to shed light on their mutual connections or to defend certain claims concerning them. For instance, according to Carolina Sartorio, moral responsibility results from a certain kind of the agent’s freedom with regards to X which in turn supervenes on the actual causal sequence producing X. Thus moral responsibility is solidly grounded on causation: there is no responsibility difference without a causal difference. Sartorio shows that some apparent counterexamples to the supervenience thesis actually end up by supporting it. Moreover, in developing her defensive argument Sartorio explicitly takes causation to be a vehicle for the transmission of an agent’s moral responsibility from one outcome to another. A causal link between two outcomes is by itself not sufficient to transmit responsibility, as for instance some relevant epistemic conditions in the agent must obtain; a causal relation, however, turns out to be a necessary condition. While many philosophers have deemed it natural to explain responsibility in causal terms—however different the details of the explanation might be—Santoro and Di Paola seem to reverse the direction of the explanation when they propose a Wittgensteinian view of causal ascriptions based on the more primitive language game of blame ascriptions. In what they call the scapegoat theory of causality, in fact, causes are pointed at just as scapegoats for explaining otherwise mysterious events; in this view, what we are doing when we point at a cause is just blaming some event for producing another event. According to their pragmatic analysis (as developed by Robert Brandom), the deployment of causal vocabulary is only possible if an inferential discursive practice of blame and responsibility ascription is present— which in turn requires the possession of an intentional vocabulary of blame and responsibility. To support their view, Santoro and Di Paola argue that we can express causal relations purely by means of intentional vocabulary, but not the reverse: we can “blame a cause”, while we cannot “cause a blame”. Taking a different path, Federico Faroldi argues that responsibility is independent from causation—and he cites group responsibility, shared responsibility and vicarious responsibility as examples of responsibility that are obtained from requisites other than causal. Although he refers to legal rather than moral responsibility, his thesis can easily be reformulated

x

Introduction

as opposing the widely accepted idea that moral responsibility requires causing as a necessary condition at least. In her essay, Sofia Bonicalzi attempts to map the different characterizations that compatibilists and libertarianists have offered of the nature of the conditions that must be guaranteed in order for moral responsibility to be obtained. While compatibilists usually admit a causal relation as a necessary condition for moral responsibility, libertarians have often tried to say that a necessary condition for being responsible—that is, being free—requires a departure from causal relations regulating the physical world, at least as long as causal relations are meant to be deterministic. Bonicalzi says that a decisive notion in both approaches is that of control, and examines how it is treated within theories positioned at either side of the boundary. Among libertarians, for example, Robert Kane has proposed to ground the agent’s control—via ultimate responsibility— on causal chains originating from a certain number of indeterministic, causally sufficient self-forming actions. Bonicalzi finally advocates a sceptical solution to free will in the footsteps of Derk Pereboom that in her opinion can rescue deontic morality through dropping the ideas of desert and blame and adopting a revisionist forward-looking conception of moral judgments. While this group of essays in the volume focuses on the relations between causation and responsibility, some others deal with the connections between causation and agency, on the one hand, and those between agency and responsibility, on the other. Consider Michael Brent’s contribution. Here the problem is to explain what happens when an agent succeeds in persisting with a resolution in the face of a compelling desire to the contrary. According to Richard Holton, whose arguments are examined by Brent, neither a traditional Humean account (where all action is explained in terms of beliefs and desires) nor a modified Humean account (where intentions, besides desires and beliefs, are considered in the explanation of actions) can adequately clear up this quandary. Holton’s solution—which, Brent argues, is not immune to serious difficulties— consists of postulating another irreducible motivational factor, namely, that of willpower. What is worth noting is that willpower can be seen as the capacity to resist motivational interference due to powerful desires threatening to undermine previous resolutions. The analogy between causal interference and unbreakable processes (Azzano), on one side, and motivational interferences and “unbreakable” resolutions (Brent), on the other, is tempting. Moreover, when Brent turns to the difficulties affecting Holton’s view, he emphasizes that Holton offers no explanation of what the causal role of willpower is and by what causal process it can make the

New Advances in Causation, Agency and Moral Responsibility

xi

agent capable of refraining from acting according to the desire threatening her previous resolution. Holton does say that willpower consists of recalling the resolution and refusing to revise it; we have no idea, however, about how these mental acts should make us capable of resisting temptations (e.g., whether by strengthening the resolution or by weakening or blocking the desire to the contrary, or something else entirely), nor are we given any help to clarify whether these mental acts are causally necessary or not to displaying strenght of will. Irene Bucelli is interested in evaluating what consequences reflective endorsement theories of agency have on responsibility. She considers Velleman’s narrative model of agency particularly promising for providing necessary—and maybe necessary and sufficient—conditions for responsibility, since “what the narrative model ends up identifying as action already has the two standard conditions for responsibility readily built in”—the two conditions being a causal (“volitional”) and an epistemic one. Yet Bucelli concludes that the project fails inasmuch as omissions can reveal themselves to be acts of negligence we are responsible for. In fact, it appears that no reflective endorsement theory can easily accommodate omissions as instances of full-blown agency. In short, as the range of actions that the model is capable of accounting for just includes deliberated actions, it necessarily leaves out non-deliberated omissions we are routinely considered responsible for. Another interesting project is that of Sarah Songhorian. Her aim is to single out a minimal concept of empathy common to all different forms of empathic capacities, and determine its causal relevance to behaviour and, in particular, to moral behaviour. She concludes that while this minimal concept of empathy has neither the necessary force to produce moral behaviour nor to entail consequences about responsibility and other normative categories, the concept of sympathy can start doing the job thanks to its metaethical dimension, and this in particular may provide a narrow path from the is of empathy to the ought of normative attitudes. According to P. Roger Turner, some profound consequences for discussions on moral responsibility arise from certain facts about the nature of truth. In particular, the obvious truism that truth depends on the world—i.e. that, for every true proposition p, ‘p’ is true because it is the case that p—entails that counterexamples to one of the main incompatibilist arguments are impossible, and that compatibilism fails. One of the remarkable corollaries of Turner’s defence of incompatibilism is the thesis that, if an agent is directly morally responsible for what p’s truth depends on (in the sense of ‘depends on’ in which truth depends on the world), then the agent is directly morally responsible for p’s truth at

xii

Introduction

least partially. This thesis provides us with a guide to what we may call the compositionality of responsibility: Turner argues that if the agent is morally responsible for a single state of affairs that a complex proposition’s truth depends upon—in the sense, for example, that the agent is morally responsible that s, and the truth of the complex proposition ‘t entails s’ compositionally depends upon s—then the agent is also morally responsible for the truth of the complex proposition (e.g., for the truth of ‘t entails s’) and is even morally responsible that the corresponding relation between states of affairs holds (e.g., that t entails s). While Turner’s approach has the important characteristic of weakening in a very particular way the connection between moral responsibility and the epistemic conditions usually required for moral responsibility to be obtained—because for him an agent who is morally responsible that s also in virtue of her knowing both that s and that she is causing s, can be ipso facto morally responsible that v although she knows neither that v nor that she is causing v, provided that v’s truth depends on s—Franziska Müller wants to ascertain to what extent another connection that many philosophers have considered to be a very strong one—i.e. that between awareness of one’s doing and intentional action—stands. Thus she puts to the test Elizabeth Anscombe’s famous statement that if someone is not aware she is doing X, doing X is not an intentional action of hers. First, Müller establishes that there are many different ways in which we may interpret Anscombe’s claim depending on what exactly the subject is supposed to be unaware of. Then—after picking out the most plausible option—she examines the numerous senses in which the subject can be said to be (and, not to be) aware of what she is doing. Her main thesis, here, is that having a dispositional belief to be doing X is not sufficient to grant awareness of one’s doing X. Philosophers still have a great deal of work to do before such complex notions as causality, agency and moral responsibility will be fully understood. We believe, however, that these essays constitute a valid contribution towards our comprehension of such matters.i Sassari (Italy), March 2, 2014 Fabio Bacchini Stefano Caputo Massimo Dell’Utri

New Advances in Causation, Agency and Moral Responsibility

xiii

Notes i

We would like to thank Silvia Negroni for her invaluable assistance and helpful advices.

CHAPTER ONE CAUSATION AND THE INTERVENTIONIST VECTOR OF EXPLANATORY DEPTH FERNANDA SAMANIEGO

1. Introduction One of the most recent and influencing proposals in modelling causal scientific explanations is the Interventionist Theory or interventionism (James Woodward 2003).This theory aims to be applicable to causal explanations from a broad variety of natural, social and health sciences. Even though the wide applicability of Woodward’s interventionism has become evident (see for example Waters 2007 or Suárez & San Pedro 2011), there is still some skepticism towards the suitability of some interventionist notions in specific cases (see Marcellesi 2010; Phyllis et al 2011; Russo 2012; Samaniego 2013; Reutlinger 2013; Baumgartner & Glynn 2013). This skepticism is, among other reasons, due to the lack of applicability of the interventionist notion of explanatory depth to cases where several causes are intertwined. Given the difficulties in applying interventionism to some relevant cases, one possible reaction would be to criticize and dismiss the theory. The alternative reaction, here adopted, is attempting to improve interventionism in such a way that it can account for the complex causal explanations postulated by scientists. The main objective of this paper is to improve interventionism by proposing a new notion of explanatory depth. It will be argued that this new notion is suitable to assess multi-causal patterns difficult to evaluate with Woodward’s original notion of explanatory depth. The content of the paper will be organized as follows: the central notions of interventionism will be presented in section 2, and the three criteria of explanatory depth will receive special attention. In the next section these three criteria will be applied to a particular case: the causal relationship between cholesterol

2

Chapter One

and atherosclerosis. In section 4 it will be discussed how the interventionist criteria of explanatory depth, as originally proposed in the interventionist theory, are not equally fulfilled in the cholesterolatherosclerosis case of study. This will lead us to propose (in sec.5) a new way of understanding the interventionist notion of explanatory depth as a three dimensional vector. Section 6 presents some final remarks.

2. The Interventionist Theory The essence of a causal explanation, according to the interventionist theory, consists in exhibiting a pattern of counterfactual dependence associated with relationships that are potentially exploitable for purposes of manipulation and control (see Woodward, 2003, pp. 13-16). In other words, a basic idea of interventionism is that causal scientific explanations do not aim simply at satisfying our intellectual curiosity, but are often guided by the goal of finding information potentially relevant for the manipulation and control of the explained events.

2.1. The notion of cause The very notion of cause in the interventionist theory is linked to manipulation. It is defined as follows: C is a genuine cause of the effect E if, given the appropriate background conditions, there is a possible manipulation of the cause C so that this is also a way of manipulating or changing the effect E (see Woodward 2003, sec.2.2, or Woodward 2008, sec.1). In other words, causal relations entail some changes upon the values of E whenever the values of C are modified. The changes of value performed over C must be reproducible in the sense that responses to the effect E must be in some way repetitive or systematic. The fact that a set of counterfactuals is associated with every causal relationship is also essential in the interventionist theory. When C is a cause of E, the associated counterfactuals will be of the following kind: “If C were manipulated by the intervention I, then E would experiment such and such changes”.

2.2. Formal definition of intervention The notion of intervention is formally defined in the interventionist theory as follows (IN) I assuming some value I =i is an intervention on C with respect to

Causation and the Interventionist Vector of Explanatory Depth

3

E if and only if I =i is an actual cause of the value taken by C, and I meets the following conditions: (IN-i) I must be the only cause of C; i.e., the intervention must completely disrupt the causal relationship between C and its previous causes so that the value of C is set entirely by I. (IN-ii) I should not itself be caused by any cause that affects E via a route that does not go through C. (IN-iii) I must not directly cause E via a route that does not go through C. (IN-iv) I leaves the values taken by any causes of E except those that are on the directed path from I to C to E (should this exist) unchanged (Woodward, 2010, sec. 5; and 2003, p. 98).1

2.3. The notion of invariance The notions of causation, explanation and invariance are closely intertwined in the interventionist theory. According to this theory, a generalization G—i.e. a relationship expressing a putative causal connection between two variables—counts as causal or explanatory if and only if it is invariant under some appropriate set interventions (see Woodward 2003, p. 15; p. 239). Like the notion of intervention, the notion of invariance is modal in the sense that it tells us whether a putative causal relationship would remain stable if, perhaps contrary to fact, certain changes or interventions were to occur. A generalization G relating changes in the cause C (from the value c to the value c’) to changes in the effect E (from e to e’) is invariant under a testing intervention I if and only if G correctly describes what the new value of E, e’, would be under this change; that is, if and only if it remains true that G (c’) = e’ (see Woodward 2003, sec. 6.2).

2.4. The notion of explanatory depth According to the interventionist theory, the import of a given explanation relies on its capacity to provide answers to counterfactual questions, i.e., questions about what would happen under circumstances different to the actual ones (from now on what-if-questions). And in satisfactory causal explanations, the patterns of dependence between causes and effects (or generalizations) are invariant under a set of interventions. The notion of explanatory depth is introduced in the interventionist theory using the following example (formulated by Haavelmo 1944, pp.

4

Chapter One

27-28). Suppose that two different explanations are offered about how to increase and decrease the speed of an automobile. The simplest explanation relates the speed to the distance of the gas pedal from the bottom of the car. And a second and more elaborated explanation details the whole inner mechanism of the car, tells us how the motor and the carburettor work and so on. Both explanations are successful in terms of providing sufficient information to operate the vehicle. Besides, both explanations are valid under the criteria of the interventionist theory because they identify patterns of counterfactual dependence that enable to control the speed of the automobile. In the first case, the postulated relation between speed and the gas pedal remains invariant under some interventions, for example, under changes in the values of the pedal inclination. However, this relation postulated by the simplest explanation fails if the car runs out of petrol, or if any element inside the car does not work properly. The second explanation, in contrast, is invariant under interventions on any part of the mechanism. And thus it answers a larger number of what-if-questions. Among other things, it explains why the car does not accelerate if the gas tank is empty. Therefore, in interventionist terms, the second explanation is deeper than the first one. The notion of explanatory depth is not properly defined in the interventionist theory in the sense that the theory does not provide us with a set of necessary and sufficient conditions for this notion. Nevertheless one may identify three criteria that Woodward’s interventionism takes into account in order to assess explanatory depth. I turn now to specify those three criteria. Criterion I. A causal explanation is explanatorily deep if the generalization figuring in the explanation is invariant under a wide range of interventions (see Woodward 2003, p. 311). Suppose we define the variables that figure in the explanation of the car’s acceleration as P=pressing the pedal, and A=acceleration. If the pedal is pressed then P=1, and if the pedal is not pressed P=0. Similarly, A=1 if the car accelerates and A=0 if the velocity of the car is constant. And suppose we have an alternative explanation that also appeals to the causal relation between P and A but, instead of defining P and A as bi-valued variables, it defines them as multi-valued variables. P can take any real value in the possible range of the pedal’s inclination, and A can take any value between 1m/s2 and 1000m/s2. Criterion I would then tell us that the explanation appealing to multi-valued variables is deeper than the explanation appealing to bi-valued variables, because it shows “how any

Causation and the Interventionist Vector of Explanatory Depth

5

one of a great number of changes in the explanans variables will lead to one of many possible changes in their explanandum variables. In other words, [it] gives us information about a much more detailed and finegrained quantitative pattern of counterfactual dependence than the ´binary´ pattern” (Woodward 2003, p. 206). This means that there are many more ways to intervene upon the multi-valued variable P than upon the bivaluated P. Criterion II. A causal explanation is explanatorily deep if the generalization figuring in the explanation is invariant, not only under a wide range of interventions (criterion I), but also under a wide variety of different kinds of interventions (see Woodward 2003, pp. 211-215). Criterion II may be illustrated again with the example of the car. In fact, Woodward (2003, pp. 259-260) uses this example to illustrate that the second and elaborated explanation is deeper than the simple one because the generalizations figuring in the elaborated explanation are invariant under a wider variety of different interventions. In addition to being able to intervene upon the pedal in many ways, the elaborated explanation allows us to perform many new interventions (upon the motor, the carburettor, the petrol tank, etc.) that were not considered in the first and simple explanation. Therefore, according to the interventionist theory, it is a deeper causal explanation of the car’s acceleration. This invariance under a variety of kinds of interventions is reflected, in turn, in the diverse and detailed counterfactuals associated to that explanation. According to the interventionist theory in order “to elucidate certain kinds of causal claims, including claims about direct causal relationships and singular causal claims, one must appeal to counterfactuals with detailed antecedents—counterfactuals that describe what will happen under combinations of manipulations or interventions, rather than under single manipulations” (see Woodward 2003, p. 21). Criterion III. A causal explanation is explanatorily deep if it is able to answer a wide range of counterfactual questions about the conditions under which the explanandum would have been different (see Woodward, 2003, p. 191). In other words, the deeper the explanation, the wider the range of “what-if-things-had-been-different” questions it answers (see Woodward 2003, p. 311).2 Again our original example of the car illustrates how the second explanation is deeper than the fist one. The second explanation is able to answer what would happen if the carburettor breaks, if the petrol does not flow from the tank to the motor and so on. Whereas the first explanation is

6

Chapter One

able to answer one single what-if-question, namely, what would happen if the pedal were not pressed. In the interventionist theory these three criteria seem to come along together. For example, Woodward comments: “A deeper explanation for the behaviour of the car would need to appeal to [generalizations and] engineering principles that are invariant under a much wider range of changes and interventions. Not coincidentally such a deeper explanation could be used to answer a much wider range of what-if-questions” (Woodward 2003, p. 260, my emphasis). I have stressed the expression “not coincidentally” because it lets us see that, according to the interventionist theory, a causal explanation that meets the first two criteria will also meet criterion 3. The following passage also suggests that the three criteria of explanatory depth come along together: “Some generalizations are not invariant under any (testing) interventions at all, and hence are non-explanatory. Other generalizations are invariant under some testing interventions (and answer some what-if-questions), and hence are above the threshold of explanatoriness, although they are less invariant and answer a narrower range of what-if-questions than others, and for this reason are less explanatory (Woodward 2003, p. 369). A consequence of a given explanation meeting criteria I, II and III, according to the interventionist theory, is that the explanation will be relevant to the manipulation and control of the explained event. And this is precisely what the second explanation of the car’s acceleration achieves. And, according to interventionism, this is what causal explanations should aim for. The interventionist theory also defines a minimal condition for explanatory depth: The minimal condition for successful causal explanation is to present a generalization G (relating the putative cause C to the effect E) and show that there is at least one intervention I under which G is invariant. It would be much better if we show that G is invariant under many interventions. But showing only one is sufficient for our explanation to be considered as minimally successful. If a generalization is not invariant under any intervention, then it will fail to qualify as invariant or explanatory. In that case both the generalization G and the putative explanation associated to G fall below what Woodward calls “the threshold of explanatoriness” (see Woodward 2003, p. 203; p. 368).

Causation and the Interventionist Vector of Explanatory Depth

7

3. Case study: cholesterol and atherosclerosis 3.1. Detailed causal explanation Let us analyze the causal relationship between cholesterol and atherosclerosis (also known as arteriosclerotic vascular disease or ASVD). Sterols are substances in our body that serve, for example, as material to build cellular plasmatic membranes, or for producing bile. Among all the different types of sterols, two are particular relevant factors for heart diseases: high-density lipoproteins and low-density lipoproteins. The low-density lipoproteins (LDL) are sterols of a very harmful type. LDL have low density because they are composed by low amounts of proteins and great amounts of lipids. If the LDL are not properly eliminated by the organism, they start accumulating in the walls of the blood vessels, forming atheromas. These atheromas may reduce or even obstruct the flow of blood, avoiding the proper oxygenation of the heart and the brain. Additionally, oxidized LDL molecules damage the vessels walls making them weaker. These are the reasons why LDL are often referred as “bad cholesterol”. “High-Density Lipoproteins” (HDL), on the contrary, are molecules highly dense in proteins, and poor in lipids, which help to prevent heart diseases. Most of the HDL in our body are produced in the liver and they are transformed in bile to be used in digestive processes. HDL also travel in the blood, collecting the cholesterol that cells did not use. Those HDL arrive to the liver and they are decomposed and recycled for bile production. For these reasons HDL are also known as good cholesterol. The algebraic addition of HDL and LDL is defined as total cholesterol TC. TC= HDL + LDL. And the following table (3-1) contains the recommendable levels of HDL, LDL and TC respectively (see Tudela 2000, pp. 33-34). Desirable concentration Limit of acceptable concentration High risk Very high risk Table 3-1

TC (mg/dl) TC < 200 200 < TC < 239 240 < TC

LDL (mg/dl) 100 < LDL < 129 130 < LDL < 150 160 < LDL < 189 190 < LDL

HDL (mg/dl) HDL = 45 The highest the concentration of HDL, the better

8

Chapter One

Using the information we have gathered so far, together with the possible values of the variables in table 3-1, someone could propose the following causal graph as a representation of the causal relationships among good cholesterol HDC, bad cholesterol LDC, total cholesterol TC, and atherosclerosis ASVD.

Fig.3-1. Causal Graph 1

Let us now imagine that, by means of the interventionist theory, we want to assess whether the putative causal relationship between the variables TC and ASVD is genuine. In order to do so, we shall firstly propose some possible manipulations to control the value of the postulated causal variable TD. Afterwards, we shall verify which of them fulfil the conditions IN (defined in sec 2.2). This will provide a set of interventions for testing the causal relationships postulated in the explanations under study. Once we have the set of testing interventions, we can proceed to analyse if the causal link between TC and ASVD remains invariant under the interventions. The higher the invariance under interventions, the deepest the explanation of heart disease caused by cholesterol would be. How can we manipulate the values of TC? A natural way to start manipulating TC values is altering the ingestion of specific ingredients that have been proven to increase or decrease cholesterol levels. The following table contains some examples (see Tudela 2000, pp. 51-71):

Causation and the Interventionist Vector of Explanatory Depth

Ingredients 1. Soya Milk and Orange juice 2. Asparagus and broccoli 3. Tomato 4. Margarine, eggs, and fat meat 5. Red wine 6. Almonds, peanuts, pistachios, etc. 7. Avocado 8. Cacao

9

Effects Reduce LDL The liver spends LDL producing the high amounts of bile required to digest these vegetables Inhibits LDL production Increase LDL Prevents LDL from sticking in the vessel’s walls; It also increases HDL They increase HDL for they contain monosaturated lipids. They avoid LDL oxidation. They must be consumed moderately Contains mono-saturated lipids, increases HDL It is a mono-saturated lipid, so it increases HDL. But it must be consumed as dark chocolate, or even better as grain, to avoid factors of obesity as whole milk and sugar

Table 3-2

Using table 3-2 several manipulations of TC can be defined by including different amounts of these ingredients in the patient’s diet. All such manipulations will refer to ingested cholesterol (exogenous cholesterol). Additionally, we can propose some manipulations of produced in the liver (endogenous cholesterol). The production of endogenous cholesterol is determined by the capacity of the liver’s cells to capture and decompose LDL molecules. There is a genetic mutation that prevents the liver from fabricating receptors of LDL molecules. As a consequence, even if the blood has a high level of cholesterol, the liver cells cannot perceive it, and they produce high amounts of HDL, causing, of course, an important race of the total cholesterol in the blood. Let me call this increment in the production of HDL due to a genetic modification “the mutation manipulation”. It is worth noting that, in fulfilment of criterion III, our knowledge about cholesterol also allows us to answer several what-if-questions: What would happen if the genetic constitution prevents the liver to absorb

10

Chapter One

cholesterol? What would have happened if the patient had not eaten eggs and bacon every morning? And so on. So, intuitively, given the proposed manipulations in endogenous and exogenous cholesterol, and given the answered what-if-question, it seems that we have collected a good amount of knowledge to provide a deep causal explanation of atherosclerosis due to levels of cholesterol. Let us see if our intuition is correct.

3.2. Interventionist analysis of the detailed causal explanation In order to prove the depth of our causal explanation, the interventionist theory indicates that firstly we must check that our manipulations fulfil the conditions IN (that is, that they count as interventions) and, secondly, we must show that the causal links in the causal graph 1 remain invariant under those interventions. The first difficulty we find is that the interventions in table 3-2 do not fulfil requirement IN-i. According to the requirement IN-i if we wiggle the values of HDL as an intervention upon TC, such intervention should be performed while disrupting the causal relationship between TC and any other previous causes (LDL). But, in this case, that is impossible as, by definition, HDL = TC – LDL, and this means that changing the values of HDL and TC, necessarily implies changing the value of LDL as well. This difficulty is however easy to solve by revising the variables selected in the causal graph 1. On the one hand, the causal variable TC is a function of two underlying factors: LDL, which is causal, and HDL, which is actually preventive. That is the reason why the manipulations proposed in table 3-2 turn out to be what Sprites and Scheines (2005) called “ambiguous manipulations”. On the other hand, the causal graph 1 seems to represent, wrongly, that LDL causes TC, or HDL causes TC. But the arrows from HDL or from LDL to TC are not causal links, but rather mathematical or definitional links, and this should be expressed in some way (see Woodward 2011, p. 24). A possible way to repair the mistaken assumptions in causal graph 1, is removing TC and keeping only the underlying factors HDC and LDL, and representing the definitional relationship between them by a double arrow (different from the simple arrows that represent causal relationships). The emended causal graph would be the following:

Causation and the Interventionist Vector of Explanatory Depth

11

Fig. 3-2. Causal graph 2

Once we accept that the causal graph 2 is correct, we proceed to check again if our manipulations in table 3-2 fulfil the conditions IN. But now a second difficulty appears: according to requirement IN-iv, the values of HDL should remain fixed while we wiggle the values of LDL, and vice versa. However, due to the intertwined relationship between HDL and LDL this seems impossible. Levels of HDL and LDL may influence each other. For example, high amounts of LDL in the blood may inhibit the production of HDL in the liver. Or, in a similar but opposite way, if LDL levels break down very drastically, the organism may perceive lack of cholesterol in the blood and an increase in the production of HDL. This interrelation between HDL and LDL levels seems to prevent all the dietetic manipulations to meet requirement IN-iv. The only manipulation that meets IN-iv is the mutation manipulation because, as the body itself stops perceiving LDL levels, the HDL production increases leaving the amount of LDL unaffected. The conclusion of interventionist analysis is that all our manipulations of exogenous cholesterol based on diet modifications fail to meet requirement IN-iv and thus they do not classify as proper interventions. The only manipulation that counts as a proper intervention is the “mutation manipulation” of endogenous cholesterol. Although the criterion III is widely fulfilled by answering several what-if-questions, the interventionist criteria I and II of explanatory depth are fulfilled only by one single intervention. The three interventionist criteria of explanatory depth work harmonically in the car’s example (in sec.2.4), but this does not occur in the cholesterol-atherosclerosis case. And it seems frustrating to accept that our explanation falls just above the threshold of explanatoriness. We

Chapter One

12

would like to claim that cholesterol levels due to diet modifications are also causally related to atherosclerosis, specially taking into account that there is a great amount of what-if-questions that can be answered using the list of ingredients in table 3-2. In the next section we will discuss what could possibly go wrong in our interventionist analysis. But, for the sake of the argument, let me first present a possible simplification of the causal explanation of atherosclerosis based on cholesterol levels.

3.3. Simplified causal explanation of atherosclerosis Imagine an absurd, but still possible, causal explanation of atherosclerosis such as one that ignores all knowledge about HDL, and ignores the mutation manipulation, and it is therefore completely based on knowledge about levels of LDL. The causal graph 3 bellow represents this of simplified causal explanation of atherosclerosis.

Fig. 3-3. Causal graph 3

In this imaginary simplified explanation, only a reduced subset of manipulations can be proposed. Out of the dietetic modifications proposed in table 3-2, only the first five can be used in this case. Nevertheless, and surprisingly, this time the interventions actually meet requirements IN because there is no other factor to fix while we intervene upon the variable LDL. Therefore, the causal relationship illustrated in the causal graph 3 turns out to be genuine, and consequently this simplified explanation fulfils the three interventionist criteria of explanatory depth for a number of interventions. Is it not paradoxical that this explanation is evaluated as deeper than the detailed explanations presented before?

4. Discussion Our analysis shows that the criteria of explanatory depth, as originally defined in interventionism, may lead to counterintuitive results. The result that the simpler explanation, which deliberately ignores a relevant causal factor, is deeper than the detailed explanation is clearly wrong, and is a counterintuitive result from the interventionist perspective itself. The detailed explanation, which accounts for a wider set of counterfactual questions, should be deeper as it happens in the example of the car’s acceleration.

Causation and the Interventionist Vector of Explanatory Depth

13

At this point one could simply assume that Woodward’s interventionism is not applicable to cases with multiple correlated causes. One could thus dismiss interventionism and look for alternative theories of causation. However, there are at least two motivations to avoid this negative reaction and, instead, continue defending interventionism. Firstly, interventionism is a practical tool to test causal relations and, maybe more importantly, it is able to tell us how to modify the scientific explanations in order to get a better control over the effects. Interventionism allows detecting genuine causal links, and identifying deeper explanations. In our case study, for example, the interventionist theory has pointed at the fact that controlling levels of total cholesterol TC was not as good a strategy to prevent atherosclerosis, as controlling the underlying factors LDL and HDL. In other words, interventionism is able to point at the factors that doctors and patients should be measuring, and it guides us to a better understanding of diseases. This fact, by itself, would be already worthy to continue using the interventionist theory. Besides, given that the majority of diseases have multiple causes that influence each other’s occurrence in several ways (see van Loo et al. 2012), having a theory of causation for assessment of this kind of cases is necessary. Secondly, Woodward (2003, p. 132; 2011) has a flexible attitude towards possible changes and improvements of interventionism. The challenge for interventionism brought about by our case study is whether the interventionist theory is capable of telling us something about the causal link between a given effect E and its putative causes C1, C2, Cn, when those causes are correlated to each other. Even in those cases, Woodward believes that we can legitimately use counterfactuals to elucidate causal claims along the lines suggested in the interventionist theory.In cases like this, Woodward (2003, sec. 5.11) suggests to verify whether two conditions—indeed two original motivations for introducing the notion of intervention– are being met or not: Firstly, we must ensure that “there is a basis for claims about what will happen to E under an intervention on C”; i.e. we should be able to associate some well-defined notion of change with C, and we have some grounds for saying what the effect, if any, on E would be of changing just C and nothing else. And secondly, there must be “a way of disentangling the effect on E of changing just C from the effects on E of changes in other potentially confounding variables” (see Woodward 2003, pp. 131-2). In a more recent paper (2011), Woodward also shows his disposition to modify and improve interventionism. This has been a strong incentive to propose here an attempt to make interventionism more suitable for cases where it seems difficult to apply it.

14

Chapter One

My claim is that, re-defining and understanding in a new way the interventionist criteria of explanatory depth, the counterintuitive results of the kind we obtained in section 4.1, can be avoided. This proposal is meant to complement Woodward’s efforts to improve interventionism. The new interventionist vectorial notion of explanatory depth, which will be developed in the next section, aims to make the interventionist theory applicable in a more natural way to multi-causal explanations in the social, natural and health sciences.

5. Proposal: Re-defining notion of explanatory depth The aim of this section is to propose a new way of understanding the interventionist notion of explanatory depth that results more suitable for multi-causal explanations. The key conceptual modification of the notion of explanatory depth, embedded in the new vectorial proposal, consists in assigning much more importance to counterfactual questions. In Woodward’s original proposal, invariance under intervention was primordial. In the proposal here presented, counterfactual answers are almost as important as invariance under intervention. There is a single sense in which interventions continue being more important than counterfactual answers: in order to cross the explanatoriness threshold, invariance under at least one intervention is still required.3 However, once the threshold has been passed, the amount of counterfactual answers is as significant as the amount of possible interventions under which the causal link postulated in the explanation remains invariant. In order to test the explanatory depth of a given causal explanation, following steps is recommended: 1st: Find a variety of manipulations of the values of the putative causes, and verify if those proposed manipulations fulfil requirements IN. If, for any reason, the manipulations fail to meet the requirements, modifications in the relevant set of variables, possible definitional relations among the variables, and consequent modifications in the causal graph must be considered. 2nd: Once the relevant variables, the causal graph, and the manipulations upon the putative causes are chosen, it must be ensured that the generalization figuring in the explanations remains invariant under at least one intervention, crossing the threshold of explanatoriness. 3rd: Build a vector whose component X is the number of different kinds of manipulations upon the causes, whose component Y is the number of proper interventions under which the generalization remains invariant, and

Causation and the Interventionist Vector of Explanatory Depth

15

whose component Z is the number of what-if questions answered. In other words, each of the original criteria of explanatory depth I, II and III is a component of this vector, which we will name “the interventionist vector of explanatory depth”. Applying this vectorial notion of explanatory depth, we not longer arrive to the counterintuitive result that the simplified causal explanation of atherosclerosis (sec 3.3) is deeper than the detailed explanation (sec 3.1). The mutation manipulation places both explanations above the threshold of explanatoriness. However, the detailed explanation presents three different types of manipulations (endogenous cholesterol, exogenous cholesterol and mutation), while the simple explanations can only make use of the last two. Furthermore, the fact that the detailed explanation is able to answer much more what-if-questions makes the magnitude of its explanatory vector much larger than the vector corresponding to the simplified explanation. Therefore, the vectorial notion of explanatory depth leads us to a correct causal assessment of the two explanations of atherosclerosis, i.e., it leads us to the conclusion that the detailed explanation is deeper than the simplified one. The vectorial notion proposed in this paper may seem, at first sight, a simple change from verbal enunciation to mathematical representation of the criteria. However, the new vectorial notion essentially affects the notion of explanatory depth for it replaces necessity by sufficiency regarding criterion II and criterion III.

6. Conclusions Interventionism not only provides an innovative methodology to test causal explanations, but also helps to improve the causal explanations already existent in different areas of knowledge, and has a practical pay off when it elicits how to manipulate the effects that are relevant for different reasons. As an example of this, in section 3.2, interventionism allowed us to identify that a causal explanation of atherosclerosis based on HLD and LDL levels, was more convenient than a causal explanation based on the total cholesterol (TC). Therefore, interventionism provides us with specific variations and proposals to improve causal explanations. This is already a good reason to defend Woodward’s interventionist theory. According to the original definition of explanatory depth, the detailed explanation would be just above the threshold of explanatoriness. The simplified explanation that ignores a lot of relevant factors of atherosclerosis

16

Chapter One

would be considered deeper than the detailed explanation. And this is a counterintuitive and incorrect result. The advantage of the re-definition of the interventionist notion of explanatory depth, here proposed, is that it allows us to understand why some explanations are deeper than others, even though the number of interventions is reduced due to the correlations among different causes. This vectorial notion of explanatory depth aims to be particularly fruitful for cases in health sciences, where the majority of diseases present complex multi-causal patterns. Nevertheless, it also aims to be suitable for multi-factorial causal explanations provided in all kind of scientific practices including social sciences and foundations of physics (see Samaniego 2013). This proposal supports the lines of defending and improving interventionism, and hopefully, helping scientists to reach a better understanding and control of relevant causal factors.4

References Baumgartner, M. and L. Glynn. 2013, “Introduction to Special Issue on ‘Actual Causation’”, Erkenntnis, http://link.springer.com/article/10.1007/s10670-013-9441-8 Haavelmo, T. 1944, “The Probability Approach in Econometrics”, Econometrica 12, 1 Marcellesi, A. 2010, “L’interventionnisme permet–il la causalité “descendante”?” Igitur, Arguments philosophiques 2 (1), http://www.igitur.org/L-interventionnisme-permet-il-la Menzies, P. and H. Price 1993, “Causation as a Secondary Quality”, British Journal for the Philosophy of Science 44, 187 Phyllis, I., F. Russo and J. Williamson (eds.) 2011, Causality in the Sciences, Oxford: Oxford University Press Reutlinger, A. 2013, A Theory of Causation in the Social and Biological Sciences, United Kingdom: Palgrave Mcmillan. Russo, F. 2012, “Explaining Causal Modelling. Or, What a Causal Model Ought to Explain” in M. D’Agostino et al (eds.) New Essays in Logic and Philosophy of Science, London: College Publications Samaniego, F. 2013, “Causality and Intervention in the Spin-Echo Experiments”, Theoria 21 (3), 477 Spirtes, P. and R. Scheines 2005, “Causal Inference of Ambiguous Manipulations”, Philosophy of Science 71, 833 Suárez, M. and I. San Pedro 2011, “Causal Markov, Robustness and the Quantum Correlations” in Probabilities, Causes and Propensities in Physics, Synthese Library 347, 173, Berlin and New York: Springer

Causation and the Interventionist Vector of Explanatory Depth

17

Tudela, V. 2000, El Colesterol: lo bueno y lo malo, Mexico: Fondo de Cultura Económica van Loo, H.M., J.W. Romejin et al. 2012, “Psychiatric comorbidity and causal disease models”, Preventive Medicine 57 (6), 748 von Wright, G. 1971, Explanation and Understanding, New York: Cornell University Press. Waters, K. 2007. “Causes that Make a Difference”, The Journal of Philosophy 104 (11), 551 Woodward, J. 2003, Making Things Happen: A Theory of Causal Explanation, Oxford: Oxford University Press. —. 2008, “Causation and Manipulability”, The Stanford Encyclopedia of Philosophy (Winter 2008 Edition), Edward N. Zalta (ed.), URL =

—. 2010, "Scientific Explanation", The Stanford Encyclopedia of Philosophy (Spring 2010 Edition), Edward N. Zalta (ed.), URL =

—. 2011, “Interventionism and Causal Exclusion”, [Preprint] URL: http://philsci-archive.pitt.edu/8651/

Notes 1 It is worth noting that human limitations or capacities are not mentioned at all, meaning that interventions (IN) may perfectly be natural or hypothetical. 2 For short, form now onwards I will refer to this kind of counterfactual questions as “what-if-questions”. 3 The reason to keep the minimal requirement of explanatory depth is to avoid problematic causal relationships due to common causes (see the barometer case in Woodward 2003, sec.1.4). 4 I am deeply grateful to Alexandre Marcellesi, Atocha Aliseda, Ana Rosa PérezRansanz, León Olivé, Mauricio Suárez, Federica Russo, Bert Leuridan, and José Luis Vázquez for their support and bibliographical recommendations. I’m especially grateful to Janneke van Lith for comments on my previous work that became particularly helpful for the development of this article. I would finally like to thank the Mexican Council of Science and Technology (CONACYT) for the Postdoctoral Fellowship CVU-166476 (2012-2014).

CHAPTER TWO EXPLANATORY DEPTH, INFERENCE TO THE BEST EXPLANATION AND PROPORTIONALITY ALEXANDRE MARCELLESI

1. Introduction It is a truism that not all causal explanations are created equal and that some are better than others. But what are the factors the quality of a causal explanation depends on? It is to answer this question that James Woodward and Christopher Hitchcock (Hitchcock and Woodward 2003; see also Woodward 2003, Chapter 7) have developed an account of what they call “explanatory depth”. As I will explain below, this account is based upon the interventionist account of causal explanation developed by Woodward (2003, Chapter 5; see also Woodward and Hitchcock 2003). According to the Woodward-Hitchcock (or “WH”, for short) account of explanatory depth, causal explanations are deeper, better or more powerful—Woodward and Hitchcock take the expressions “explanatory depth”, “explanatory quality” and “explanatory power” as synonyms—the more invariant the generalizations they involve are. For interventionists, causally explaining an event is “a matter of exhibiting systematic patterns of counterfactual dependence” (Woodward 2003, p. 191) between this event and its causes. And the more invariant a generalization is, the better it helps us exhibit such patterns. After some expository work in Section 2, I will argue in Section 3 that depth, as the WH account defines it, is just predictive power by another name and therefore is not a properly explanatory notion. I will also argue that it is not, as a result, a notion advocates of Inference to the Best Explanation should appeal to, despite what some have suggested (see e.g. Harker 2012 or Mackonis 2013). Then, in Section 4, I will argue that the WH account conflicts with the view that causal explanations are better

Explanatory Depth, Inference to the Best Explanation and Proportionality

19

when they cite causes that are “proportional” to their effects, a view many—including Woodward himself—find plausible. The conclusion I will draw from these arguments is that, if what one is looking for is an adequate account of the factors the quality of causal explanations depends on, then one should reject the WH account.

2. The WH account of explanatory depth In order to introduce the WH account of explanatory depth, I must first briefly present the interventionist account of causal explanation it is based on. This account is primarily an account of event explanation (I will use “event”, “phenomenon” and “fact” interchangeably in what follows), and the explananda it targets are of the form “The event of variable Y taking value y” or, more compactly, “Y = y”.1 As I mentioned above, causal explanation is, for interventionists, a matter of exhibiting patterns of counterfactual dependence between the explanandum and its causes. One does so by answering “what-if-things-had-been-different” questions, or “w-questions”, about the explanandum. What are w-questions? Consider two binary variables X and Y each taking either value 1 or value 0 (it does not matter here what these variables represent). Assume that, in the actual world, it is the case that X = 1 and Y = 1. Assume also that one is interested in explaining the event Y = 1. The following is a w-question with respect to this event: (WQ) What would the value of Y have been had the value of X been 0 instead of 1 as the result of an intervention? An intervention on X with respect to Y is a manipulation that results in a change in the value of X and has an effect on the value of Y, if at all, only via its effect on the value of X. The intuition underlying the appeal to interventions is that if a change in the value of X that is induced by a manipulation that has no independent effect on the value of Y is followed (temporally) by a change in the value of Y, then the causal “responsibility” for the change in the value of Y can be attributed to the change in the value of X. The details of Woodward’s definition of “intervention” are not important for the purpose of this paper and so I omit them here (but see Woodward 2003, p. 98). There are two possible answers to (WQ), assuming the relationship between X and Y to be deterministic:

Chapter Two

20

(WA1) Had the value of X been 0 instead of 1 as the result of an intervention, then the value of Y would have been 0 instead of 1. (WA2) Had the value of X been 0 instead of 1 as the result of an intervention, then the value of Y would have been 1. Let me call “w-answers” those answers to w-questions which, like (WA1), relate changes in one variable to changes—by contrast with an absence of changes—in another variable. As I am about to explain, interventionists require that generalizations entail at least one w-answer in order to be causally explanatory.2 Consider the following generalization relating our binary (0, 1) variables X and Y, assuming again that the actual value of both is 1 and that the explanandum is the event Y = 1: (G) Y = X G—and, more generally, equations of the form Y = f(X)—should be read as implicitly prefixed with a universal quantifier ranging over values of X. What G says, then, is that for any value x of X, Y takes a value y that is such that x = y. In other words, G says that Y = 1 whenever X = 1 and that Y = 0 whenever X = 0.3 According to interventionists (see e.g. Woodward and Hitchcock 2003, p. 6), G causally explains event Y = 1 if and only if it is both: (i) true of the actual values of X and Y, and (ii) invariant under at least one possible testing intervention on X. A generalization that satisfies conditions (i) and (ii) will entail at least one w-answer, as I will illustrate below. A generalization might, of course, entail more than one w-answer, a feature the WH account of explanatory depth will exploit. Note that the label “invariant” is often used to refer to generalizations that satisfy conditions (i)-(ii) and I will sometimes follow this custom for the sake of brevity. What would it mean for G to satisfy conditions (i) and (ii)? Start with condition (i), the condition requiring that G be “true [...] of the actual values” of the variables it relates, as Woodward and Hitchcock put it (2003, p. 6). Here, since I have assumed that the actual value of both X and Y is 1, G is indeed true of the actual values of the variables it relates (or just “true”, for short) and it therefore satisfies condition (i). What about

Explanatory Depth, Inference to the Best Explanation and Proportionality

21

condition (ii)? Testing interventions are relative to generalizations. A testing intervention on X with respect to Y relative to G is an intervention which sets X to a non-actual value and is such that G predicts that, under this intervention, Y will take a non-actual value (Woodward 2003, p. 253). An intervention that sets X = 0 thus is a testing intervention on X with respect to Y relative to G, since 0 is a non-actual value of X and since G predicts that, when X = 0, Y also takes a non-actual value, namely 0. If one assumes that, if one were to set X = 0 by an intervention, then Y would in fact take value 0, then G is invariant under such a testing intervention, since G is true when both X and Y take value 0. To say that G is invariant under an intervention setting X = 0 here is simply to say that it would remain true were such an intervention to occur. And if G is invariant under a testing intervention setting X = 0, then it entails some w-answer— namely (WA1)—and thus causally explains event Y = 1.4 Now that I have presented the essential elements of the interventionist account of causal explanation—it can easily be extended to multivariate generalizations and to the indeterministic case—it is time to turn to the WH account of explanatory depth that is based upon it. Explanatory depth is primarily a property of invariant generalizations—by which I mean, remember, generalizations that satisfy conditions (i) and (ii)—and only derivatively a property of the causal explanations these generalizations are involved in. What determines how deep an invariant generalization is? The basic idea is straightforward: As Woodward puts it, “other things being equal, relationships [i.e. generalizations] that are more invariant [...] provide better [i.e. deeper] explanations.” (Woodward 2003, p. 243) Matters become more complex, however, when one tries to spell out what the “more” in “more invariant” means. Woodward (2003) distinguishes three ways a generalization G might be “more” invariant than another generalization G' with respect to an explanandum Y = y while Hitchcock and Woodward (2003) distinguish seven. Thankfully, it does not matter for the purpose of this paper whether, and how, these two versions of the WH account of explanatory depth may be reconciled. What matters here is that, however one understands the details of this account, the depth of an invariant generalization is a function of the w-answers it entails (note: not of the number of w-answers it entails). Two generalizations G and G' that entail the exact same wanswers about some explanandum Y = y are equally deep relative to this explanandum. Moreover, if the set of w-answers entailed by G' is a proper subset of the set of w-answers entailed by G, then G is deeper than is G' relative to Y = y (see Woodward 2003, pp. 260-262). It is important to note that the WH account is an account of comparative—not absolute—

22

Chapter Two

explanatory depth. It thus does not allow for judgments of the form “Generalization G is deep to degree d relative to explanandum Y = y” but only for judgments of the form “G is deeper than G' relative to Y = y”. Let me assuage a worry that might arise from the many caveats Woodward and Hitchcock attach to their account of explanatory depth. As they are keen to emphasize, “explanatory depth is not one-dimensional” (Hitchcock and Woodward 2003, p. 188) but, rather, is a “complicated and multidimensional” notion, as Woodward (2003, p. 265) puts it. One might be tempted to read these claims as implying that the depth of a generalization is a function of other factors besides the w-answers it entails. This interpretation, however, is misguided: Each one of the dimensions referred to by Woodward and Hitchcock corresponds to one of the seven ways a generalization might be “more” invariant distinguished in (Hitchcock and Woodward 2003). And all of these make the depth of a generalization a function of the w-answers it entails. As Brad Weslake puts it in a recent paper, these seven ways of greater invariance are all ways in which “a generalization can provide the resources to describe a greater range of true counterfactuals concerning possible changes to the system in question—that is, to answer more w-questions.” (Weslake 2010, p. 278, emphasis original) It should be noted, moreover, that if this was not the case—i.e. if a generalization’s depth did depend on other factors than the w-answers it entails—then there would be no such thing as the WH account of explanatory depth. Woodward and Hitchcock would have given us, at best, an account of but one component of explanatory depth, which is clearly not what they take themselves to be doing. Now that the WH account of explanatory depth has been introduced, let me turn to two arguments to the effect that it is inadequate as an account of the factors the quality of causal explanations depends on.

3. Explanatory depth and inference to the best explanation Advocates of Inference to the Best Explanation (IBE) hold the view that, when faced with a set of incompatible hypotheses H1, ..., Hn all of which fit the available empirical evidence E equally well (or are empirically equivalent), one should infer to the truth of hypothesis that best explains this evidence (see e.g. Harman 1965, pp. 90-91). A fullfledged defense of IBE requires that one specify the criteria by which to assess the relative explanatory qualities of H1, ..., Hn. In other words, it requires an account of explanatory virtues. David Harker (2012) has recently suggested that explanatory depth, as defined by the WH account, is an explanatory virtue advocates of IBE can appeal to (see also Mackonis

Explanatory Depth, Inference to the Best Explanation and Proportionality

23

2013).5 Harker’s concern is to identify what he calls “peculiarly explanatory virtues” (Harker 2012, §1)—i.e. virtues of hypotheses that differ from their theoretical virtues (simplicity, fruitfulness, predictive power, ontological heterogeneity, etc.)—and this in order to prevent IBE from collapsing into the view that, when faced with empirically equivalent incompatible hypotheses, one should infer to the truth of the hypothesis that possesses the “best mix” of these theoretical virtues. Harker’s endeavor seems sensible: If IBE is to be a distinct mode of inference, then one must be able to identify peculiarly explanatory virtues. Unfortunately, explanatory depth as defined by the WH account is not such a virtue. It is simply predictive power in disguise, contrary to what Harker (2012, §3) himself claims. I will substantiate this claim below, but let me here note two of its implications: First, and obviously, if this claim is true then the notion the WH account is an account of cannot be appealed to by advocates of IBE. Second, if this notion is just predictive power, then why accept Woodward and Hitchcock’s claim that it is an explanatory notion? IBE in fact provides us with a guide, albeit an imperfect one, for assessing whether a virtue of hypotheses or generalizations is explanatory. Can advocates of IBE appeal to this virtue without IBE thereby collapsing into some other mode of inference, e.g. inference on the basis of theoretical virtues? If yes, then we have a prima facie reason to classify this virtue as explanatory. If not, however, then we have a prima facie reason not to classify this virtue as explanatory. Let me provide an illustration by way of evidence for the claim that depth, as the WH account defines it, is just predictive power. Consider two continuous variables W and Z—which might for instance represent two physical quantities—and the two following generalizations relating them, where f  g: (G1) W = f(Z) (G2) W = g(Z) Assume now that we have a data set composed of 50 observations {(z1, w1), ..., (z50, w50)} and that both G1 and G2 are “true” of these observations, i.e. that wi = f(zi) = g(zi) for any i ‫[ א‬1, 50]. Given these two assumptions, G1 and G2 are empirically equivalent. At this point, neither G1 nor G2 causally explains any of the 50 observations according to the interventionist account of causal explanation. In order for these generalizations to causally explain one of these observations, they must both entail at least one w-answer regarding

24

Chapter Two

the observation in question. Let me here stipulate that the target explanandum is W = w1. If one also assumes both the truth of the following counterfactual: (1) Had the value of Z been z51 instead of z1 as the result of an intervention, then the value of W would have been w51 instead of w1. and also that w51 = f(z51) = g(z51), then both G1 and G2 are invariant under at least one possible testing intervention. As a result, they both entail one w-answer, namely (1), and so causally explain the event W = w1.6 But notice that the fact that both G1 and G2 entail (1) simply means—as a matter of definition—that they both accurately predict what would happen were the value of Z set to Z = z51 by an intervention.7 We have yet to observe a situation in which Z takes value z51 as the result of an intervention. And both G1 and G2 tell us—here, accurately—which value of W we should expect to observe in such circumstances. This is just prediction. The step to the claim that depth, as defined by the WH account, is just predictive power is a short one. Assume the truth of the following counterfactual: (2) Had the value of Z been z52 instead of z1 as the result of an intervention, then the value of W would have been w52 instead of w1. And assume also that w52 = f(z52)  g(z52). Given these two assumptions, G1 entails w-answer (2) but G2 does not. The set of wanswers entailed by G2 thus is a proper subset of the set of w-answers entailed by G1. As a result, G1 is explanatorily deeper than G2 according to the WH account, i.e. the former provides us with a better causal explanation of event W = w1 than does the latter. But this simply means that G1 provides us with more true predictions regarding the value of W— and the way it changes under interventions on Z—than does G2. G1 is deeper than G2 simply in virtue of the fact that it correctly predicts the value W would take under an intervention setting Z = z52 while G2 does not, i.e. simply in virtue of the fact that it has greater predictive power. Here is how Harker describes what he calls the “more ambitious” version of the strategy that consists in advocates of IBE appealing to depth as defined by the WH account: “If pursuing invariance helps us achieve deeper explanations, for example, and deeper explanations indicate a more truthlike theory, then we connect a distinctively explanatory virtue to perhaps the ultimate scientific achievement.” (Harker 2012, §3) The fact

Explanatory Depth, Inference to the Best Explanation and Proportionality

25

that depth, as the WH account defines it, is just predictive power by another name is both good and bad news for those who hope to follow this strategy. It is good news because the connection between depth thus defined and truthlikeness is seemingly straightforward. Remember that generalizations G1 and G2 are to be read as implicitly prefixed with a universal quantifier ranging over values of Z. The fact that G1 entails wanswers (1) and (2) while G2 only entails (1) means that G1 is true of more values of Z than is G2. While G1 is true of 52 values of Z, G2 is true of 51. In this sense, G1 is closer to the truth than is G2, i.e. it is closer to being true of all the values of Z. It should also be clear that, in this sense of “truthlike”, the deeper a generalization is, the more truthlike it is. Mechanically, greater depth guarantees greater truthlikeness, but trivially so: Greater depth simply is greater truthlikeness. The bad news for advocates of IBE who wish to appeal to depth as defined for the WH account is obvious: As I noted above, if IBE is the view that, when faced with empirically equivalent incompatible hypotheses (or generalizations), one should infer to the truth of the deepest one, then IBE simply collapses into the view that one should infer to the truth of the hypothesis that is closest to the truth. It is presumably true, if not trivially true, that, when faced with empirically equivalent incompatible hypotheses, one ought to infer to the truth of the one that is closest to the truth (at least if one is a scientific realist). But IBE was supposed to give us an independent purchase on truthlikeness via explanatory depth or power. It should be clear that, if one weds IBE and depth as defined by the WH account, then IBE cannot deliver on its promise. As I indicated above, there is a broader conclusion to draw here: If depth as the WH account defines it is just predictive power, then why think that this account is adequate as an account of the factors the quality of causal explanations depends on? And why think that the interventionist account the WH account of depth is based on is properly seen as an account of causal explanation? To be sure, the predictions entailed by invariant generalizations are predictions of a specific kind, namely predictions regarding what would happen under various interventions. But predicting what would happen under various interventions is still predicting. I am not here claiming that there is no relationship at all between prediction and explanation. I am simply questioning the view adopting the WH account commits one to, namely the view that causal explanation and what one might call “causal prediction” are one and the same thing.8 A detailed critical examination of this identification is beyond the scope of this paper. But if you think—as most contemporary

26

Chapter Two

philosophers of science do—that prediction and explanation are conceptually distinct, then you should have serious doubts regarding Woodward and Hitchcock’s claim to have provided accounts of causal explanation and explanatory depth. I hope to have done enough here to convince the reader (i) that there are good reasons to be skeptical of Woodward and Hitchcock’s claim to be providing accounts of explanatory, as opposed to predictive, notions and (ii) that, as a result, depth as the WH account defines it is not a virtue advocates of IBE should appeal to. Let me now turn to a distinct shortcoming of the WH account of depth, one that subsists even if one bites the bullet and follows Woodward and Hitchcock in identifying causal explanation with causal prediction and explanatory depth with predictive power.

4. Explanatory depth and proportionality Many philosophers—including Woodward himself (see e.g. Woodward 2010)—believe that causal explanations are better when they cite causes that are “proportional”, in the sense of (Yablo 1992, 277), to their effects. Indeed, some—e.g. Christian List and Peter Menzies (2009)—even think that proportionality is necessary for causation. The case canonically used to introduce Yablo’s notion of proportionality is that of Sophie, a pigeon who pecks at all and only red targets (Yablo 1992, 257). Imagine that, on some occasion, I present Sophie with a crimson target and, crimson being a shade of red, she pecks at it. What is it that caused Sophie to peck at the target? There are at least two ways to answer this query for a causal explanation. The first is to say that the cause of Sophie’s pecking was the target being red. And the second is to say that the cause was the target being crimson. According to advocates of proportionality, the explanation that mentions the target being red as the cause of Sophie’s pecking is the better one because it cites a cause that is proportional to its effect. The explanation that cites the target being crimson cites a cause that is too “specific” for its effect since, by assumption, Sophie would have pecked at the target had it been red but non-crimson.9 To put it briefly, a cause is proportional to its effect in Yablo’s sense just in case it is both necessary and sufficient for its effect. Being presented with a red target is both necessary and sufficient for Sophie to peck since, as I stipulated above, Sophie pecks at all and only red targets she is presented with. Being presented with a crimson target, by contrast, is sufficient but not necessary for Sophie to peck, since she will peck at red but non-crimson targets.

Explanatory Depth, Inference to the Best Explanation and Proportionality

27

What does all this have to do with explanatory depth as defined by the WH account? As I will show below, there are cases in which the WH account implies that a generalization citing a cause that is not proportional to its effect is just as deep as a generalization that does cite a proportional cause. As a result, one cannot consistently endorse both the WH account of depth and the view that causal explanations are better when they cite causes that are proportional to their effects. Suppose that you are seeking to causally explain the fact that some sample of water located in a room the temperature of which is below 0°C is in a frozen state (this case is adapted from Craver 2007, pp. 205-206). And assume that the two following generalizations are available: (G3) S = C (G4) ܵ ൌ ቄ

Ͳiffܶ ൑ Ͳ ͳ iffܶ ൐ Ͳ

Here S is a binary variable taking value 0 when the sample of water is frozen and value 1 otherwise (i.e. if it is either liquid or gaseous). Variables C and T embody alternative ways of representing the temperature of the room the sample of water is located in. C is a binary variable taking value 0 when the temperature is less than or equal to 0°C and value 1 when it is strictly greater than 0°C. And variable T is a discrete variable taking its values in Է and representing the temperature of the room on the Celsius scale. I assumed above that the sample of water of interest is in a frozen state. This means that the actual value of S is 0. Let me also stipulate that the temperature in the room is -19.6°C, so that the actual values of C and T are 0 and -19.6, respectively. Both G3 and G4 are thus true of the actual values of the variables they relate. Assume now that the two following counterfactuals are true: (3) Had the value of C been 1 instead of 0 as the result of an intervention, then the value of S would have been 1 instead of 0. (4) Had the value of T been 10 instead of -19.6 as the result of an intervention, then the value of S would have been 1 instead of 0. The truth of (3) straightforwardly implies that G3 is invariant under at least one testing intervention. As a result, G3 entails a w-answer, namely (3) itself, and therefore causally explains the event S = 0. The truth of (4)

Chapter Two

28

likewise implies that G4 is invariant under at least one testing intervention. G4 thus entails w-answer (4) and causally explains the event S = 0. As should be obvious from the way C and T were defined above, the following relations hold between values of these two variables: (B1) C = 0 ļ T ” 0 (B2) C = 1 ļ T > 0 Given B1 and B2, any w-answer entailed by G3 will also be entailed by G4, and vice-versa. Imagine that somebody asks you “What would the value of S be were T to be set to 10 by an intervention?” Since, by B2, any intervention that sets T = 10 also sets C = 1, you can answer this wquestion without knowing G4, simply with the help of G3. And if you rely on G3 in order to answer this query, the answer you will give will be none other than w-answer (4), since G3 tells you that S = 1 when C = 1. Likewise, you can answer the w-question “What would the value of S be were C to be set to 1 by an intervention?” without knowing G3, simply with the help of G4. The answer you will then give will be w-answer (3), since any intervention that sets C = 1 also sets T to a value > 0 and since G4 tells you that S = 1 when T > 0. Thanks to the relationships holding between values of C and values of T, generalizations G3 and G4 thus entail exactly the same w-answers with respect to explanandum S = 0. According to the WH account, this means that they provide us with equally good causal explanations of the fact that the sample of water is frozen. The problem here is that, for advocates of the view that causal explanations are better when they cite causes that are proportional to their effects, G3 provides us with a better causal explanation of the fact that S = 0 than does G4. This is because G3 cites a cause that is proportional to its effect while G4 does not. The cause cited by G4, namely the exact temperature of the room the sample of water is located in, is intuitively too “specific” insofar as the target explanandum is the relatively coarse-grained fact that the sample of water is frozen. The analogy with the case of Sophie, Yablo’s pigeon, is straightforward. In Yablo’s case, we know—because we stipulated it—that Sophie will peck if and only if the target is red. In the case at hand, we know that the sample of water will be frozen if and only if the temperature in the room is less than or equal to 0°C. To cite the precise shade of the red target Sophie was presented with, namely crimson, as the cause of her pecking is to cite a cause that is too “specific” because Sophie would have pecked at a red but non-crimson target. And, here, to cite the precise temperature of the room,

Explanatory Depth, Inference to the Best Explanation and Proportionality

29

namely -19.6°C, is to cite a cause that is too “specific” because the water would have been frozen had this temperature been different from -19.6°C as long as it was less than or equal to 0°C. The WH account and the proportionality view thus yield conflicting verdicts. According to the former, G3 and G4 provide us with equally good causal explanations of the fact that the sample of water is frozen. And according to the latter, G3 provides us with a better causal explanation than does G4. One thus cannot consistently endorse both the WH account and the proportionality view. If you think, as many (but by no means all) philosophers do, that the proportionality view is correct—and that G3 provides us with a better causal explanation because it tells us exactly what we need to know, no more and no less, to causally explain the coarse-grained fact that S = 0 while G4 gives us more than we need—then the argument developed above, if sound, provides you with a good reason to reject the WH account of explanatory depth.

5. Conclusion I have argued above that depth as the WH account defines it is not an explanatory notion but, rather, is simply predictive power by another name. As I explained, this means that this notion is not one advocates of IBE looking for what Harker calls “peculiarly explanatory virtues” should appeal to. I have also argued that the WH account conflicts with the popular view that causal explanations are better when they cite causes that are proportional, in Yablo’s sense, to their effects. If I am right that depth as the WH account defines it is just predictive power, then this is hardly surprising. If proportionality is a norm governing causal explanation while depth as defined by the WH account is a norm governing what I have called “causal prediction” then, given that prediction and explanation are conceptually distinct activities, there is no reason to expect these two norms to agree. If what you are looking for is an adequate account of the factors the quality of causal explanations depends on, then, there are good reasons for you to reject the WH account of explanatory depth.10

References Craver, C. 2007, Explaining the Brain, New York: Oxford University Press Harker, D. 2012, “Inference to the Best Explanation and the Importance of Peculiarly Explanatory Virtues”, http://philsci-archive.pitt.edu/9278/

30

Chapter Two

Harman, G. 1965, “The Inference to the Best Explanation”, Philosophical Review 74 (1), 88 Hempel, C. and P. Oppenheim 1948, “Studies in the Logic of Explanation”, Philosophy of Science 15 (2), 133 Hitchcock, C. and J. Woodward 2003, “Explanatory Generalizations, Part 2: Plumbing Explanatory Depth”, Noûs 37 (2), 181 List, C. and P. Menzies 2009, “Non-reductive physicalism and the limits of the exclusion principle”, Journal of Philosophy 106 (9), 475 Mackonis, A. 2013, “Inference to the Best Explanation, Coherence and Other Explanatory Virtues”, Synthese 190 (6), 975 Weslake, B. 2010, “Explanatory Depth”, Philosophy of Science 77 (2), 273 Woodward, J. 2003, Making Things Happen, New York: Oxford University Press —. 2010, “Causation in Biology: Stability, Specificity, and the Choice of Levels of Explanation”, Biology and Philosophy 25 (3), 287 Woodward, J. and C. Hitchcock 2003, “Explanatory Generalizations, Part 1: A Counterfactual Account”, Noûs 37 (1), 1 Yablo, S. 1992, “Mental Causation”, Philosophical Review 101 (2), 245

Notes 1 Both the interventionist account of explanatory depth and the interventionist account of causal explanation are formulated in terms of variables, and I will follow interventionists here in talking of causal explanatory relations as holding between variables (rather than between the features of the world these variables represent). Variables will be denoted by capital letters (X, Y, Z, etc.) while particular values taken by variables will be denoted by lower-case letters (x, y, z, etc.) 2 Woodward uses “support” where I use “entail” and deliberately chooses to leave it vague what he means by “support” (see e.g. Woodward 2003, p. 279). As will become clear below, however, the conditions a generalization G must satisfy in order to support a w-answer W are such that G supports W if and only if it entails it (see below note 4). I will thus stick to the more precise “entail” in what follows. 3 Though G is an equation, it is also a “change-relating generalization” in Woodward and Hitchcock’s terms (see e.g. Hitchcock and Woodward 2003, p. 181) and so has the right format to potentially be causally explanatory. 4 The invariance of G under an intervention setting X = 0 guarantees that it entails (WA1) simply because the truth of (WA1) is a necessary condition for G to be invariant under such an intervention. If (WA1) was false and it was not the case that Y would take value 0 were the value of X to be set to 0 by an intervention, then G would not remain true under an intervention setting X = 0. Invariant

Explanatory Depth, Inference to the Best Explanation and Proportionality

31

generalizations thus trivially entail the w-answers which they “support”, to use Woodward’s term. 5 Because depth, as characterized by the WH account, is a property of causal explanations, to appeal to it in a defense of IBE is to restrict the scope of IBE to causal hypotheses, i.e. to hypotheses that posit the existence of certain causal relations. 6 Note that the fact that w2 = f(z2) = g(z2), for instance, does not imply that either G1 or G2 entail the following counterfactual: (2) “Had the value of Z been z2 instead of z1 as the result of an intervention, then the value of W would have been w2 instead of w1”. It might be that, as a matter of actual fact, w2 = f(z2) = g(z2) but that this equality would not hold in counterfactual circumstances in which Z comes to take value as z2 as the result of an intervention (which need not have actually been the case). 7 Note that I use “predict” in a temporally neutral sense here, so that it covers both prediction as commonly understood (about the future) and retrodiction (about the past). 8 The view that prediction and explanation, whether causal or not, are the same thing was famously endorsed by Carl Hempel and Peter Oppenheim, who claimed that the distinction between the two notion “is of a pragmatic character.” (Hempel and Oppenheim 1948, p. 138) This view has fallen out of favor for a variety of reasons. But note that, regardless of what one thinks of this view, there is an important difference between Hempel and Oppenheim on the one hand and Woodward and Hitchcock on the other. According to Hempel and Oppenheim’s Deductive Nomological (DN) account of explanation, explaining an event such as W = w1 is the same thing as predicting (or retrodicting, assuming that we normally seek to explain events that have already occurred) the occurrence of this very event. By contrast, for Woodward and Hitchcock, causally explaining W = w1 is the same thing as predicting the occurrence of some other event, e.g. of event W = w52. 9 For advocates of the view that proportionality is necessary for causation, e.g. List and Menzies, this second explanation is not even properly called “causal” insofar as, on their view, the target being crimson is not a cause of Sophie’s pecking, since it violates the proportionality requirement. 10 I wish to thank Craig Agule, Nancy Cartwright and Gil Hersch for their comments on earlier drafts.



CHAPTER THREE COMPLEX CAUSATION IN THE RACE DEBATE LUDOVICA LORUSSO

1. Introduction In biomedicine, self-identified races (SIRs) are used to make predictions about individuals’ race-specific genetic susceptibility to complex diseases; to be more precise, the property of being of a specific SIR is thought to be a projectible property for race-specific genetic properties assumed to be component causes of complex diseases in specific SIRs (Risch et al. 2002; Burchard et al. 2003; Kumar et al. 2010). The question I would like to ask here is: are we justified in using the property of being of a specific SIR to predict an individual susceptibility to complex diseases? My point is that in order to justify this use of SIRs we must clarify what are the factors likely to be the main causes of an individual susceptibility to complex diseases in human populations, and then investigate the differences in such factors being the main causes of the differences in the occurrence of complex diseases among racial groups. In this paper I analyse the causal explanation of differences in the occurrence of complex diseases among racial groups, by focusing on recent studies of complex disease causation in human populations. First (sec. 2), I introduce the use of SIRs in biomedicine as proxies for genetic properties correlated with ancestry and co-varying with other genetic properties causing a population-specific genetic susceptibility to complex diseases; second (sec. 3), I show that biological theory and evidence from human population genetics do not support this use of SIRs; third (sec. 4), I argue that the property of being of a specific SIR is projectible for environmental, socioeconomic, cultural, and psychological (ESCP) properties, for interactions between ESCP properties and the body, and for epigenetic changes due to interactions between ESCP properties and individual-based genetic properties. Since current biological evidence



Complex Causation in the Race Debate

33

supports a fundamental role of ESCP properties, interactions between these properties and the body, and epigenetic changes, in the causation of the development of complex diseases in human populations, then ESCP properties and their interactions with genes and the body must be seen as causes of a non-genetic susceptibility to complex diseases in the individuals. As a consequence, SIRs are projectible for this kind of nongenetic susceptibility. Finally, I claim that differences in ESCP properties and in epigenetic changes among racial categories are the main causes of the observed differences in the occurrence of complex diseases among them.

2. The use of race in biomedicine The most used racial classifications in biomedicine are called “selfidentified” because they are constructed on the basis of self-declarations about racial identity. A self-identified human classification, therefore, relies on a mix of different kinds of psychological properties dependent on a specific historical, social, and cultural context; in biomedicine SIRs are assumed to represent some genetic variation correlated with a specific ancestry. Often, since genetic studies based on races are mostly conducted in the U.S., what researchers report as self-identified racial categories are made up by means of a self-categorization in one of the racial categories established by the OMB1. SIRs, therefore, are certainly both a “social” and “cultural” construction, in the sense that they are determined by both social and cultural properties. Moreover, SIRs are a “psychological” construction, since for example the property of being of SIR Black is a property that depends on personal beliefs of an individual about her “black” identity, and recent studies have shown that such beliefs may change over time (Saperstein and Penner 2012). Anyway, the fact that SIRs represent a social, cultural, and psychological construction does not mean that they cannot be also “biological”; some of the properties that may be used in the process of self-identification are in fact both social and biological. Consider for instance the property of having skin colour black: it is used in the process of self-identification in a specific racial group, but at the same time is certainly a biological property. This point is often used by researchers in biomedicine as one of the main arguments to justify the use of socially determined race concepts as genetic concepts, by claiming that racial categories are in fact a mix of genetic and non-genetic properties:



34

Chapter Three Unlike a biologic category such as sex, racial and ethnic categories arose primarily through geographic, social, and cultural forces and, as such, are not stagnant, but potentially fluid. Even though these forces are not biologic in nature, racial or ethnic groups do differ from each other genetically, which has biologic implications” (Burchard et al. 2003, p. 1171).

Nevertheless, claiming that individuals of a specific SIR might share distinctive genetic properties does not mean that they actually do. When geneticists use a self-identified classification as a proxy for a genetic classification of human beings they actually use the property of being of a specific SIR as a proxy for having a specific genotype. This particular use, however, is justified only if we assume that a self-identified classification can reliably reflect genetic differences among human populations. There are three main problems with this assumption. First, self-identified classifications are ambiguous because they are contingent upon the specific culture of the place to which people belong; as a consequence of their ambiguity, self-identified classifications cannot be good proxies for genetic classifications (about the issue of the ambiguity of self-identified classifications see, e.g., Hunt and Megyesi 2008). Second, biological traits relevant to the process of selfidentification such as skin colour are adaptive traits reflecting a specific genetic variation under selection; therefore, these traits are representative of a specific environment but not necessarily of a specific ancestry (Templeton 2013). Third, because genetic traits are not generally correlated in their variation over the planet, there is not a unique and unambiguous genetic classification of human beings; as a consequence, a classification based on the variation of some genetic properties cannot be used to predict other genetic properties of the genome (see section 3). In other words, from the colour of the skin of an individual we are not able to infer any genetic property correlated to her ancestry, and from an individual genetic property—correlated or not with ancestry—we are not generally able to infer other genetic properties of the genome. In spite of these important considerations, SIRs are used in biomedicine as proxies for specific genetic properties correlated with ancestry (or “genetic ancestry”), as well as for specific genetic properties causing a genetic susceptibility to complex diseases and co-varying with genetic ancestry. In particular, geneticists who use race as a proxy for genetic causes of diseases explain the evidence of the differences in the occurrence of particular diseases among racial groups in the U.S. by means of the existence of differences among SIRs in unknown genetic properties co-



Complex Causation in the Race Debate

35

varying with a specific ancestry and involved in causing a genetic susceptibility in individuals of a specific SIR. I call this explanation, the “genetic explanation”. Consider the case of the existence of differences in a complex phenotypic character like lung function between two SIRs in the U.S.: African Americans and European Americans. Under the genetic explanation, these differences are mainly due to differences in populationspecific genetic variants co-varying with genetic ancestry (Aldrich et al., 2012). Note that differences in genetic ancestry are not considered to be by themselves the causes of the differences, because they mostly occur in non-coding sequences. Thus unknown genetic variants are introduced ad hoc in order to explain the differences in pulmonary function between the two SIRs: Genetic ancestry itself is unlikely the cause of these differences, but may be a proxy for population-specific rare genetic variants contributing to racial/ethnic differences in lung function and susceptibility to tobacco smoke (Aldrich et al. 2012, p. 4).

In the genetic explanation, therefore, the property of being of a specific SIR is considered to be projectible for a specific genetic ancestry that is supposed to be a proxy for unknown genetic variants, which are taken as component causes of the disease. The hypothesis of the existence of unknown genetic properties covarying with genetic ancestry relies on the assumption that genetic properties are likely to be correlated in their variation across human populations—which, as I have said, is not the case. In the following section I will show in more details that biological data and theory do not support this assumption. In section 4, finally, I will show that differences in non-genetic variants as opposed to differences in genetic variants among SIRs in the U.S. are likely to be the main causes of the differences in the occurrence of complex diseases among them.

3. Genetic boundaries among human populations We have just seen that the use of SIRs in biomedicine is based on the assumption that SIRs may be projectible for a genetic variation co-varying with a specific genetic ancestry, and this assumption relies on the possibility that an unambiguous genetic classification of human beings exist. Most of human population geneticists agree on the fact that such a



36

Chapter Three

classification does not exist, even if it is possible to represent some kinds of correlations between geographical origin and genetic variation through clustering techniques (e.g., Rosenberg et al. 2002). Biological data show that only few genetic differences can be found among human populations—most of differences being among individuals (e.g., Lewontin 1972; Templeton 2013). Moreover, most of the few genetic differences that may be found among human populations are continuous, in the sense that they are present in all human populations with different frequencies (Barbujani and Belle 2006). There are in fact two possible kinds of genetic differences among human populations: continuous and discontinuous differences: x We have discontinuity in a specific genetic trait if, given two populations, that trait in an individual’s genome is found almost exclusively in one of the two populations. Therefore, from this trait it is possible to unambiguously predict the population to which the individual belongs. x We have continuity in a specific genetic trait if, given two populations, that trait in an individual’s genome is found with a different probability in the two populations. Therefore, from this trait it is possible to predict with a certain probability the population to which the individual belongs. In the case of discontinuity in one or more traits, from a specific genotype of a single individual it is possible to unambiguously determine the population to which the individual belongs. However, if traits were not correlated in their variation, populations constructed on different traits would rarely coincide. Only in the case in which discontinuities in the traits were correlated, populations constructed on different traits would tend to coincide. In this case there would be projectibility from one trait to another and, by attributing an individual’s genotype to a genetic cluster, one would obtain information on the whole individual’s genome. In case of continuity, where the distribution of one or more traits is probabilistic, the same reasoning applies and only in case of correlation between traits it is possible to have projectibility from one trait to another. Rare cases of discontinuities in single traits are possible, but only in extreme situations of populations that have been isolated for a long time (e.g., Barbujani and Colonna 2010). Studies on the humankind’s genome variation show that populations are characterized by a great continuity in the variation of most traits, and that these traits are not generally correlated in their variation, making any genetic classification arbitrary because



Complex Causation in the Race Debate

37

dependent on the trait or traits chosen; therefore, the biological significance of each classification strictly depends on the trait or traits it is based on. For this reason, even if genetic clusters can be created from statistically significant boundaries among genotypes, they are not unambiguous and they are not projectible for the whole genome variation. Attributing an individual to a population will not tell us anything about her whole genome: With the possible exception of some very isolated island or remote indigenous areas, most human populations exchange mates (often by rule) with other populations, at least their neighbors, and this obscures population boundaries and membership. Population boundaries, mating choices, and movements of people are also highly stochastic. All of these factors make human population strata and boundaries multilayered, porous, ephemeral, and difficult to identify (Weiss and Long 2009, p. 704).

Now we may ask: what is exactly the “genetic ancestry” that SIRs are assumed to be projectible for? Such “genetic ancestry” is supposed to be a genetic variation correlated with specific continents and representative of the genetic variation of ancestral populations of those continents (Rosenberg et al. 2002; Corander et al. 2004; Li et al. 2008); however, even if the variation of some genetic traits can be found to be correlated with geography—i.e. being the traits present with a different frequency in different continents—the absence of correlation among different traits does not allow to infer information about the whole genome of individuals only by the presence in their genome of certain traits considered to be informative of a specific ancestry. In addition to the problem of lack of correlation, it must be pointed out that the concept of genetic ancestry is ambiguous, in the sense that there are different possible levels of genetic ancestry depending of the time considered for testing a specific genetic ancestry; two populations, therefore, may be at the same time similar—if we consider one level of genetic ancestry—and different—if we consider another level. Recently, many geneticists have also questioned the biological significance of the tests used to infer genetic ancestry (e.g., Bolnick et al. 2007; Lee et al. 2009); in fact, the genetic panels used to test genetic ancestry are based on the genetic variation of specific proxy populations sampled from specific geographic regions, so that this variation actually reflects the specific genetic variation of the individuals having an ancestry in a certain region at a certain time rather than the specific ancestral genetic variation of a certain continent.



38

Chapter Three

It is certainly possible that a person identifying herself as an African American does share more genetic variation with a given group of African individuals with respect to a person identifying herself as an European American; but it is not possible to establish if this is really the case, and how much/which variation the first individual may actually share with the group of African individuals, since her ancestors might have been living in a different region of Africa—so that she might possess really different genetic properties from those considered in the genetic panels as representative of an “African ancestry”: Most African Americans derive from different areas of Africa that existed over a span of several centuries and that might not be well represented by those present-day proxy populations (Royal et al., 2010, p. 667).

I have just offered a rough overview of the problems of attributing specific genetic properties to SIRs. Finally, the hypothesis that SIRs may be characterized by specific genetic properties causing a population-specific genetic susceptibility to complex diseases and co-varying with a specific genetic ancestry is introduced ad hoc in the genetic explanation, since there is no evidence of the existence of such genetic properties, and in general the biological evidence does not support the hypothesis of co-variation of genetic traits across populations. If SIRs are unlikely to be projectible for specific genetic properties causally relevant for a population-specific susceptibility to complex diseases, the genetic explanation is undermined. In the following section I am going to investigate the possible causal roles of both genetic and non-genetic factors in the causation of complex diseases in human populations; I will show that current biological evidence supports the causal relevance of non-genetic over genetic factors in this kind of causation, giving therefore the kiss of death to the genetic explanation.

4. Complex causation and race When we consider a causal explanation of a difference in the occurrence of a disease between two populations, first we must face the issue of the causes of the development of that specific disease in the individuals within the two populations, and analyse the causes of the variability in the occurrence of the disease within them; for instance, if we find out that a variability in non-genetic factors is likely to be the most relevant cause of the variability in the occurrence of that disease in the two



Compllex Causation in the Race Debbate

39

populations,, it is importannt to investigaate the causall relevance of that kind of variabilitty between the t two popu ulations in thhe explanatio on of the difference inn frequency of o that diseasee between them m. Of course a portion of the wholee variability inn the occurren nce of that diisease between n the two populations may still be explained e by means m of a gennetic variabiliity. In our specific case we have seeen that such a genetic varriability—if itt exists at all—is still uunknown. In the caausation of coomplex diseasees like lung fu function declin ne, hearth diseases, annd cancer—the kinds of disease for whiich a differen nce in the frequency aamong SIRs in U.S. is observed—noon-genetic factors are known to be the most reelevant causess together witth an individu ual-based genetic preddisposition; moreover, m in this kind off causation, not only multiple cauusal factors may m intervenee in the causaal chain leadin ng to the developmennt of the diseaase, but they may have a ddifferent caussal power and interacct differently with each other througgh different temporal dynamics (V Vineis and Krriebel 2006). Therefore, diifferences amo ong SIRs should be seearched not onnly in differen nces in the cauusal factors, but b also in differences in the caussal powers they t may haave, in the kinds of interactions they may makke with each other, and in tthe temporal dynamics d of these inteeractions. One of tthe classic reepresentations of a multipl e kind of cau usation is Rothman’s ““pie” (Rothm man 1976), in which w multiplle factors are involved in different sufficient cauusal complexees (the “pies”)) of a complex x disease. magines the hyypothetical sittuation in whiich a disease has three Rothman im sufficient caausal complexxes, each havin ng a different number of co omponent causes (see figure 1). Inn this schemee, if a compoonent cause like l A is present in eeach of the coomplexes, it is to be conssidered as a necessary n cause of thee disease. On the t other hand d, the componnent causes th hat do not appear in evvery causal com mplex are nott necessary cauuses of that diisease.

Figure 1. Rotthman’s “conceeptual scheme for f the causes oof a hypotheticaal disease” (Rothman 19976, p. 589). Each E causal co omplex or “piee” represents a sufficient cause for the development of the disease.



40

Chapter Three

When we consider the explanation of the differences in the occurrence of a complex disease between two SIRs, a difference in the supposed relevant pies must exist between the two SIRs in order to explain the different occurrence of that disease between them. Such a difference may be either in the presence/absence of one or more pies, in the frequency of one or more pies, in the causal powers of the component causes within one or more pies, in the different interactions among the component causes, or in the temporal dynamics of such interactions. In the first two cases, some component causes or combinations of component causes may respectively be either present in one population but not in the other, or more frequent in one population than in the other, thus making specific pies to be respectively either present in one population but not in the other, or more frequent in one population than in the other. A component cause that may be present only in one of the two populations in a specific social/historical context is, for example, the psychological property of being discriminated (BD); other non-genetic properties may instead differ in frequency between the two populations, like eating healthy food, doing physical activity, etc. In the third case, the component causes may be more or less powerful in specific social/cultural contexts in the two populations: for example, some pathological conditions like chronic inflammation may be component causes of more severe pathological conditions like cancer, and can be found with the same frequency but different intensity in the two populations; in this case we may say that same component causes have different causal powers between the two populations. In the fourth case, the component causes may interact with each other and may get their causal powers enhanced/reduced as an effect of this interaction; for example, chronic inflammation and junk food are two separate component causes that may interact making chronic inflammation much more intense, and therefore enhancing its causal powers within the pie/s. Moreover, new component causes may emerge as a consequence of interactions: gene-environment interactions, in particular, may have a very serious role in the causation of a complex disease, as they may cause epigenetic changes that may turn out to be new component causes either for that disease or for other diseases, and may be carried across generations (see below). In the fifth case, different kinds of component causes may be present at different times during the life of individuals of different SIRs, thus interacting with each other at different times; for example, while in Blacks living in U.S. stress is likely to be present at very early stages of their life, caused for instance by growing up in stressful neighbourhoods, in Whites



Complex Causation in the Race Debate

41

living in U.S. stress is likely to be present later in their life: such differences in the time of occurrence of certain non-genetic component causes in the individuals’ life obviously change the temporal dynamics of their possible interactions with other genetic and non-genetic component causes, impacting the possibility of a causal complex to be sufficient for the development of the disease. Temporal dynamics are of particular relevance in gene-environment interactions. Consider for example the case in which, given two SIRs A and B, some component causes are present only in SIR B, like for instance the psychological property BD; this property is of course context-dependent, in the sense that it characterizes SIR B and not SIR A only under a specific historical and social context. In that specific context BD is SIR-specific and present in SIR-specific pies associated only to SIR B; thus in this context the property of being of SIR B is projectible for BD, as well as for any susceptibility to pathological conditions to which BD is likely to be a component cause. SIR-specific properties, however, may be weakened in their causal powers, be reduced in number, or even disappear (in this case making the pies in which they are present to disappear) in a different historical/social context in which, respectively, people of SIR B are less discriminated, fewer people of SIR B are discriminated, or no one of SIR B is discriminated at all. In other different contexts BD may characterize both SIR A and SIR B, either with a different frequency or with a different intensity and therefore different causal powers within the pie/s. With all these possibilities in mind, let us now consider how we can proceed in explaining the differences in the occurrence of a complex disease among different SIRs in the U.S.. Three are the main possible causal explanations we must distinguish here: the genetic explanation; the ESCP explanation; the epigenetic explanation. While the three explanations are not mutually exclusive, they consider differences in different kinds of causes as the most relevant causes of the differences in the occurrence of complex diseases among SIRs. While in the first explanation being of a specific SIR is a projectible property for genetic properties causing a specific genetic susceptibility to a complex disease, in the ESCP and in the epigenetic explanations being of a specific SIR is a projectible property, respectively, for ESCP properties and epigenetic properties, causing a specific non-genetic susceptibility to a complex disease. The epigenetic properties consist in being characterized by specific DNA alterations that have been caused by gene-environment interactions, where these alterations are modifications of DNA structure (like, for example, DNA methylation) that modify gene expression.



42

Chapter Three

We have seen that the genetic explanation involves a strong ad hoc assumption about the existence of genetic differences among SIRs covarying with specific genetic ancestries and causing specific predispositions to the disease; in the pie scheme, one or more genetic component causes within some of the sufficient causal complexes involved in causing the development of the disease must either be present in some SIRs and absent in others, or differ in frequency among SIRs. In this explanation the causal powers of genetic properties may be seen as the “penetrance”2 of those genetic properties in the SIRs. The genetic explanation assumes that being of a specific SIR is a projectible property for genetic properties being either SIR-specific or mainly present in a specific SIR, and causing a specific susceptibility to the disease; individuals being of SIRs who are characterized by these genetic properties will be considered as more susceptible to the disease in respect of individuals belonging to SIRs who are not characterized by those properties. In the hypothetical case the disease only occur in some SIRs, such genetic properties will be both SIR-specific and necessary component causes of the development of that specific disease. In the case of the ESCP explanation, being of a specific SIR is supposed to be a projectible property for one or more ESCP properties, which are component causes in one or more causal complexes. What an ESCP property can be? It can be eating saturated fats; living in a polluted environment; being under a psychological stress; having low education; having low income; having low access to health care; etc. In fact, most of these properties are highly correlated with the mix of properties that make an individual identifying herself as being of a specific SIR. The same cannot be said about the genetic properties correlated with a specific ancestry, for which the property of being of a specific SIR is not a projectible property, as we have seen above. However, if we want to consider differences in ESCP properties among SIRs as the most relevant causes of the differences in the occurrence of complex diseases among SIRs, we have to consider whether these properties are likely to be causally involved in the development of complex diseases (see below). In the case of the epigenetic explanation, finally, being of a specific SIR is supposed to be projectible for specific epigenetic changes due to interactions between ESCP properties and the genome. On this explanation we need a more dynamic view of the pie scheme, where different kinds of genetic and non-genetic properties interact with specific temporal dynamics. The presence/absence of each component cause is not the only relevant element in the causation; two other elements play an essential role in the causation: the interactions between genetic (individual-based) and



Complex Causation in the Race Debate

43

non-genetic component causes, and the temporal dynamics of such interactions. Moreover, this kind of explanation also considers inherited epigenetic changes as component causes. The crucial point is: is there any biological evidence for SIRs being projectible for ESCP properties and/or epigenetic properties that are in turn causes of a non-genetic kind of susceptibility to complex diseases? Epidemiological studies that investigate the differences in genetic traits involved in causing the differences in the occurrence of complex diseases among human populations already take into account some social properties like, for instance, the socioeconomic status. In particular, it has been observed that differences in the occurrence of certain complex diseases among SIRs still remain after adjusting for some socioeconomic properties like having a specific income and having a specific education; this evidence is often taken to support the genetic explanation (Thakur et al. 2014). However, properties reflecting the socioeconomic status do not represent the complexity of a racial identity, which relies on several psychological, social, and cultural properties. The concept of racial identity is a complex and fluid concept that needs to be studied and embedded in the epidemiological hypotheses about societal dynamics of health and disease. Since this concept strongly depends on the societal structure and the local culture of the place in which people live, it cannot be trivially captured by general indexes constructed to represent socioeconomic factors like income and education. Moreover, specific kinds of psychological stress characterizing specific SIRs in a specific social and historical context may even be the main relevant causes of the development of some particular pathological conditions in those SIRs. A psychological property like BD, for example, is certainly not detectable by means of a simple test of socioeconomic status, but it may deeply characterize the individuals belonging to a specific SIR; as it has been said, race can get “under the skin” (Smedley et. al. 2008), not in a genetic sense, but in a social and physiological sense. Therefore, the fact that health differences remain among racial groups—even when a statistical analysis has been adjusted for the socioeconomic status—does not necessarily support the genetic explanation. Nowadays, recent data coming from genome-wide association studies3 seem to suggest that gene-environment interactions and epigenetic changes, in addition to an individual-based genome variation are very likely to be relevant causes of an individual susceptibility to complex diseases (Manolio et al. 2009). Moreover, some studies (Vineis and Kriebel 2006) have pointed out that temporal dynamics of geneenvironment interactions are of great importance in biological mechanisms



44

Chapter Three

of complex causation. Recent studies on epigenetics have contributed to change the meaning of concepts like “heritability”; while so far heritability of a susceptibility to a disease was considered to be only “genetic”, nowadays heritability is proved to be also “epigenetic” (for recent evidence supporting a transgenerational inheritance of epigenetic changes, see e.g. Morgan and Whitelaw 2008; Slatkin 2009; Handel et al. 2010; Sullivan 2013). Moreover, a susceptibility to complex diseases can be acquired during life time, either by means of a direct action of ESCP properties on people’s body or by means of epigenetic changes occurring at different times—even very early—in people’s life (Zhu 2013). Evidence of an acquired susceptibility and of an epigenetic inheritance supports the causal role of differences in non-genetic properties—both ESCP and epigenetic properties—among SIRs in the causation of differences in the occurrence of complex diseases among them: in other words, such evidence supports both the ESCP and the epigenetic explanations. Therefore, while being of a specific SIR may be considered as a projectible property for gene-environment and body-environment interactions, and for either acquired or inherited epigenetic changes, biological evidence does not support its being projectible for populationspecific genetic properties. The current evidence suggests instead that genetic properties causally involved in the causation of complex diseases mainly differ among individuals rather than populations (Manolio et al. 2009). Turning back to the pie scheme, for each individual there will be many different pies for the same disease, constituted by both SIR-specific nongenetic properties like BD and individual-based genetic properties. Pies that contain SIR-specific properties should be considered SIR-specific pies. Concerning the explanation of the differences in the frequency of complex diseases among SIRs, the biological evidence does not support the existence of causal pies characterized by population-specific genetic component causes, and supports instead the existence of causal pies characterized by population-specific non-genetic properties like BD, which may interact either with individuals’ body (causing for example a chronic inflammation) or with individuals’ genome variation by means of specific temporal dynamics. Therefore, the differences that are likely to explain the differences in the frequency of complex diseases among SIRs are differences in non-genetic factors like ESCP properties, differences in their causal powers, and differences in the interactions either among nongenetic factors, between non-genetic factors and the body, or between non-



Complex Causation in the Race Debate

45

genetic factors and an individual-based genome variation following specific temporal dynamics. In figure 2 we can see an illustration of the pies for a complex disease in one individual of SIR A (the three pies in the first row) and in one individual of SIR B (the four pies in the second row). In the pies “Gi” represents individual-based genetic properties causing an individual-based genetic susceptibility; “P” represents different kinds of psychological properties characterizing both individuals of SIR A and SIR B, even if maybe with different intensity (anyway, P is supposed to be non SIRspecific); “SE” represents socioeconomic properties that may be differently distributed in the two SIRs; “E” represents environmental properties that may be differently distributed between the two SIRs; and “C” represents the cultural properties that may be differently distributed between the two SIRs. As we can see, the individual of SIR B is characterized by an additional SIR-specific pie, in which there is the SIRspecific component cause BD.

Figure 2. The first three pies are in common between one individual of SIR A and one individual of SIR B, while the fourth is SIR B-specific because only individuals of SIR B are discriminated in the social/historical context where individuals of SIR A and SIR B live. BD is a SIR B-specific P property.

To conclude, the hypothesis I suggest here is that SIRs can be considered to be projectible for SIR-specific causal pies characterized by SIR-specific non-genetic component causes. SIRs may be used to represent a variation of specific non-genetic component causes of complex



46

Chapter Three

diseases, and a variation of interactions between component causes and temporal dynamics of these interactions.

5. Conclusion The current use of race in biomedicine is grounded on a genetic determinism that is not supported by the current biological evidence on the role of the environment and the epigenetics in the causation of complex diseases in human populations. So far SIRs have been considered to be relevant from a biological point of view in virtue of their being projectible for some kind of genetic variation existing among human populations; more precisely, the property of being of a specific SIR has been considered in biomedical research to be projectible for population-specific genetic properties causing a population-specific genetic susceptibility to complex diseases. Current biological evidence, however, supports the existence of an acquired susceptibility caused both by ESCP and epigenetic properties, and of an inherited epigenetic susceptibility to complex diseases; therefore, current evidence supports a strong role of ESCP properties involved in causing a susceptibility to complex diseases through interactions with individuals’ body and individual-based genome variation. This evidence should push biomedical research towards a fundamental change of perspective, in which SIRs become projectible for a non-genetic kind of variation, for body-environment and gene-environment interactions, and for epigenetic changes. Finally, geneticists who use the property of being of a specific SIR as a projectible property for populationspecific genetic properties must seriously reconsider the epistemic role of race in the investigation of the mechanisms of causation of complex diseases.4

References Aldrich, M.C., Kumar, R., Colangelo, L.A., et al. 2012, “Genetic Ancestry-Smoking Interactions and Lung Function in African Americans: A Cohort Study”, PLoS One 7, e39541 Barbujani, G. and E. Belle 2006, “Genomic Boundaries between Human Populations”, Human Heredity 61, 15 Barbujani, G. and V. Colonna 2010, “Human Genome Diversity: Frequently Asked Questions”, Trends in Genetics 26, 285 Bolnick, D.A., Fullwiley, D., Duster, T., et al. 2007, “The Science and Business of Genetic Ancestry Testing”, Science 318, 399



Complex Causation in the Race Debate

47

Burchard, E., Ziv E., Coyle N., et al. 2003, “The Importance of Race and Ethnic Background in Biomedical Research and Clinical Practice”, The New England Journal of Medicine 348, 1170 Corander, J., Waldmann, P., Marttinen, P., et al. 2004, “BAPS 2: Enhanced Possibilities for the Analysis of Genetic Population Structure”, Bioinformatics 20, 2363 Handel, A.E., G.C. Ebers and S.V. Ramagopalan 2010, “Epigenetics: Molecular Mechanisms and Implications for Disease”, Trends in Molecular Medicine 16, 7 Hirschhorn J.N. and M.J. Daly 2005, “Genome-wide Association Studies for Common Diseases and Complex Traits”, Nature Reviews Genetics 6, 95 Hunt, L.M. and M.S. Megyesi 2008, “The Ambiguous Meanings of the Racial/Ethnic Categories Routinely Used in Human Genetics Research”, Social Science & Medicine 66, 349 Kumar, R., Seibold, M.A., Aldrich, M.C., et al. 2010, “Genetic Ancestry and Lung Function Predictions”, New England Journal of Medicine 363, 321 Lee, S.S., Bolnick, D.A., Duster, T., Ossorio, P., and K. TallBear 2009, “Genetics. The Illusive Gold Standard in Genetic Ancestry Testing”, Science 325, 38 Lewontin, R.C. 1972, “The Apportionment of Human Diversity”, Evolutionary Biology 6, 381 Li, J.Z., Absher, D.M., Tang, H., et al. 2008, “Worldwide Human Relationships Inferred from Genome-Wide Patterns of Variation”, Science 319, 1100 Manolio, T.A., Collins, F.S., Cox, N.J., et al. 2009, “Finding the Missing Heritability of Complex Diseases”, Nature 461, 747 Morgan, D.K. and E. Whitelaw 2008, “The Case for Transgenerational Epigenetic Inheritance in Humans”, Mammalian Genome 19, 394 Risch, N., Burchard, E., Ziv, E., and H. Tang 2002, “Categorization of Humans in Biomedical Research: Genes, race and disease”, Genome Biology 3, 1 Royal, C.D., Novembre, J., Fullerton, S.M., Goldstein, D.B., Long, J.C., Bamshad, M.J., and A.G. Clark 2010, “Inferring Genetic Ancestry: Opportunities, Challenges, and Implications”, American Journal of Human Genetics 86, 661 Rosenberg, N.A., Pritchard, J.K., Weber, J.L., et al. 2002, “Genetic Structure of Human Populations”, Science 298, 2381 Rothman, K.J. 1976, “Causes”, American Journal of Epidemiology 104, 587



48

Chapter Three

Saperstein, A. and A.M. Penner 2012, “Racial Fluidity and Inequality in the United States”, American Journal of Sociology 118, 676 Slatkin, M. 2009, “Epigenetic Inheritance and the Missing Heritability Problem”, Genetics 182, 845 Smedley, B., Jeffries, M., Adelman, L., et al. 2008, “Race, Racial Inequality, and Health Inequalities: Separating Myth from Fact.” http://www.emfp.org/MainMenuCategory/Library/ResearchResourceL inks/RaceRacialInequalityandHealthInequitiespdf.aspx Sullivan, S. 2013, “Inheriting Racist Disparities in Health: Epigenetics and the Transgenerational Effects of White Racism”, Critical Philosophy of Race 1, 190 Templeton, A.R. 2013, “Biological Races in Humans”, Studies in History and Philosophy of Biological and Biomedical Sciences 44, 262 Thakur, N., Martin, M., Castellanos, E., et al. 2014, “Socioeconomic Status and Asthma Control in African American Youth in SAGE II”, Journal of Asthma 14, 1 Weiss, K.M. and J.C. Long 2009, “Non-Darwinian Estimation: My Ancestors, My Genes’ Ancestors”, Genome Research 19, 703 Vineis, P. and D. Kriebel 2006, “Causal Models in Epidemiology: Past Inheritance and Genetic Future”, Environmental Health: A Global access Science Source 5, 21 Zhu, J. 2013, “Epigenetics(-omics) Takes Center Stage”, Journal of Genetics and Genomics 40, 323

Notes  1

According to the U.S. Office of Management and Budget (OMB), there are five races in the U.S.: Black, White, Asian, American Indian, Pacific Islander. 2 In medical genetics, the penetrance of a specific genetic trait causally associated to a disease is the proportion of individuals with that specific trait who exhibit clinical symptoms of that disease. 3 Genome-wide association studies are genetic studies that investigate associations between common diseases (e.g., cancer) and genome variation in human populations. In a study of association, “a genetic variant is genotyped in a population for which phenotypic information is available (such as disease occurrence, or a range of different trait values). If a correlation is observed between genotype and phenotype, there is said to be an association between the variant and the disease or trait” (Hirschhorn and Daly 2005, p. 96). 4 This research was funded through a Regional grant n. CRP-18397- 2009, Regione Sardegna (Italy), and a P.O.R. SARDEGNA F.S.E. 2007-2013 fellowship.



CHAPTER FOUR CAUSAL SUFFICIENCY AND INTERFERENCES LORENZO AZZANO

1. What is an interference? To some extent, that of “interference” is a concept of our everyday life. Plans are continuously altered, hindered and obstructed by an unpredictable environment, for instance when the subway breaking down prevents me from getting to the university in time. But is there some philosophically more precise definition of interference? According to a very simple definition, an interference is whatever prevents an expected outcome from happening. We should think of the “expected outcome”, not just as what we believed was going to happen, but as what was supposed to happen by nature itself; or, in another words, interferences disrupt what we believe to be a necessitated outcome. In this paper I am not going to give an account of causation itself, or describe, if there is one, the productive aspect of reality (even if I will occasionally discuss such issues); the problem at hand is, rather, whether or not causes bring about their effects by necessity and the role of interferences in assessing these matters. First of all, we need to understand what causal necessity is, and what an interference is. To say that the effect necessarily follows the cause means that the cause was somewhat sufficient for it, so that whenever the cause happens, the effect cannot but happen as well. This can be represented through the use of the strict implication: (suff1) C causes E entails Ƒ(CĺE) In standard possible worlds semantics, truth-conditions of the right part of (suff1) are that all C-worlds are E-worlds. To show that C may not be sufficient for E, logically, means to find a C‫ר‬൓E-world. This is where

50

Chapter Four

interferences come into the picture. Interferences are disruptive factors in a causal process that seem naturally meant to provide counterexamples to (suff1) as ensuring sufficiency in causation. According to Mumford and Anjum (2011), they precisely show that there is no causal sufficiency, and so that (suff1) is false. C may actually be the cause of E, but it could have been the case that, after C, something happened that made the causal process to take another route. Hence, no necessity. That the cause somehow “forces” the effect to happen is a very strong intuition on causation, one that even one of causation’s worst enemies (allegedly), Bertrand Russell, seemed to have: In order to be sure of the expected effect, we must know that there is nothing in the environment to interfere with it. But this means that the supposed cause is not, by itself, adequate to insure the effect.1

The inescapable presence of interferences provided, according to Russell, “insuperable difficulties” in the understanding of the notion of cause; and this seems to mean that causation is taken to be a necessary relation. As I’m not doing Russell’s exegesis, I content myself to substitute “adequate” with “sufficient”. Now, let’s define “interference”. First of all, note that we cannot just define an interference, say, from cause C to effect E, as a lack of sufficiency of C for E, as there may be some C‫ר‬൓E-world even if actually the causal process from C to E is successful and non-interfered. Therefore, we cannot define (actual) interferences straightforwardly as cases of failure of sufficiency; interference is itself a case of causal interaction, so it’s more likely that we cannot define it without causal notions appearing in the definiens. But we can use the notion of causal sufficiency to state some feature of interferences: (int)ĺ effect E, from cause C, is interfered ĺ C is not sufficient for E Interferences mess with the sufficiency condition of causation. Logically, any counterexample to the sufficiency of C for E needs to entail that it’s possible that C‫ר‬൓E. Therefore a condition for interference is the following: D Interferences don’t alter directly the cause. Provided D , whether some usual phenomenon of interference falls in the scope of D depends on what we consider to be the process involved

Causal Sufficiency and Interferences

51

and its cause. But how does an interference work? I would like to introduce a twofold characterization of interference by pointing to some classical feature of causation. Let’s consider the following scenario; a match is struck and it lights. A case of successful causation. As properties or events causes are taken to be, it’s uncontroversial that the match lighted because it had a head with proper chemical features, and it was struck against a properly shaped surface, and so forth. But if we want the cause to be sufficient for the effect, way more has to be specified in order to force the lighting to happen. For instance, the match could be struck in a vacuum, and without oxygen no fire would be produced. Or, watchful aliens on very little flying saucers could have the mysterious desire to instantly spray water on the match when it is struck, thus preventing it from lighting. We thus need what are usually called background conditions. To some extent, the distinction between background conditions and full-fledged causes is chauvinistic. This is the case, for instance, of oxygen: oxygen is a causally operative factor in the production of fire, one that is ontologically on a par with the chemical features of the head of the match. However, because the lack of oxygen is a quite unorthodox situation, its presence is not typically considered as a relevant feature in a process on lighting. But some background conditions are different, as the case of the little aliens show. Some other time we need, instead of mentioning the presence of some factors, to exclude it. No alien, and obviously no spray water is required for a lighting process; so they are extra causal factors that alter the final outcome. If there has to be fire, the intervention of water, whatever the source, has to be excluded. For brevity I’ll call the first kind of background conditions positive background conditions, while the second, negative background conditions. Based on this distinction it’s easy to distinguish two kinds of interferences, and respectively: subtractive and additive interferences. Subtractive interferences: lack of a positive background condition. Some partial cause is taken out. Example: some amount of substance X is bound to cause cancer in the subject. A healthy diet potentially decreases the quantity of X in the body. A healthy diet is thus a subtractive interference to the insurgence of cancer, given that it prevents the body from having enough X to produce it. Additive interferences: lack of a negative background condition. Some other element is added. Example: a magnet progressively attracts a metal ball towards point A; at some point, however, a more powerful magnet is placed in a way such that the ball starts to move towards point B instead.

52

Chapter Four

Three remarks on my twofold distinction. 1) Additive interferences seem somewhat dependent, or even reducible, to subtractive ones. For instance, let’s consider when water is sprayed on the struck match before it lights. The match needs to be dry because only when it’s dry its head is supposed to light, and this, again, is a positive background condition. The presence of water is thus a subtractive interference. While I agree that many supposedly genuine additive interferences may be identified with subtractive ones, I don’t believe a general argument for the reduction of the former to the latter is forthcoming. We can’t a priori deny that sometimes, somehow, something goes bad because something else happens. Sometimes, the difference between additive and subtractive may be again a linguistic or epistemic difference. For instance, “dry” may be a positive background condition if we consider it an additional property of the individual or a negative one if we consider it as expressing a lack of water or humidity. I take this confusion to be relative to the concept “dry” rather than to the general concept of interference. If we are realist about properties and/or states, we cannot fail in the task of understanding some difference between adding and taking out. 2) The distinction between interferences and “standard” causal factors may appear to be pragmatic, as the difference between partial causes and background conditions. Answer: it is of no interest for our present concern: our task right now is not to define the category of interference independently from causal notions, but rather to find out whether or not a given set of causal factors is sufficient for its effect. This leads to the third remark. 3) I previously said that interferences do not alter the cause. However, both the additive and the subtractive interferences seem precisely to do that. Now this is a tricky point. The key assumption of my twofold understanding of “interference” is that a real interference cannot alter the course of a process by magic, but interacting with the it, and this interaction is considered to be, itself, causal interaction (as (int)ĺ shows). And of course we may wonder, as in remark (2), what do interfering factors have to differentiate them from “standard” causal factors. In a case of interferences, the original process gets hindered because the set of causally operative factors, at some point, has changed: as we will later see, interferences may seem to be therefore confined to processes with intermediate steps. Given that a causal process is also determined by its final effect, a further complication lies within this third remark on interferences contexts,

Causal Sufficiency and Interferences

53

and it goes as follows: it’s unclear what we are referring to when we claim that a process has been interfered with, insofar as no real process is present that can satisfy the requirement of being interfered with. If a fragile vase is struck with sufficient force but it is somehow protected by a guardian angel as in Martin (1994), then there is no process that leads to the breaking, precisely because of the presence of the interference. Merely a potential process is interfered: but then we obtain the somewhat paradoxical conclusion that all that can be authentically said to be interfered with, does not exist. If interferences are a case of interaction that can be subsumed under the category of causal interaction, there can be interferences only if, as we said, there are intermediate steps. If we consider transitive causation, so if a set of causes is responsible for what happens even after some time (perhaps there is a gradual decrease of causal dependence), when an interference may pop up, then there is a causal relation jeopardized. It’s difficult to spell out precisely what does it happen to interfered processes, and trying would mean to give a full account of what causation is. As I said, this is not the aim of this paper. Roughly speaking, it’s like if the same causal relation turns into something else (from the production of the non-interfered effect to the production of the interfered effect).2

2. Is interference a macrophenomenon? I suggested myself the possibility that some, very simple and brief causal processes may be unbreakable (I take “unbreakable” as a synonym of “such that it cannot be interfered with’”); given our understanding of interference as causal interaction, provided the D condition, only mediated causal relations seem to be subject to this phenomenon; has this more to do with how the transitivity condition is conceived in causal talk or with interferences themselves? So, are there unbreakable processes? On what assumptions? First of all, I would like to spell out a little more what does the D condition mean. As I said, technically all background conditions are about causal interactions of the same kind. Let’s say that a “positive background condition for E” names a causally operative factor for E, while a “negative background condition for E” stands for a lack of a causally operative factor for ൓E. When a process towards E starts, all positive background conditions are present; in other words, we have a complete cause of E taking place (one for which we may want to apply (suff1)). The exclusion of a “negative background condition for E” is an exclusion of a “positive

54

Chapter Four

background condition for ൓E” that, in the relevant context, generates a complete cause of ൓E (an interference that is not strong enough for altering the potential outcome from E to ൓E does not need to be excluded). Therefore, given that one cannot have at the same time a complete cause towards E and a complete cause towards ൓E, when a complete cause for E take place and a process towards E starts, both positive and negative background conditions for E are present. Therefore D0) If a complete cause C for effect E takes place, when C takes place both positive and negative background conditions are present. We previously argued for the existence of two types of interference: the additive type, where a positive background conditions for another effect is added, and the subtractive type, where a positive background condition for that same effect is taken away. The first type consists in a lack of a negative background condition for effect E, while the second one in a lack of positive background condition for E.3 We can take the requirement for a failure of sufficiency as stating that if we want to show how a cause is not sufficient for some effect, the complete cause needs to happen (we need to have a C‫ר‬൓E scenario where C is a per se sufficient cause of E). So, if an interference for E needs to show that its causes were not sufficient, the complete cause for E, C, needs to occur. Now, because of (D0), and because of the fact that both kinds of interferences are lacks of some background condition (additive: negative, subtractive: positive) at the time when C happens, no other interaction can be present, given that both positive and negative background conditions for E are present. So: D Interferences don’t alter directly the cause is true. And given that both additive and subtractive interferences are in fact situations of causal interaction with the original set of causes, we can add another requirement E Interferences don’t alter directly the effect Why E ? let’s consider a simplified trumping-like preempted cause C is stopped from being the cause of E, up cause C* takes place bringing forth E* without interaction between the C and the C* process. There

situation where because a backany cutting or is apparently a

Causal Sufficiency and Interferences

55

violation of E , given that there is no disruption in the C process, yet somehow E fails to happen. Yet this example, like all examples of trumping preemption,4 is quite problematic; that a process towards something may be altered without interaction is something that relies too much on hidden assumptions; in our simplified situation, there must be some law, or some sort of governing principle such that “if both C and C* happen, E* takes place instead of E”, where C and C* do not relate in any significant way. Why would either of them take place, if we have complete causes for both of them, and they are mutually incompatible? This is not an interference: this is a failure of Sufficient Reason. Therefore, was each interference to be an interaction consisting in adding or taking away partial causes (positive background conditions), the E condition immediately holds. Without this understanding, as we will see, no unbreakable process can be granted, as even the most simple ones could be vulnerable for trumping preemption. The central claim of the supporter of causal sufficiency may be that some causal processes are just too simple to be interfered with, for instance when it comes to partless objects (mereological atoms) acquiring or losing structureless properties (properties supposedly without a causal basis, such as spin, or charge). We can imagine the following example: particles A and B are coming close, and at some point they touch, or they come close enough for one of them, say, B, to acquire positive charge. Exactly because both the property and the individual acquiring it are so simple, this property acquisition does not happen progressively, or bit by bit. A and B come close, they touch, and then bam!, the effect immediately takes place. The easiest way to account for this kind of situation is to claim, on two-event causation, that causation is discrete. This would mean that even if many of the causal processes that we witness in our everyday life, if analyzed, would show that the simple relation between cause and effect actually hides a system of more complex interactions between those individuals and/or temporal segments of the events (a causal chain), there are causal processes made up by a single causal relation between two events (a simple link). So, even if the description “the rock hitting the window caused it to shatter” describes a causal chain, as the descriptions of the events involve very complex physical processes, “particle A touching B caused B to acquire positive charge” may be the description of a genuine causal link. It’s easy to see why simple links are not subject to interferences, given the D andE conditions for interferences.

Chapter Four

56

Interferences need to operate on intermediate steps, leaving untouched both the cause and the effect; but in a simple link there is no such thing. It is interesting to note that the very viability of eventual metaphysics seems to rest on the assumption that causation is discrete; if there were no simple links, chains would thus be made of more chains, and a regressum ad infinitum would be generated. The importance of any of those relata, given the inexistence of immediate relations, would more likely be trivialized.5 But what is more, the discreteness of causation involves the discreteness of time, a very robust claim philosophers are (or should be) very careful in arguing about. Let’s see why, working with a dense time model, and through a reductio ad absurdum. On a dense time model, for each ǻt0-t1 time interval delimited by two instants t0 and t1 there is an extra instant t0.5 between them; ǻt0-t1 is thus composed by ǻt0-t0.5 and ǻt0.5-t1. The first problem of eventual causation on this model is the existence of a time gap between the cause and the effect; if the cause-event and the effect-event are conceived as contiguous and not overlapping,6 there is no way to avoid the existence of a time interval between them, as there is no next instant for the effect to be placed into, after the cause. Still, a supporter of eventual causation may resort to some continuous properties of events themselves in order to solve this problem.7 If we map events over half-open intervals, we can define a cause as an event that lasts for each instant until, say, t1, and the effect as the event taking place from t1; the cause lasts for all the infinite instants before the effect, and there is no time gap between them. Yet I believe that the most pressing problem for eventual causation on a dense time model is not about time gaps, but about the fact that, even in the case of a simple link, cause and effects are indefinitely temporally extended. The existence of causal processes made by simple links on such a model generates the following paradox: (1) (2) (3) (4) (5)

a process is temporally extended if (1) holds, a process has temporal parts contiguous temporal parts of a causal process are causally related time is dense there are simple links (causation is discrete)

All together, they imply a contradiction. In fact, (1), (2) and (3) jointly require that each causal relation that is temporally extended for more than two instants (one for the cause and one for the effect), is in fact a chain.8

Causal Sufficiency and Interferences

57

But according to (4), no process only lasts two instants. So no causal relation is a simple link; but that’s precisely what is stated in (5). How can we get out of the tangle? I don’t see how to make sense of the denial of (1), even in fundamental contexts; how could temporally extended macroprocesses be constituted by microprocesses, if the latter are just “positions” in time? (2) represents a fairly acceptable truth that cannot be denied without further requirements.9 (5) is the claim the supporter of unbreakable processes wants to hold in conjunction with (4), a claim on which we previously agreed to keep neutral about. Thus it seems that (3) has to go. At first sight, it makes sense to consider (3) wrong, in our case. As extended in time “A touching B” and “B acquiring positive charge” may be, even if they have temporal parts, such parts do not display any intrinsic difference with one another, so there is no change throughout their entire extension. “B acquiring positive charge” is an event that does not seem a further process constituted by smaller events causally chained to one another. So a supporter of unbreakable processes could put a little amendment to an eventual metaphysics of causation limiting (3) to nonextreme cases like rocks shattering windows. But (3) would not hold in our limiting case of A causing B to acquire positive charge. Still, such a position would have nasty consequences, exactly because the temporal parts of events so simple would not display any qualitative difference with one another, nor with the event they constitute. For example, we have “A touching B” and “B acquiring negative charge”: C and E; C and E are both temporally extended, and because time is dense they are indefinitely temporally extended. But for now, let’s just stick with the claim that they have each three temporal parts, C1, C2 and C3 and E1, E2, and E3, in this order; none of them bear any causal relation with the next one, or anyone else of them. But because the effect E is just a very simple event such as “B having positive charge”, all of those parts of E would equally be describable as “B having positive charge”. E, E1, E2 and E3 would be qualitatively indiscernible states, just a little shorter. Let’s consider a case where A touches B and B successfully acquires charge; whether or not B retains this property after acquiring it is not something that pertains to the causal process that made it acquire it in the first place. So after E1, say, B loses the charge, and the complete event E does not occur. Therefore, because according to the rejection of (3), E is the rock-bottom level of causality, no successful process towards B acquiring positive charge never happened: but that is clearly false; and we cannot select E1 as the right effect to avoid the problem, because E1 is as temporally extended as E.

58

Chapter Four

More generally, the strength of eventual causation, that of being capable of analyzing apparently simple causal processes into more complex ones (chains), without (3) has completely vanished. It thus seems that eventual causation, by selecting changeless process segments known as “events” as the basic unit of causation, requires discrete slots of time as well.10 Or, the supporter of two-event causation has to argue against the paradox; for instance, assuming the consistency of some sort of action and causation at a temporal distance. Yet, I won’t say more on the topic.11 This is not the only way to achieve unbreakable processes; D andE together imply that interferences, as localized interactions, need some intermediate step in some intermediate moment between cause and effect to operate. But almost any theory of causation requires some sort of zerotemporal distance condition to hold because of the impossibility of action at a temporal distance.12 Another account of causation may take causes and effects to be simultaneous, yet not instantaneous, for instance when we consider a quantity x, constantly changing at some rate, which continuously causes another quantity y to change, for some amount of time, at the same rate. This shift would move our compass from causes and effects, considered as events, to segments of processes in a state of change themselves; from there, some kind of “process” account of causation is more likely to follow. And obviously if cause and effect are simultaneous, causality is not a transtemporal relation and, because of the D – condition, no genuine interference is possible. Take for example the sun slowly warming a rock from two degrees to fifteen degrees C°; this (undoubtedly) causal process does not display a finite number of links, or perhaps even no link at all. There is just a change in a quantity slowly and continuously causing another quantity to change. How to make sense of this kind of causation within the boundaries of two-event causation is something that I leave to its supporters. Finally, if some zero-temporal distance condition holds, together with the claim that interferences are interactions of some intermediate step of a process, there is no need of any other requirement for the inexistence of interferences: all that we need is to claim that causation is intransitive. Was that to be the case, no non-contiguous steps of a process would be causally related, and thus subject to disruptions. This could be said for both eventual and process causation; in process causation, this problem partially overlaps with the question as to how can this account countenance mediated temporally extended processes (the traditional chains) if “causation” is not a transtemporal relation; because if cause C,

Causal Sufficiency and Interferences

59

extended from t0 to t2 is simultaneous with the effect E, E cannot in turn be the cause of anything that extends beyond t2 itself. How then can any interference hope to get the job done? The intransitivity of causation is a metaphysically robust claim, and is not clear to me how this endorsement is not backed up by a circular argument: let’s consider for instance a rock thrown at a window, but, before hitting it, being caught by someone’s hand. One could claim that catching the rock at place c1 interfered with the rock shattering the window, but someone else may reply that the rock flying in place c1 was not the cause of the window shattering, given that it was just too distant, spatially and causally, from the effect. Why was it not a cause? Simply because we believed causes to bring about their effects by necessity.

3. Commitments So we have, at some level and on some assumptions, unbreakable processes. But in order to dispel Mumford & Anjum’s argument more has to be done. In fact, for the argument based on interferences to work against causal sufficiency in (suff1), all that we need is a single case. One counterexample is more than enough. So the implicit claims that the supporters of causal sufficiency have to put forward are the following: 1) Macrocausation is somehow reducible to microcausation. This point is fairly clear. I put my sugar cube in water, but the water slowly freezes before the dissolution is complete. The process has been interfered with, and no claim about discrete links, or causal simultaneity, can take that away. Unless, of course, we hold causation at the macrolevel to be apparent, for instance if we take properties to be causal relata and we believe macroproperties to be reducible to microproperties. So the unbreakability of causal processes and the truth of (suff1), given that some everyday example of interferences seems genuine, rely on the truth of some kind of microphysicism. 2) For similar reasons, if all causal processes are really unbreakable, we should explain how the illusion of interferability is generated in the first place. The question has to be solved also if we want to exclude interferences by claiming that causation is not a transitive relation. For a counterexample to (suff1) to hold, all that we need is a possible interference (for a strict implication to be false we only need to find one world where the antecedent but not the consequent holds), so even if we were to find some non-actually interfered type of process, it wouldn’t be enough to dismiss Mumford & Anjum’s argument.

60

Chapter Four

To deny the existence of all interferences as I defined them is a monstrous job; one needs to endorse extremely robust claims on the nature of causation (such as the endorsement of some model, or a choice in the structural properties of the relation such as transitivity) on the mathematical structure of space-time and on the macro/micro debate. The price is causal sufficiency, that is to say, necessity in causation, something that we intuitively believe to hold. But what is truly causal sufficiency? As an analysis of preempting-like interferences may quickly show, we have little to nothing to back up our intuitions on the matter. We don’t need to fall on the other extreme and claim that causes and effects are to be found in random patterns–we don’t need to deny the existence of an interesting modal value in causation; but we may be forced to abandon the claim that this value equates with necessity.13

References Chakravartty, A. 2005, “Causal realism: Events and processes”, Erkenntnis 63 (1), 7 Huemer, M. & Kovitz, B. 2003, “Causation as simultaneous and continuous”, Philosophical Quarterly 53 (213), 556 Martin, C.B. 1994, “Dispositions and conditionals”, Philosophical Quarterly 44 (174), 1 Mumford, S. and R. Anjum 2011, Getting Causes from Powers, Oxford: Oxford University Press Russell, B. 1912, “On the Notion of Cause”, Proceedings of the Aristotelian Society 7, 1 Schaffer, J. 2000, “Trumping preemption”, Journal of Philosophy 97 (4), 165

Notes 1 Russell (1912, p. 7). Interestingly enough, he presented the problem of interferences as providing counterexample to eventual causation on a dense time model, something that I will face later on. 2 Our structure of definition may be threatened by interferences modeled after preempting counterexamples to the counterfactual analysis of causation. There are, broadly speaking, two of those cases. First case: late preemption cases present subtractive interference (a late cutting) even if apparently they do not entail a failure of sufficiency. So either (suff1) is not an adequate definition of “causal sufficiency”, and needs either a temporal or a causal qualification, or (int)ĺ is wrong given that the presence of some interference does entail a failure of causal sufficiency. Secondly: trumping preemption, on the contrary, present clear-cut

Causal Sufficiency and Interferences

61

cases of failures of sufficiency even if they fail to show the presence of an interference (no cutting). So either interferences are more than just additive or subtractive interactions, or they are not necessary for a failure of causal sufficiency (whatever we think of (int)ĺ). As those cases do not exclude that there is at least some uncontroversial case against (suff1), such as the one in Mumford and Anjum (2011), we treat them as unnecessary complications and we won’t discuss them now. 3 This suggests a useful interdefinability between positive and negative background conditions: Positive background condition for E =df lack of a negative background condition for ¬E Negative background condition for E=df lack of a positive background condition for ¬E. There is slight asymmetry between positive and negative conditions in causation; for instance, that there is actually a positive background condition for E does not entail that a negative background condition for ¬E is not met, if the former is not enough to alter the process towards the ¬E outcome. So in both definitions with “positive background condition for x” what we really mean is “positive background condition for x that, in the relevant context, generates a complete cause of x”. 4 Schaffer (2000). 5 Actually whether or not the regressum ad infinitum is really that problematic, and whether events as causal relata would be trivialized by the presence of underpinning chains is a somewhat more controversial claim. See point (1) in the last section. 6 On eventual causation, instantaneous causation is usually forbidden insofar as it would not allow for time-extended causal chains. 7 Chakravartty (2005). 8 Plus a non-time gap assumption that sanctions that no “time slot” is empty. But, as I said, this is not about time gaps. Even if there was a time gap, the paradox would still be engendered given that the cause and the effect themselves are temporally extended. It would dissolved if both there are time gaps and cause and effects are not temporally extended. Causation would just be a “time leaping” relation from one position in time to another. I consider this position too farfetched to be considered. 9 There is at least some level of agreement in metaphysics of persistence between endurantists and perdurantists that processes and events are temporally extended by having distinct temporal parts in different moments. 10 See Huemer and Kovitz (2003). But there could even be other requirements: for example, eventual causation seems to take a stand in the so called “process philosophy” by claiming that change is not real, in the sense that any process of change can be reduced to a string of changeless events. 11 Another way to put the problem is the following: if time is dense, we have infinitely small process segments, that may display differences with one another all

62

Chapter Four

the way down (there is change all the way down); yet, if causation is discrete, there is not “causal” change all the way down. 12 Someone more malevolent may even wonder to what extent is the zero-temporal distance requirement theoretically independent from any claim of causal sufficiency. 13 I would like to thank Andrea Bottani, Florian Fischer, Carolina Ana Sartorio and all the participants of the 10th national SIFA conference for useful criticisms, suggestions and discussions. As they say, I’m the only responsible for all of the above.

CHAPTER FIVE ABSOLUTE COINCIDENCES AND SALMON'S INTERACTIVE FORK MODEL ALESSANDRA MELAS

Introduction In philosophical literature the word “chance”, as for the French “hasard”, is commonly used to indicate a number of different things, such as lack of causes, randomness (von Mises, Kolmogorov), and so on.1 Even though all of the definitions of chance are quite fascinating, the present study considers only a restricted meaning for the word “chance”, taking into consideration a causal conception of chance, that is chance intended as absolute coincidences. According to Jacques Monod, absolute coincidences are the result of intersections between different processes that belong to totally independent causal chains:2 Mais dans d’autres situations, la notion de hasard prend une signification essentielle et non plus simplement opérationnelle. C’est le cas, par exemple, de ce que l’on peut appeler les “coïncidences absolues”, c’est-àdire celles qui résultent de l’intersection de deux chaînes causales totalement indépendantes l’une de l’autre.3

I will call this type of coincidences “causal absolute coincidences”.4 The first part of the present work provides a precise definition of causal absolute coincidences; the second one presents the strong relation between causal absolute coincidences and Wesley Salmon's common cause model, that is interactive forks (Salmon 1984, pp. 168-174). As it shall be observed, causal absolute coincidences are events that can be divided into intersecting causal components, and they give origin to consequences. Equally, Salmon's interactive forks are characterized by two, or more, intersecting processes and two, or more, ensuing processes.

64

Chapter Five

1. Causal absolute coincidences: a definition As we have already seen, causal absolute coincidences are simply the effect of the intersection of independent causal processes. The independence between two processes, A and B, that belong to different causal chains, can be explained in the following terms: 1) A and B are independent if they are probabilistically independent, so that: P(A/™B) = P(A/B) = P(A)

(1)

P(B/™A) = P(B/A) = P(B)

(2)

and

2) The probabilistic independence between A and B is not due to a common cause in their past.5 To clarify this point let us consider Monod's example (Monod 1970, p. 128) as it is represented in Figure 1-1. Dr Dupont is going to visit a patient for the first time. In the meanwhile, Mr Dubois is fixing a roof in the same area. When Dr Dupont comes across Dubois' work site, Dubois' hammer falls inadvertently down and the trajectory of the hammer intersects the trajectory of Dr Dupont, who dies. In Figure 1-1: P(A/™B) = P(A/B) = P(A)

(1)

P(B/™A) = P(B/A) = P(B)

(2)

and

The fact that Dr Dupont goes to visit his patient is probabilistically independent of the fact that the hammer falls down, and the fact that the hammer falls down is probabilistically independent of the fact that Dr Dupont goes to visit his patient. The two dotted lines in Figure 1-1 represent the two independent causal histories of A and B.

Absolute Coincidences and Salmon's Interactive Fork Model

65

Figure 1-1: Monod’s example of absolute coincidences

As already observed, according to Monod, absolute coincidences are the result of intersections between different processes that belong to totally independent causal chains: “c’est-à-dire celles qui résultent de l’intersection de deux chaînes causales totalement indépendantes l’une de l’autre” (Monod 1970, p. 128). From this conception a total independence between the causal chains involved follows, and it means that: x There is not any direct (or indirect) causal link between the causal lines we are taking in consideration.

66

Chapter Five

x The causal lines involved do not share any common cause in their past, neither in the near past or in the far past.6 To sum up, causal absolute coincidences are events that can be divided into components totally independently produced by some causal factor and those components are joined together. Let us consider a different example. Suppose that I am watching a TV programme on Boris Pasternak. In the meanwhile, my best friend, not knowing what I am doing and not knowing anything about that TV programme, is reading Pasternak’s Doctor Zhivago. We would say that it is a coincidence that at the same time (but in different places) my friend and I are doing something that concerns Boris Pasternak. However, since there is no physical direct interaction between the two processes, someone may conclude from this example that something is a coincidence only in the eye of the beholder. This case is different from the one proposed by Monod. Hence, we need a more precise definition of what a causal absolute coincidence could be according to Monod’s view: We speak of a causal absolute coincidence whenever there is an intersection, which is also a physical interaction, of two or more totally independent causal processes, in the same space and at the same time.

This is exactly the case with Monod’s example. In that circumstance the two independent causal processes intersect, and physically interact, in the same space and at the same time. No one may conclude that the intersection is a coincidence only in the eye of the beholder.7 What can be noted here is that chance is, according to Monod’s conception, a fundamental notion, something which comes to exist independently to the mind of the observers, as made explicit in the following passage: Le contenu de la notion de hasard n’est pas simple et le mot même est employé dans des situations très différentes. Le mieux est d’en prendre quelques exemples. Ainsi on emploie ce mot à propos du jeu de dés, ou de la roulette, et on utilise le calcul des probabilités pour prévoir l’issue d’une partie. Mais ces jeux purement mécaniques, et macroscopiques, ne sont de “hasard” qu’en raison de l’impossibilité pratique de gouverner avec une précision suffisante le jet du dé ou celui de la boule. Il est évident qu’une mécanique de lancement de très haute précision est concevable, qui permettrait

Absolute Coincidences and Salmon's Interactive Fork Model

67

d’éliminer en grande partie d’incertitude du résultat. Disons qu’à la roulette, l’incertitude est purement opérationnelle, mais non essentielle. Il en est de même, comme on le verra aisément, pour la théorie de nombreux phénomènes où on emploie la notion de hasard et le calcul des probabilités pour des raisons purement méthodologiques. Mais dans d’autres situations, la notion de hasard prend une signification essentielle et non plus simplement opérationnelle [...].8

Moreover, in cases such as Monod’s the intersection gives origin to some consequence, as it is illustrated in Figure 1-2.

Figure 1-2: Consequences of coincidences

68

Chapter Five

2. The connection between Salmon's Interactive Fork Criterion and causal absolute coincidences Causal absolute coincidences are events that can be divided into components which intersect (and also interact) in certain spaces and at certain times. As seen above, the resulting intersection gives origin to consequences. This section aims to show the strong relation between causal absolute coincidences and Salmon's common cause model. More precisely, following a presentation of Salmon's model, it will argue that coincidences can be entirely described in terms of interactive forks.9 A brief overview of Reichenbach’s common cause model is necessary to begin with.

2.1 Reichenbach’s Conjunctive Fork Model: a brief overview Every time an improbable coincidence has occurred we seek a common cause, and that common cause is invoked to explain the improbable coincidence.10 That is the reason why Hans Reichenbach introduced the Principle of Common Cause in his The Direction of Time: The schema of this reasoning illustrates the rule that the improbable should be explained in terms of causes [...]. The logical schema which governs it may be called the principle of the common cause. It can be stated in the form: If an improbable coincidence has occurred, there must exist a common cause.11

Every time some events occur together more frequently than they would if they were probabilistically independent of each other, there is some common cause C which takes place prior to those events. Here a Salmon's example: Suppose [...] that several members of a travelling theatrical company who have spent a pleasant day in the country together become violently ill that evening. We infer that it was probably due to a common meal of which they all partook. When we find that their lunch included some poisonous mushrooms that they had gathered and cooked, we have the explanation.12

If C is the common cause and A and B are two effects, the causal link between those three events can be represented in the following way:

Absolute Coincidences and Salmon's Interactive Fork Model

69

Figure 2-1: Reichenbach's common cause

Reichenbach explicated the Principle of the Common Cause in terms of a probabilistic structure called “conjunctive fork”. The notion of conjunctive fork is defined according to the following conditions: P(A š B/C) = P(A/C) u P(B/C)

(3)

P(A š B/™C) = P(A/™C) u P(B/™C)

(4)

P(A/C) > P(A/™C)

(5)

P(B/C) > P(B/™C)

(6)

The first equation, better known as “screening-off”, provided P(A/C)  0, entails: P(B/C) = P(B/A š C)

(7)

This means that C screens off A from B. According to the first equation, the common cause C makes the two effects, A and B, probabilistically irrelevant to one another. Similarly, the second equation entails that the absence of C also screens off the two

70

Chapter Five

effects. The third and the forth relations assert that C is a positive cause of the two effects A and B. In fact, the probability of each effect is greater in the case C has occurred than in the case C has not occurred.

2.2 Salmon's Interactive Fork Model The common cause model presented by Wesley Salmon in his Scientific Explanation and the Causal Structure of the World (Salmon 1984, pp. 168-174) is different. As Salmon says: Consider a simple example. Two pool balls, the cue ball and the 8-ball, lie upon a pool table. A relative novice attempts a shot that is intended to put the 8-ball into one of the far corner pockets, but given the positions of the balls, if the 8-ball falls into one corner pocket, the cue ball is almost certain to go into the other far corner pocket, resulting in a ‘scratch’. Let A stand for the 8-ball dropping into the one corner pocket, let B stand for the cue ball dropping into the other corner pocket, and let C stand for the collision between the cue ball and the 8-ball that occurs when the player executes the shot. We may reasonably assume that the probability of the 8ball going into the pocket is also about 1/2 if the player tries the shot, and the probability of the cue ball going into the pocket is also about 1/2. It is immediately evident that A, B, and C do not constitute a conjunctive fork, for C does not screen off A and B from one another. Given that the shot is attempted, the probability that the cue ball will fall into the pocket (approximately 1/2) is not equal to the probability that the cue ball will go into the pocket, given that the shot has been attempted and that the 8-ball has dropped into the other far corner pocket (approximately 1).13

Salmon's interactive forks are considered as spatio-temporal intersections between processes. The space-time diagram of the interactive forks has the shape of an x, as illustrated in Figure 2-2. In Figure 2-2, P1 and P2 are two processes which intersect in C. E1 and E2 are the two emerging processes from the intersection. The intersection produces an interaction and the interaction, according to Reichenbach’s mark criterion (Reichenbach 1956, pp. 197-205), has the capacity to produce changes in the properties of the two separate processes that come from the intersection. In order to characterize Salmon's forks the following condition is needed: P(E1 š E2/C) > P(E1/C) u P(E2/C)

(8)

Absolute Coincidences and Salmon's Interactive Fork Model

71

Figure 2-2: Salmon’s interactive fork

The intersection in C makes the two effects probabilistically dependent on each other. The interactive forks are considered as spatio-temporal intersections that in general violate Reichenbach’s screening-off condition. However, the screening-off condition may represent a limit case for interactive forks. There seems to be a kind of screening-off that is valid for some macroscopic interactive forks and it can be represented by the following condition (Salmon 1984, p. 177): P(E1 š E2/C) = P(E1/C) u P(E2/C) = 1

(9)

which violates the relation (8). Salmon calls this sort of forks “perfect forks”. According to Salmon, we can say:

72

Chapter Five The main point to be made concerning perfect forks is that when the probabilities take on the limiting values, it is impossible to tell from the statistical relationships alone whether the fork should be considered interactive or conjunctive.14

Which is, then, the most important difference between conjunctive and interactive forks?15 The answer is as follows: It seems essential to have two processes going in and two processes coming out in order to explain the idea of mutual modification.16

Moreover, according to the conjunctive fork model, the common cause can produce one of its effects without producing the other one, and vice versa. According to the interactive fork model, the common cause (the intersection) cannot produce one of its effects without producing the other one, and vice versa. Furthermore, as a result of the physical interaction, in the case of interactive forks, the two emerging processes have something (a mark) of each other and this is not true in the case of the conjunctive forks.

2.3 The coincidences x-shape and interactive forks Interactive forks are characterized by two, or more, intersecting (and interacting) processes and two, or more, ensuing processes. Equally, absolute coincidences are events that can be divided into components which intersect (and interact) in certain spaces and at certain times, and the resulting intersection gives origin to consequences. It seems that causal absolute coincidences can be represented by an x-shape, like interactive forks. I will call this x-shape coincidences x-shape, and I will try to investigate its relation with x-shapes that characterize interactive forks. 2.3.1 The first part of the coincidences x-shape and interactive forks Salmon provides some instances in which the two intersecting processes, P1 and P2, are probabilistically independent of each other, as in our definition of causal absolute coincidences. The following example may be quoted: In everyday life, when we talk about cause-effect relations, we think typically (though not necessarily invariably) of situations in which one event (which we call the cause) is linked to another event (which we call

Absolute Coincidences and Salmon's Interactive Fork Model

73

the effect) by means of a causal process. Each of the two events in this relation is an interaction between two (or more) intersecting processes. We say for example, that the window was broken by boys playing baseball. In this situation there is a collision of a bat with the ball (an interactive fork), the motion of the ball through space (a causal process), and a collision of the ball with the window (an interactive fork).17

According to the example above, the ball travelling towards the window and the window are two intersecting processes; the collision between them represents the intersection. In this case, the two intersecting processes belong to independent causal chains, since the fact that the window is there is probabilistically independent of the fact that the ball travels towards the window, and vice versa. Salmon does not give any importance to the distinction between intersecting processes that belong to independent causal chains and intersecting processes that belong to non-independent causal chains. This is not a relevant point for a good definition of interactive forks. Interactive forks are not sufficient criteria to say that an interaction occurs by coincidence, since a fork can always be found in which the interacting processes do not belong to independent causal chains. 18 However, cases of intersecting processes that belong to independent causal chains are not ruled out by Salmon's account, and Salmon's model can still be adopted to describe such cases. The first part of the coincidences x-shape can be described with Salmon's model. 2.3.2 The second part of the coincidences x-shape and interactive forks Let us consider again Monod's example, as it is represented in Figure 1-2. In that case we may reasonably assume that the probability of Dr Dupont dying D is of 1/2 if there is a collision C between the hammer and Dr Dupont's head, and that the probability of the hammer having bits of brain E is also of about 1/2. In this case, it is evident that D and E, given C, are probabilistically dependent on each other: P(D š E/C) > P(D/C) u P(E/C)

(10)

Given that the hammer collides with Dr Dupont's head, the probability that the hammer will have bits of brain (approximately 1/2) is not equal to the probability that the hammer will have bits of brain, given that the hammer has collided with Dr Dupont's head and Dr Dupont has died [P(E/C š D) | 1]. That is:

Chapter Five

74

P(E/C š D) > P(E/C)

(11)

Moreover, given the collision between the hammer and Dr Dupont's head, the probability that Dr Dupont will die (approximately 1/2) is not equal to the probability that the doctor will die, given that the hammer has collided with Dr Dupont's head and the hammer has bits of brain [P(D/C š E) | 1]. That is: P(D/C š E) > P(D/C)

(12)

The intersection in C makes the two effects, D and E, probabilistically dependent, like in Salmon's example of the two billiard balls. However, as it has been seen already, the condition of screening-off represents a limit case for interactive forks. There appears to be a kind of screening-off which is valid for some macroscopic interactive forks. That limit case is represented by the following condition: P(E1 š E2/C) = P(E1/C) u P(E2/C) = 1

(9)

Let us consider the two billiard balls example once more. Suppose that our novice attempts shot from time to time. Since practice helps improve one’s skills to perfection, the novice becomes so good that he can invariably make the cue ball and the 8-ball collide (C) in the manner that the 8-ball drops into one of the far corner pockets (E1) and the cue ball goes into the other far corner pocket (E2). We may reasonably assume that the probability of the 8-ball going into the pocket is of about 1, if the player tries the shot, and the probability of the cue ball going into the pocket is also of about 1. It is immediately evident that C does screen E1 and E2 off from one another. Up until the moment when our player has perfected his technique, the results of his shots exemplified interactive forks, and it would be absurd to claim that when he achieves perfection, the collision of the two balls no longer constitutes a causal interaction, but it must now be considered as a conjunctive fork. It is an arithmetical accident that, when perfection occurs, the equation (9) is fulfilled while inequality (8) must be violated (Salmon 1984, pp. 177-178). Let us go back to Monod's example once again. We may reasonably assume that the probability of Dr Dupont dying (D) is of about 1, if there is a collision between the hammer and Dr Dupont's head (C), and that the probability of the hammer having bits of brain (E) is also of about 1. In this latter case, it is evident that C screens D and E off from one another:

Absolute Coincidences and Salmon's Interactive Fork Model

P(D š E/C) = P(D/C) u P(E/C) = 1

75

(13)

However, it would be absurd to claim that the collision between the hammer and Dr Dupont's head no longer constitutes a causal interaction, but must now be considered as a conjunctive fork. According to this last example, the common cause (C) cannot produce one of its effect without producing also the other one, and vice versa. Moreover, the two emerging processes (D and E) have something of each other. It is an arithmetical accident that, when perfection occurs, the equation (13) is fulfilled while the inequality (10) is violated. According to what I said in this section, even the last part of the coincidences x-shape can be described using Salmon's interactive fork model.

3. Some problematic cases In the previous sections I have shown how coincidences can be described through Salmon's common cause model. As noted, interactive forks are characterized by two, or more, intersecting (and interacting) processes and two, or more, ensuing processes. Equally, coincidences are events that can be divided into components which intersect (and interact) in a given space and at a given time, and the resulting intersection gives origin to consequences. Hence, it seems that causal absolute coincidences can be represented by an x-shape, as is the case with interactive forks. However, there are particular instances that deserve, at least, a brief mention. Consider, for example, a hydrogen atom, which absorbs a photon and then exists for a given time in an excited state (Salmon 1984, p. 181). We can consider the hydrogen atom and the photon as two intersecting processes that belong to independent causal chains, so that their intersection represents a coincidence. However, in this case, two different processes come together, and they seem to fuse in a single outgoing process. The shape of the space-time diagram is no longer an x but a O. Therefore, some coincidences can be represented with a O-shape (Salmon 1984, p. 181). Can Salmon's interactive forks be represented by a O-shape too? According to Salmon, “it seems essential to have two processes coming \out in order to exploit the idea of mutual modification” (Salmon 1984, p. 182). In other words, interactive forks can only be represented by an xshape. Some coincidences cannot be described by Salmon's common

76

Chapter Five

cause model; nevertheless, it is still open to discussion whether the Oshape phenomena can be reassessed as x-shape phenomena. In addition, in the case quoted above a situation in which a photon is completely absorbed by a hydrogen atom is extremely rare. By contrast, the situation in which a photon is not completely absorbed by the hydrogen atom and there are two intersecting processes and two other processes coming out, that is the hydrogen atom in an excited state and the photon with less energy, is more common. These kinds of phenomena also have an x-shape. It may be worth noticing how some cases in Biological and Social Sciences seem to be more problematic, but I will not explore the possibility of reassessing them as x-shaped phenomena in this context, and I will leave that discussion for a future study. To end this section, most coincidental phenomena show clearly at least two intersecting processes and at least two outgoing processes, and these phenomena can be described by way of Salmon's interactive fork model.

4. Conclusion As the present study has highlighted, according to a causal conception of chance, chance events are simply the result of intersecting causal lines. More precisely, the intersections between independent causal chains are the origin of accidental events that can be called “causal absolute coincidences”. In line with this view there is a relation between chance and the Principle of Causality, according to which whatever begins to exist has a cause. In addition, there seems to be even a strong relation between causal absolute coincidences and a precise causal model, that is Wesley Salmon's common cause model. To sum up, the primary reasons to assert that Salmon’s interactive forks can describe causal absolute coincidences are: 1. Interactive forks are characterized by two, or more, intersecting (and interacting) processes and two, or more, ensuing processes. Equally, causal absolute coincidences are events that can be divided into components which intersect (and interact) in a given space and at a given time, and the resulting intersection gives origin to consequences. Causal absolute coincidences can be represented by an x-shape, as with interactive forks. 2. Cases of intersecting processes that belong to independent causal chains are not ruled out by Salmon's account and Salmon's model can be

Absolute Coincidences and Salmon's Interactive Fork Model

77

used to describe such cases. The first part of the coincidences x-shape can be described by way of Salmon's criterion. 3. Salmon's interactive fork model can be easily used to describe the last part of the coincidences x-shape. Finally, not only there is a connection between causal absolute coincidences and the Principle of Causality, but also there is the possibility to describe them in terms of a precise causal model.

References Cournot, A.A. 1851, Essai sur les Fondements de nos Connaissances et sur les Caractères de la Critique Philosophique, Paris: Hachette Hacking, I. 1990, The Taming of Chance, Cambridge: Cambridge University Press Julienne Junkersfeld, M. 1945, The Aristotelian-Thomistic Concept of Chance, Notre Dame, IN: University of Notre Dame Monod, J. 1970, Le Hasard et La Nécessité. Essai sur la philosophie naturelle de la biologie moderne, Paris: Éditions du Seuil Owens, D. 1992, Causes and Coincidences, Cambridge: Cambridge University Press Reichenbach, H. 1956, The Direction of Time, Berkeley/Los Angeles: University of California Press Salmon, W.C. 1984, Scientific Explanation and the Causal Structure of the World, Princeton: Princeton University Press

Notes 1

For an extended enquiry see Hacking (1990). In Metaphysics, Aristotle maintains that the existence of per accidence causes is a sign of the existence of per se causes. In commenting on Aristotle's Metaphysics, Saint Thomas also says that if we treat accidental beings as things produced by per se causes, many things may be by accident, such us the meeting of independent causal lines. Although it is very important for the philosophical historiography, the Aristotelian-Thomistic notion of chance has not been taken into account in the present discussion. For an extended enquiry see Julienne Junkersfeld (1945). 3 Monod (1970, p. 128). A similar conception can be found in Antoine-Augustin Cournot: “Les événements amenés par la combinaison ou la rencontre d’autres événements qui appartiennent à des séries indépendantes les une des autres, sont ce qu’on nomme des événements fortuits, ou des résultats du hasard” (Cournot 1851, p. 52). 4 This is in order to make a distinction between this notion of coincidence and other notions of coincidence. 2

78

5

Chapter Five

I want to avoid the case in which an event X is a screening-off common cause of the two processes A and B. In that case, according to Reichenbach's screening-off condition, given X, A and B are independent of each other. For a more extended discussion on Reichenbach's screening-off condition see Salmon (1984, pp. 159163). 6 Clearly, a common cause of the causal lines involved could always be found if we trivially consider the range of all the physical laws and all the background conditions as common causes. In order to identify specific common causes, it is necessary to hold both the physical laws and the background conditions fixed, and to see whether extra common causes exist. Moreover, Monod’s view seems to exclude what David Owens calls “partial coincidences”. According to Owens, causal lines could be at least partially independent, even if they share a common cause in the past: “how much of a coincidence it is will depend on the weight and salience of the causes shared by its components relative to those of the causes which are not shared” (Owens 1992, p. 8). 7 One may say that coincidences are also unexpected events and therefore ask whether a non-unexpected intersection between processes that belong to independent causal chains is still to be considered as a coincidence. Though this is a very interesting point, it will be left out of the present discussion. 8 Monod (1970, pp. 127-128). 9 A similar work can be done using Bayesian networks, but that is left to further discussion and to another paper. 10 It is worth noticing that in the present context the word “coincidence” is not intended as synonymous with “absolute coincidence”. 11 Reichenbach (1956, p. 157). 12 Salmon (1984, p. 158). 13 Salmon (1984, pp. 168-169). 14 Salmon (1984, p. 178). 15 For a detailed discussion on conjunctive forks see Salmon (1984, pp. 158-168). 16 Salmon (1984, p. 182). 17 Salmon (1984, p. 178). 18 The reader shall be recalled that the aim of the present paper is to show that interactive forks are a necessary condition for coincidental phenomena.

CHAPTER SIX THE METHOD OF CONTRAST AND THE PERCEPTION OF CAUSALITY IN AUDITION ELVIRA DI BONA

Assuming that there is causality in nature, there are several questions related to our possibility of perceiving it. First of all, there is the crucial distinction between the mere perception of causality, namely our possibility of perceiving that something causes something else to happen (Michotte 1963; Duncker 1935), the fact that our experience represents causality as a part of perceptual content (Siegel 2009), and the idea that we can perceive causality only through a sort of strict analogy with other senses (Butterfill 2009)—namely we see some causal interactions in the same sense as that in which we can hear speech.1 With a corpus of about 150 experiments, Michotte (1963) tried to demonstrate that adults could perceive that something causes something else to happen. Michotte showed that subjects were inclined to describe scenes of launching and entraining in causal terms. In launching, an object a moves towards a stationary object b and, upon being touched, object b immediately begins to move in the same direction as a while a stops. In the entraining case, the same thing as in launching happens, but a continues to move along b. Michotte uses different kinds of objects in his experiments, such as hefty wooden balls, lights and shadows projected onto a screen, or a combination of both. Even if subjects knew that no causal relations were in force in a significant number of situations, they still described the situations shown in causal terms. Michotte’s results demonstrate that adults regularly describe launching and entraining in causal terms using sentences such as: “The red square moved the blue one along,” or “The ball pushed the shadow.” Therefore, Michotte’s experiments do not prove that we can genuinely perceive that something causes something else to happen, but that we can describe a specific

80

Chapter Six

situation in causal terms. To prove that we can perceive causality, we need theoretical assumptions about the relationship between experience, beliefs and reports. Moreover, Michotte isolated the exact parameters that have to be respected in order for subjects to describe an action as causal, such as the exact rate at which objects moved, the acceptable delay between the hit of object a and the moving of object b, or the dimensions of the objects on the screen. Duncker (1935) has also studied the correspondences between the properties of causes and of their effects, extrapolating some specific parameters that have to be satisfied in order for us to experience causality. For example, a sound has to be heard there where an object is seen to strike; a sheet of paper acquires a crease where it is folded; fire ensues shortly after a match is applied to an object. There are also pronounced similarities in content and form between cause and effect. The shape of a footprint corresponds to that of a shoe; a hot object transmits heat to its surroundings; a wet object moistens things in contact with it. In addition, Duncker correlates the accelerated rhythm of the motions of knocking parallels to the changing rhythm of the sounds produced. Both Duncker’s and Michotte’s reflections are useful for evaluating the correct conditions that have to be satisfied in order to experience causality, but that does not imply that we can represent or perceive causality. Then, another issue related to the theme of causality concerns whether there is a method that allows us to discover if we can perceive causality and if such a method is applicable to all sense modalities. In what follows, I shall consider Siegel’s method of phenomenological contrast as a good tool to discover whether a high-level property, such as causality, can be perceived. I shall state that the method works not only within the case of vision, but also in the specific case of audition.

1. The contrast method The method of contrast is a useful tool in the philosophy of perception since, by virtue of the examples upon which it is based, it may help us to discover something that is characteristic of perceptual experience. Moreover, in order to apply the method correctly, we need to specify the features of our intuition of difference upon which the method is grounded. As a starting point, it is necessary to establish what the relata connected by the relation of causality are. At this stage, I am not considering the casual theory of perception as sustained by Lewis (1980), Grice (1961), or Armstrong (1968). In their versions of the causal theory, they discuss the conditions in which, during the visual experience E of an object O, the object O would cause having the experience E. In fact, as in

The Method of Contrast and the Perception of Causality in Audition

81

every theory of perception, the causal theory also tries to explain the nature of our perceptual states and how these are connected to the world. For the causal view, the relata are, on the one hand, my own experience and, on the other hand, the object of the world that caused it. In addition, since we are within the sphere of auditory perception, what I am not considering as relata are sound and its source. I am not going to show whether sound sources cause their own sounds and whether we can perceive them as causally related, since I think they can be considered as a sole event, therefore we cannot perceive them as causally related. The relata of the causal relation I shall be taking into account are two different sonorous episodes. I shall sustain that the causality we represent is not that connecting sound and sound sources within the same sonorous episode, but the one linking two different sonorous episodes together. With “sonorous episode” I mean a situation in which you have a sound produced by a sound source, namely the jiggle of jingling keys or the music of Death and the Maiden performed by a string quartet. Hence, I shall hold that the causality we perceive is not that connecting sound and sound sources within the same sonorous episode, but one linking two different sonorous episodes together. The phenomenal contrast method is used to “discover” not only the contents of perceptual experiences, i.e. to find out, once a perceptual state has somehow acquired a content, which properties (or objects) enter such content. It is also used to discover if there is something within the perceptual experience that cannot be exhausted by the representational content, since it is irreducible—namely a form of qualia or the so-called “phenomenal character” of experience. The phenomenal contrast situation is composed of a “target experience” and a “contrasting experience,” constituting a minimal pair. The constraints on the “target experience” should be that: a) it does not differ from the “contrasting experience” in any other respect than the tested property; and b) it should clearly contain the target property. The method predicts that, if such a contrast were to be obtained, then “the experiences contrast phenomenally because one of them has the hypothesized content, while the other one does not” (Siegel 2009, p. 158). The method is based on the analysis of three different kinds of situations: 1) a scenario that changes with respect to a specific trait and that is perceived in two different ways by the same perceiver; 2) a scenario that remains the same and that has been perceived in a different way by two perceivers alike in all respects but for a difference; 3) a scenario that remains the same and that has been perceived in two different ways by the same perceiver. The majority of the applications of the method has been based on the first and on the second option. My aim here is to see which

82

Chapter Six

among the three options listed above would be the best instance for the correct application of the method.

2. Applications of the method An example of the first option is Koksvik’s “cinema screen argument”: Imagine that you are sitting in a comfortable seat in the middle of a dark movie theatre. You are not in pain, you are not hungry or thirsty, and you are sitting still. The screen turns a uniform, pleasant green. You stare at the screen, concentrating on it. This is the first situation. You relax, closing your eyes. When you open them again, the screen is a uniform garish red. You stare at it, concentrating on it. This is the second situation. Clearly, the characters of your overall phenomenal experiences would be different in the two situations. The best explanation for this is that perceiving something green makes a different contribution to experience than does perceiving something red (Koksvik, forthcoming, p. 6).

Even if Koksvik criticizes the use of the method, the cinema screen argument is meant to show that there is a difference in the way in which a perceiver could perceive a scenario that remains alike but for a specific trait, i.e. the colour of the screen. The idea is that, since the colour of the screen changes, the experience will consequently change. Nonetheless, the nature of such a difference is not obvious at all; it may change, depending on what your theory of perception is. An example of the second option is given by Kriegel (2007, p. 125): Suppose Prosop and Aesop are modal counterparts, living their almost indistinguishable lives in two different possible worlds. Everything about Prosop and Aesop is the same, and everything that ever happens to them is the same—with one exception: Prosop is, but Aesop is not, prosopagnostic. And now, on their 21st birthday, Prosop and Aesop are looking tenderly into their respective mothers’ eyes. Intuitively, it seems there is a difference in what it is like for them to undergo their respective perceptual experiences at this moment. There is an element that is phenomenologically manifest in Aesop’s experience but not in Prosop’s. This element is the feel of recognizing mommy’s face. Therefore, the property of being mommy’s face is phenomenologically manifest in Aesop’s perceptual experience.

Aesop and Prosop, with the latter being prosopagnostic, are going to perceive the same scenario differently, therefore the example is meant to show that a high-level property, such as mommy’s face, is a property that can be phenomenologically manifest. In this case, mommy’s face remains

The Method of Contrast and the Perception of Causality in Audition

83

constant, while the perceivers differ in a specific trait. Therefore, it seems obvious to claim that there is a difference between their experiences explicable, in Kriegel’s view, in terms of the representation of a high-level property. A further example of the second option is Sigel’s synchronic case of the perceptual experience of causation (2009, p. 158). The target situation is when you open the curtain of your room and allow some light to come in. Thus, you have the event of drawing the curtain and the event of the increasing the room’s illumination. The contrasting scenario is based on the idea that the curtain does not block out any light in the first place since it is translucent. Therefore, just as you uncover the window, the sun is let in as in the target case, but not simply by virtue of drawing the curtain, which is translucent, but because, as the curtain is drawn open, the sun comes out from behind a dark cloud. Yet, the uncovering of the window causes the room to light up progressively anyway, as the curtain is drawn open. Siegel’s example is meant to show that what is phenomenologically manifest is the property of causality, a high-level property that will be represented in the target experience—where the perceiver represents the two events as completely unified—and not in the contrasting experience, where, on the contrary, the perceiver would perceive them as distinct. There is an element in the scenario that makes it different, namely the thickness of the curtain, but Siegel sustains that this feature is not relevant to her thesis; hence the perceiver will perceive the two scenarios differently. I shall propose two examples (a and b) of the third option in which there is no change in the scenario, but which are, nonetheless, perceived differently by the same perceiver. Suppose you have never seen a pine tree before and are hired to cut down all the pine trees in a grove containing trees of many different sorts. Someone points out to you which trees are pine trees. Some weeks pass, and your disposition to distinguish the pine trees from the others improves. Eventually, you can spot the pine trees immediately: they become visually salient to you. Like the recognitional disposition you gain, the salience of the trees emerges gradually. Gaining this recognitional disposition is reflected in a phenomenological difference between the visual experiences had before and those had after the recognitional disposition was fully developed (Siegel 2011, p. 132).

In example a) the scenario remains stable, but the perceiver might perceive it differently. A person might find herself in two very similar situations, so that the capacity of immediately recognizing pine trees is

84

Chapter Six

exercised in one but not in the other. Here, the target experience is the over-all visual experience one has after learning to recognize pine trees (of which the specific visual experience of recognizing a pine tree is part), and the contrasting experience is the over-all experience one has before learning to recognize pine trees. What is phenomenologically manifest is the ability to recognize a pine tree, and this would explain the difference between the two experiences. Another example of the third option is Siegel’s diachronic case of causality perception (example b). Suppose you are playing catch indoors. Then, by mistake, the ball lands in a potted plant. You see the ball landing and, after that, the lights go out. The ball’s landing in the potted plant does not cause the lights to go out and, presumably, you do not believe it does, but it seems that the landing of the ball causes the lights to go out. In the second case, again, you see the ball landing in the potted plant and the lights go out, but we are assuming that in this case you have no feeling that the ball’s landing caused the lights to go out. Siegel says that the phenomenal difference between the two experiences stems from how we perceive the connection of the two events. In the second case, we just see two discrete events as occurring in a quick succession, while in the first one we perceive them as experientially unified in a way that is not merely temporal. Such an experiential unification that we perceive visually is an experiential representation of causation. A phenomenological contrast argument invokes visually experienced causality as the best way to explain the phenomenological difference. I listed three different ways of applying the phenomenal contrast method. I would say that only the third option is acceptable, since it is the only application in which intuition is the basis of the features necessary to avoid a massive use of introspection. The method starts from the idea that the comparison between two experiences might reveal something new about perceptual experience. The method is based on two passages: 1) the appeal to introspection, which is necessary in order to build the minimal pair argument where the difference appears; 2) the use of arguments in order to investigate the nature of that difference and to choose the best explanation for it, after having excluded possible alternatives. The second passage is based on an inference to the best explanation. In addition, in order to avoid misunderstanding, the same label of “phenomenal contrast method” has to be substituted with the simpler label of “contrast method,” so that the “phenomenal” attribute is not confused with the more technical use of the term “phenomenal”: such as phenomenal content, or quale. The elimination of “phenomenal” renders the method more impartial; it is a guarantee that the method is a neutral tool that is used to discover the

The Method of Contrast and the Perception of Causality in Audition

85

property represented in our perceptual experience. The difference that holds between the target and the contrasting experiences is a perceptual difference. This difference is based on an intuition that constituted the first passage of the method and has the specific properties that I am going to describe. Among the three available applications of the method listed above, I shall state that only the third one—where the scenario remains the same but a perceiver perceives it in two different ways—represents a good application of the method.

3. First passage of the method Siegel (2006, p. 130) claims that the phenomenal contrast method appeals to introspection only indirectly. She admits that it is just the starting point of the method, useful to the extent of helping to rule out inadequate contents that may be proposed as forming part of the experience but, at the same time, it is not a totally reliable tool: Introspection can rule out many proposed contents as inadequate to the phenomenal character of the experience, and thus does not seem to be completely useless as a means of finding out which contents an experience has. It thus seems to take us some way toward finding out which contents a specific experience has. But it does not take us far enough. Both the colorshape hypothesis and the cherry-content hypothesis seem prima facie plausible—neither is obviously phenomenally inadequate. If introspection could tell us which of them were correct, then it would tell us directly about the phenomenal character and content of experience to a degree of precision that would be needed to decide between these hypotheses. And if it did that, one of the hypotheses would seem obviously phenomenally inadequate. Since it doesn’t, we can conclude that introspection does not decide between these hypotheses.

In the above passage, Siegel refers to the impossibility for introspection to establish which theory is correct in the case of a contrasting situation. For instance, supposing you are looking at a bowl of expertly crafted wax cherries, you can affirm either that your visual experience represents the colours and shapes of the wax fruits, but does not go so far as representing that they are fruits, or that your visual experience represents the being property of cherries, and its contents include that there are cherries in the bowl. Introspection will not help you to tell which of the two representations is correct, so it is a limited tool, limited to the construction of the minimal pair of the contrast method. Actually, even if Siegel sustains a relative power of introspection, her examples of the application of the method, in the pine tree case and in the case of the drawn curtains

86

Chapter Six

and room illumination, do not use introspection with parsimony. This claim will become clear only after we would have defined the requirements of the intuition upon which the first passage of the method is based. Intuition is part of the first stage of the method and it is required to make the minimal pair acceptable, namely to make it evident that there is a clear difference between the contrasting and the target experiences. Arguments do the further job of deciding the best explanation of that difference. In order to be widely accepted, the difference between the two experiences has to have two features between which a tension is created: the difference has to be immediately evident, and, at the same time, it does not have to be easily explicable, it has to be complex and multifaceted. The McGurk effect shows how an intuition of a perceptual difference with those properties is applied. It is perceptually evident to every normal perceiver that the multimodal perception under this illusion would change with the two different ways in which the speaker moves his mouth. There is a difference in the way the two scenarios are perceived. We need an explanation that is not already given in the presentation of the McGurk effect. Generally speaking, the immediate evidence will contribute to the general agreement on the intuition of the difference; a serious explanation of it should avoid assuming that the specific intuition of a difference in itself already contains a possible path to follow in order to define the nature of that difference. For instance, Jackson’s black and white room is a very good example of the use of an intuition of a difference that possesses the afore mentioned characteristics (Jackson 1986). Mary, a neuroscientist who has all the possible knowledge regarding colours, is forced to live in a black and white room her entire life. When she escapes from the room and sees a ripe tomato for the first time, she is going to learn something new, namely what it is like to see red. Without evaluating the validity of the argument, Jackson’s reasoning is based on a very clear, immediately evident intuition, such as the fact that there is a difference between Mary’s knowledge in and out of the room. It is hard to deny that such a difference exists, as it is vividly and unquestionably apparent. Yet, providing an explanation of the nature of such a difference is not an easy task,2 and such a difficulty is the confirmation of the complex and multifaceted character of intuition. If a specific difference in a minimal pair were enough to offer a guide for a possible interpretation, the power of argumentation, which is central in the second passage of the method, would be enormously diminished. Kriegel’s method of knowability3 (2007, p. 131) will help to sketch a very general picture of the way in which it is possible for us to acquire the

The Method of Contrast and the Perception of Causality in Audition

87

intuition of a difference. Kriegel claims that in order for knowledge to be phenomenologically manifest, it has to be first-person knowable and, since we can have a first-person knowledge of the contents and attitudes of our conscious judgements, such contents and attitudes are phenomenologically manifest. I do think that the intuition of difference is immediately evident only if it is first-person knowable. As in Kriegel’s method, also in my analysis of intuition, being phenomenologically manifest and being firstperson knowable are intertwined traits. If the intuition is first-person knowable, then it is probably also phenomenologically manifest. With “phenomenologically manifest,” I mean what simply appears to constitute our perceptual experience. Therefore, I am not using “phenomenologically” as a technical term, as if the basic difference that constitutes the starting point of the contrast method were already considered a kind of subjective property, a quale, or something that is irreducible to the representational content. I am using the term in a more neutral way, so that “phenomenologically manifest” refers to the status of the intuition which has to appear evident to us. The way in which this intuition appears evident, or is manifest to us, is through the first-person knowledge. Regarding first-person knowledge, I cannot give an exhaustive notion of it but, to follow Kriegel’s example, it is possible to define it via the difference between first-person knowledge and third-person knowledge: Consider the fact that I am right now visualizing a smiling kangaroo. Both you and I are in possession of knowledge of this fact: we both know what I am visualizing. Yet we know this differently. I know it effortlessly, you know it effortfully; I know it without the mediation of inference, you know it with the mediation of inference; I know it quickly, you know it slowly. One may disagree, or be unclear on, one or all of these characterizations of the difference between my knowledge and yours. But surely there is a difference between the two. However one ends up characterizing the difference between my knowledge and your knowledge of the fact that I am visualizing a smiling kangaroo, it is clear that I know it one way whereas you know it another way. The way I know it we will call “firstperson knowledge”, the way you know it we will call “third-person knowledge” (ivi, p. 132).

Being effortless, quick and non-inferential (ivi, p. 133), first-person knowledge is what makes intuition immediately evident. The second trait of the intuition of difference, namely it being complex and multifaceted, would be explained away by the fact that we do not possess any immediate plausible explanation of it but, as stated by the second passage of the method, the choice of the best explanation has to be made through argumentation.

88

Chapter Six

4. Analysis of the three options based on the first passage of the method In this paragraph I shall evaluate if the specific intuitions of the difference on which the five examples of the method are based respect the characteristics I listed, namely immediate evidence and the complexity and multifacetedness traits. Therefore, the analysis will concern the first passage of the method and will not evaluate if the explanation of each example offers sound arguments to purport its validity. This will be tackled in the last paragraph of the paper, but only in relation to the property of causality. I would claim that example b) of the third option, where the scenario remains the same but the perceiver perceives it in two different ways, is the best application of the method. But I shall start by analysing the other examples. The example of the pine tree, instance a) of the third option, is a case in which the intuition of difference, even if it is immediately evident, is not complex and multifaceted, but in itself it already contains a possible explanation. The example was meant to show that it is possible to perceive a pine tree as a tree and as a pine tree. The ability to perceive a pine tree as pine tree is due to the acquisition of a recognitional disposition that is reflected in the phenomenological difference between the visual experiences you had before and after the recognitional disposition was fully developed. Furthermore, such recognitional disposition might be result from a process of learning. Both facts, the idea of recognitional disposition and the idea of a process of learning, seem to arise as the most spontaneous and immediate explanation of the intuition of difference, as if the intuition in itself already contained such an explanation and as if introspection alone was sufficient to give such an explanation, like an invasive suggestion. We said that in order to limit the use of introspection, it would have to be used only as a starting point to construct the target and the contrasting case. For instance, in the pine tree example its use is pervasive, therefore the example is not a good one. The same kind of criticism—the idea that the core intuition already contains its explanation so that it is not complex and multifaceted, even if it is immediately evident—might be applicable in the other examples, albeit with a small difference. In the first example, the cinema screen argument, there is a property of the scenario that changes, namely the screen’s colour property, and such change corresponds to a difference in the perception of the perceiver. It seems that when the argument entails a different scenario between the contrast and the target experience, this difference already suggests the corresponding difference in terms of the perceiver’s

The Method of Contrast and the Perception of Causality in Audition

89

perception. For example, the fact that the cinema screen argument bases its intuition of difference on the represented property of the screen would contain the explanation of the intuition in itself. As if the changed property of the screen was the cause of the difference in perception and, by virtue of it being the cause, it also gives the possible explanation of that difference, using introspection excessively. As I stated beforehand, introspection has to be used only to state the intuition of difference, to present it, without giving any additional clue towards the explanation of it. The examples of Prosop and Aesop and Siegel’s synchronic argument are not acceptable for the same reason for which the cinema screen argument is not acceptable. In Prosop and Aesop example, the scenario, that is mommy’s face, remains the same. There is no property of the scenario that changes, but this is perceived in a different way because of the prosopagnosia of Prosop. The two brothers perceive their mother’s face differently, so the idea is that the high-property of being a mommy’s face can be represented in the content of perceptual experience. Even though the intuition of difference is immediate enough here, as in the case of the cinema screen argument, the fact that the brothers perceive it objectively in a different way would already guide, through introspection, the explanation of such a difference. And that explanation, suggested by the stated difference between the twins, does not make the intuition complex and multifaceted. The stated difference in this case lies not in the scenario’s architecture, but on the side of the perceivers. At the same time, Siegel’s synchronic example faces the same problem. In her example, the scenario changes, since the curtain in the contrasting experience is translucent, whereas in the target experience it is not. The curtain case is described in such a way that the colour and thickness of the curtain differ in the two scenes, and it would be plausible to infer that by virtue of this difference there are corresponding differences in the contents of experience. Thus, the difference of the scenario would already force one to give an explanation, letting intuition lose its complexity and multifacetedness. Siegel actually explains the difference in terms of the representation of causality—since what is at issue is not the relation between the curtain and light in general, where the representation of colour and thickness would matter, but the relation between the lighting of the room and the uncovering of the window—but, at this stage, her answer is not relevant, even if she is right. On the contrary, what matters is the fact that the intuition of difference implicitly carries its possible explanation. The first characteristic of the intuition of difference, immediate evidence, is respected in both Siegel’s and the twin’s examples, since both propose a

90

Chapter Six

situation in which the difference of perception between the target and the contrasting experience is undoubtedly plausible. The example of the ball landing in a potted plant is the ideal application of the method of contrast. Both of the necessary features are respected. It is likely that we might perceive the landing of a ball in a potted plant and the illumination of the room—if the events happen without a big temporal gap—either as somehow united, tied, or as two distinct and separate events. It is immediately evident that the two events could be perceived in two different ways, but we would not find the explanation of this difference as immediately as the intuition in itself. And this is due to the fact that the scenario remains the same; there are no changes in the property represented. In addition, the perceiver is constituted by a single person, that is there are no twins with an evident difference which might influence the explanation of it. Furthermore, the intuition remains complex since there is no visible and obvious explanation of it. Therefore, the tensive aspect of the intuition of difference is preserved, therefore the use of introspection is confined within a legitimate use.

5. The perception of causality in auditory perception Siegel’s case of the ball landing in a potted plant is a good way to apply the method of contrast, at least with respect to the first passage of the method. The second passage consists in the evaluation of the different possible explanations of the intuition of difference, ruling out implausible accounts and leaving just the best explanation. Instead of assessing her arguments, I shall propose a different example. The idea is to present two examples, a synchronic and a diachronic case, which will show that we can perceive causality within the sense modality of audition. Therefore, my diachronic case will have the same schema as that proposed by Siegel. Then, I shall evaluate the best explanation of the intuition of difference, and I shall state that the perception of causality is the best answer. I shall also present a synchronic case that, on the contrary of Siegel’s, will contain the features necessary to intuition. I shall conclude with an evaluation of the possible explanations of that last example, arguing that we can auditorily perceive causality. The diachronic example is the following: suppose you are in a country fair and you decide to have a go at the shooting gallery. The task is to shoot a gun and try hitting a target in the form of a small iron bell. It is possible for a perceiver to hear these two sonorous episodes, the gunshot and the bell’s jiggle, in two different ways, either as two unified events,

The Method of Contrast and the Perception of Causality in Audition

91

bound by an impression of causality, or as totally separate and distinct events. If Siegel’s potted plant example works, at least in the general formulation of the example, then, also in our case, we have a clear and immediate intuition of a difference between the two experiences, excluding an already suggested explanation. Therefore, we may say that such intuition is complex and multifaceted. Regarding the synchronic case, let us suppose that in the exact moment in which you hear a note A produced by a single horn you hear a sound produced by a crystal glass that resonates at the same frequency of the horn. The idea is that, as in the diachronic case, you might perceive the two sonorous episodes as somehow united or totally distinct, and the immediacy and the complexity of the intuition of difference is preserved.4

6. The second passage of the method The second passage of the method consists in the evaluation of the possible explanations of the difference. My explanation is that the sense of unity we experience is due to the perception of causality, but it is possible to affirm that what helps us to perceive a certain unity in both cases is a form of Husserlian retention. Suppose you hear five sounds: the jingle of jiggling keys, the stomping of a foot, the creak of a door, the clink of a cup and the rustling of a peace of paper. Husserl (1928) claims that if you compare this auditory experience to the experience of a series of five sounds as parts of a melody, you would perceive the notes of the melody as unified in a way in which the five motley sounds are not perceived, even if at each moment you could remember the sounds from the previous moments. Husserl’s idea is that in the case of melody, we “retain” the previous notes, which is a form of remembering, whereas in the motley sound case we do not. The opponent of the causal content thesis would say that the difference between the contrasting and the target experiences is due to the sense of retention which is present in the target but not in the contrasting case. We can reply to this objection by saying that in the diachronic case we are talking of motley sounds, not a melody of sounds, so that we cannot refer to retention. Also in the synchronic case, even if there is a horn which produces the note A, the note is related to the sound of the crystal glass through resonance, not melody. It is possible to disregard causality by stating either that the impression of unity can be explained away by reference to retention and not causality, or that we do not experience any sense of unity at all. In addition, the difference may be due to non-causal contents. For example, in the diachronic case, one would say that there is a time gap between the shot

92

Chapter Six

and the bell’s jiggle that varies in the two experiences. And this variation is responsible for the difference in their perception. But I am assuming that the scenario remains exactly the same, the time gap between the shot and the bell’s jiggle also remains stable, so that the difference is not due to this non-causal content. The same answer is applicable to other possible noncausal content explanations, such as the duration property of the respective sounds of the gun and the bell in the diachronic case, or the sound volume of the horn’s note A and the crystal glass sound. The contrast method is an excellent strategy that enables us to discover that a high-order property, such as causality, can be part of our auditory perception. Nevertheless, the intuition of difference has to have specific features in order for the method of contrast to be acceptable. Within the experience of auditory perception, via a diachronic and a synchronic case, it is possible to perceive causality as a relation between two different sonorous episodes.

References Armstrong, D. M. 1968, A Materialist Theory of the Mind, London: Routledge and Kegan Paul Butterfill, S. 2009, “Seeing Causes and Hearing Gestures,” Philosophical Quarterly 59 (236), 405; now in Hawley and Macpherson (2011), 36 Duncker, K. 1935, “Zur Psychologie des Produktiven Denkens”, J. Springer, Berlin; English translation: “On Problem-Solving”, Lynne S. Lees, Washington: American Psychological Association, 1945 Grice, H.P. 1961, “The Causal Theory of Perception”, Proceedings of the Aristotelian Society, Supplementary Volume 35, 121 Hawley, K. and F. Macpherson (eds.) 2011, The Admissible Contents of Experience, Chichester, West Sussex (UK) - Malden, (MA): WileyBlackwell Husserl, H. 1928, “Vorlesungen zur Phänomenologie des Inneren Zeitbewusstseins (1893-1917)”, hrsg. von Martin Heidegger, Halle: Niemeyer; English translation: “On the Phenomenology of the Consciousness of Internal Time”, J.B. Brough, Dordrecht: Kluwer, 1980 Jackson, F. 1986, “What Mary Didn't Know”, Journal of Philosophy 83: 291 Kriegel, U. 2007, “The Phenomenologically Manifest”, Phenomenology and the Cognitive Science 6, 115 Koksvik, O. Forthcoming, “Phenomenal Constrast: A Critique”

The Method of Contrast and the Perception of Causality in Audition

93

Lewis, D. 1980, “Veridical Hallucination and Prosthetic Vision”, Australasian Journal of Philosophy 58 (3), 239 Michotte, A. 1963, The Perception of Causality, Methuen: London Nida-Rümelin, M. 2002, “Qualia: The Knowledge Argument”, The Stanford Encyclopedia of Philosophy, Fall 2008, http://plato.stanford.edu/entries/qualia-knowledge/ Siegel, S. 2006, “How Can We Discover the Contents Of Experience?” Southern Journal of Philosophy 45, 127 —. 2009, “The Visual Experience of Causation”, in Hawley and Macpherson (2011), 150 —. 2011, The Contents of Visual Experience, New York: Oxford University Press

Notes  1

I shall not discuss this third option in this paper. It would be enough to look at the Stanford Encyclopedia of Philosophy entry on “Qualia: The Knowledge Argument” (Nida-Rümelin 2002) to realize that many different explanations of the new knowledge Mary gets out of the room have been given, and to verify how Jackson’s intuition of difference is not immediately definable, and that, in addition, plenty of arguments are needed to support one view or another. 3 Kriegel considers his method as an alternative to the method of contrast, an alternative that does not face the problem of not distinguishing between contingent phenomenology and necessary phenomenology. 4 Siegel aims to show not that we can perceive causality, but that the property of causality is part of the representational content of our experience. For her, experiential representation of properties or relations is not factive, so you can represent that p even if p is not true. In fact, her examples are cases in which there is no real causal link between the events, nonetheless the property of causality is considered as being represented as a part of the content of experience. In my paper I shall not discuss the difference between perceiving and having an experiential representation of causality. For this reason, I proposed two cases where the sonorous episodes are linked by a real causal relation, so that, at least with regard to my examples, I do not have to draw the difference between the auditory perception of causality and the experiential representation of it. 2



CHAPTER SEVEN ACTUAL-SEQUENCE FREEDOM CAROLINA SARTORIO

1. Introduction Actual-sequence views of freedom (the kind of freedom relevant to moral responsibility) are motivated by “Frankfurt-style” scenarios, scenarios of the kind that Frankfurt once offered in his argument against the “principle of alternative possibilities:” Principle of Alternative Possibilities: The kind of freedom relevant to moral responsibility requires the ability to do otherwise.1 Frankfurt-style scenarios motivate the rejection of the principle about alternative possibilities because they apparently show that the kind of freedom relevant to moral responsibility does not require the ability to do otherwise. Here is an example of a Frankfurt-style case: Frank and Furt: Frank makes the choice to shoot Furt completely on his own, motivated by a desire for revenge (Furt has just seriously insulted him), and after having carefully considered reasons for and against killing Furt. A neuroscientist has been secretly monitoring Frank’s brain processes; had Frank hesitated, the neuroscientist would have intervened by directly manipulating Frank’s brain in such a way that Frank would still have made the same choice (the choice to shoot Furt). Given that the neuroscientist never feels the need to intervene, it seems that Frank makes his choice to shoot Furt freely, and is morally responsible for that choice. However, given the presence of the neuroscientist, it seems that he couldn’t have made any other choice. This suggests that the principle of alternative possibilities is false: the kind of



Actual-Sequence Freedom

95

freedom relevant to moral responsibility doesn’t require the ability to do otherwise. In recent years, compatibilists about determinism and free will have found this apparent lesson of Frankfurt-style cases encouraging because, if freedom doesn’t require the ability to do otherwise, this means that at least one important reason for thinking that determinism might be incompatible with freedom is absent: we can’t say that determinism is incompatible with freedom because it rules out the ability to do otherwise. If Frankfurt-style cases successfully show that freedom doesn’t require the ability to do otherwise, then, even if determinism ruled out the ability to do otherwise, we could still be free, in the sense that matters for moral responsibility. At the same time that Frankfurt-style cases cast doubt on the principle of alternative possibilities, they motivate an alternative picture of freedom: one according to which the kind of freedom relevant to moral responsibility is exclusively a function of “actual sequences.” Frank’s choice to shoot Furt appears to be free, in the relevant sense, because he made it completely on his own, on the basis of his own reasons, and, in particular, not compelled by the neuroscientist in any way. In other words, the actual sequence of events issuing in his act appears to be of the right kind for his choice to be free. This is so even if the presence of the neuroscientist guaranteed that he would make that very choice. Even if he couldn’t have made any other choice, what matters to whether his choice was made freely is how he actually came to make it, and not whether there was any alternative open to him at the time. This is, apparently, the important lesson that Frankfurt-style cases teach us. So, according to an actual-sequence view of freedom, freedom is “exclusively” a function of actual sequences (and not at all a matter of the existence of alternative possibilities). It is natural to understand this claim as a supervenience claim about actual causal histories, or, at least, as a claim from which the supervenience claim follows. The supervenience claim states: Supervenience: An agent’s freedom with respect to X supervenes on the actual causal history of X (or on the relevant elements of the actual causal history of X).2 That is to say, if two acts have (relevantly) the same causal history, the acts are equally free or non-free. In other words, there cannot be a difference in freedom without a (relevant) difference in the causal history. In this paper I will examine some potential challenges to the supervenience claim (and thus, to actual-sequence views of freedom,



96

Chapter Seven

understood in this way). Those challenges arise from certain examples that advocates of the principle of alternative possibilities once presented in favor of their view, in order to counteract Frankfurt’s attack against the principle. Here I will not be interested in the challenge that those examples pose to Frankfurt’s argument, but only in the challenge that they pose to an actual-sequence view of freedom, and to the supervenience claim in particular. I will argue that the challenge can be successfully addressed by focusing on the role that causation (the relevant concept of causation) has to play in an actual-sequence view of freedom.

2. Challenges to supervenience Consider the following pairs of scenarios, originally offered by van Inwagen (1983) in his defense of the principle of alternative possibilities against Frankfurt’s attack: Phones: I witness a man being robbed and beaten. I could easily pick up the phone and call the police, but I decide against it, out of a combination of fear and laziness. No Phones: Everything is the same as in the Phones scenario except that, unbeknownst to me, I couldn’t have called the police: the phone lines were down at the time. Not All Roads Lead to Rome: Ryder’s horse, Dobbin, has run away with him. Ryder can’t slow Dobbin down, but he can steer him in different directions with his bridle. When they get to a crossroad, Ryder realizes that only one of the roads leads to Rome. Since he hates Romans and wants some Romans to get hurt, he steers Dobbin in that direction, predicting that some Romans will get hurt by the passage of the runaway horse. As predicted, some Romans are hurt as a result. All Roads Lead to Rome: Everything is the same as in the Not All Roads Lead to Rome scenario except that, unbeknownst to Ryder, in fact all roads lead to Rome (his belief that only one road leads to Rome is false). Van Inwagen points out that the agent’s moral responsibility is importantly different in each of the cases in the pairs. In both Phones and No Phones I am responsible for not trying to call the police; however, I am responsible for not calling the police in Phones, but not in No Phones. Similarly, Ryder is responsible for intending the harm to the Romans in both Not All Roads Lead to Rome and All Roads Lead to Rome; however,



Actual-Sequence Freedom

97

Ryder is responsible for the harm to the Romans in Not All Roads Lead to Rome, but not in All Roads Lead to Rome. Van Inwagen suggests that this supports the principle of alternative possibilities because it suggests that, when the agent is responsible in these kinds of cases (as in Phones, and Not All Roads Lead to Rome), the agent’s responsibility is accounted for, at least partly, by the fact that they could have done otherwise. I could have called the police in Phones, and, similarly, Ryder could have prevented the harm to the Romans in Not All Roads Lead to Rome; according to van Inwagen, this helps account for the agent’s responsibility in each case. Again, I am not interested in van Inwagen’s claim that these examples support the principle of alternative possibilities, but in a related claim that one could make: the claim that these examples undermine the supervenience principle that (I have argued) is central to actual-sequence views of freedom. Here is why one could make such a claim: although there are some important differences between the cases in each pair, these differences apparently don’t concern the actual causal histories themselves. What actually happens seems to be exactly the same, even if what happens in other possible worlds is not. For example, take Phones and No Phones. In both scenarios I fail to pick up the phone out of the same combination of fear and laziness, and I don’t call the police. Had I tried to pick up the phone and call the police, then things would have been different in the two cases: I would have succeeded in Phones, but not in No Phones. But, as a matter of fact, I never even tried to call the police. Similarly for Not All Roads Lead to Rome and All Roads Lead to Rome: in both scenarios Ryder steers Dobbin in the same direction (the direction that he takes to be the only road leading to Rome) motivated by the same desire to harm some Romans. What would have happened if he had decided not to steer the horse in that direction is different in each case, but (apparently) not what actually happens, or the actual causal history. These are, then, important challenges to the supervenience claim, and thus to actual-sequence views of freedom, to the extent that those views are cashed out, at least partly, in terms of the supervenience claim. These examples threaten to undermine the supervenience claim because, on the face of it, it appears that the actual causal histories issuing in the relevant outcomes are the same in each case, but the responsibility of agents, and the freedom relevant to responsibility, is quite different with respect to those outcomes. But this is strictly forbidden by the supervenience claim, which states that no differences in freedom are possible without an underlying difference in the actual causal histories.



98

Chapter Seven

3. Preserving supervenience I will argue that, despite initial appearances, the examples we have looked at are not genuine counterexamples to the supervenience claim, or to actual-sequence views of freedom. They are not genuine counterexamples because, although the freedom of agents is different (relative to the relevant outcomes), the actual causal histories are also relevantly different. In Sartorio (2011) I put forth a similar argument on the basis of certain claims about causation that (I argued) are intuitively plausible. Now I want to try a different strategy. This is not because I think that the earlier strategy fails, but because I think it has to be adequately complemented by some claim to the effect that those facts about causation are facts about the relevant concept of causation, or about the relevant kind of causal relation. Some philosophers believe in the existence of multiple concepts of causation (see, e.g., Hall 2004 and Hitchcock 2007). I won’t take a stand on this issue here. But what I want to suggest is that, if there are multiple concepts, then (of course) we should focus on the relevant one. And, I will argue, the relevant concept of cause is one that can be used to show that the apparent counterexamples to supervenience we have reviewed (and similar ones one could cook up) are not genuine counterexamples. Not only this but, as we will see, on the basis of that concept of cause, actualsequence views yield exactly the right kinds of results that one should expect to see in such cases. In other words, far from undermining those views, the examples we have reviewed support actual-sequence views of freedom, or are further evidence of their truth.3 What’s the concept of cause that is at play in contexts where the freedom and moral responsibility of agents is at stake? Arguably, it is one that is connected to the concept of moral responsibility in certain important ways. In particular, here is one way in which we tend to think that the concept of cause and the concept of responsibility are connected: under circumstances of the right kind, the existence of a causal connection can let an agent’s responsibility for something be derived from the agent’s responsibility for other things (things for which the agent is responsible in a more basic way). In other words, causation can act as a vehicle for the transmission of responsibility (I have developed this point more fully in Sartorio 2005a). Roughly, here is a general principle that captures this connection: (Causation and Responsibility) If an agent is responsible for X, X causally results in Y, and the relevant epistemic conditions for responsibility obtain (for example, the agent could foresee that X would causally result



Actual-Sequence Freedom

99

in Y, and in roughly the way it did), then the agent is also responsible for Y. More specifically, here is a particular instance of this principle that will be especially useful for our current purposes: (Causation and Blameworthiness) If an agent is blameworthy for X because, in particular, he could predict that X would causally result in Y, and X causally resulted in Y in (roughly) the way the agent predicted, then the agent is also blameworthy for Y. In other words, this principle states sufficient conditions under which an agent’s blameworthiness extends from one thing to another thing, and an important part of those conditions concerns the existence of a causal connection between the relevant items. So now consider, again, the putative counterexamples to the supervenience claim: the pairs Phones/No Phones and Not All Roads Lead to Rome/All Roads Lead to Rome. The first thing to note is that, in both pairs, we can identify a certain piece of behavior by the agent that the agent is blameworthy for. In the first pair, this is my failure to pick up the phone, or to try to alert the police. I am blameworthy for that, in both cases, because in both cases I had every reason to believe that I would have been able to call the police if I had tried. In the second pair, the piece of behavior that Ryder is blameworthy for is his steering the horse in a certain direction. He is blameworthy for that in both cases because in both cases he took that direction to be the only direction leading to Rome. Now, we also know that, in each pair of cases, there is one scenario where the agent is blameworthy for some relevant outcome and one where he is not. I am blameworthy for not calling the police in Phones but not in No Phones; Ryder is blameworthy for the harm to the Romans in Not All Roads Lead to Rome but not in All Roads Lead to Rome. I will now argue that, in conjunction with the Causation and Blameworthiness principle stated above, these claims entail that there is a difference in the causal histories between the scenarios in each pair. This is why. Consider, first, No Phones. As I have explained, the following claims are true of No Phones: (1) I am blameworthy for not trying to call the police (in particular, because I could predict that not trying to call them would result in my not calling them). (2) I am not blameworthy for not calling the police.



100

Chapter Seven

In conjunction with (1) and (2), Causation and Blameworthiness entails: (3) My not trying to call the police didn’t causally result in my not calling them, or it only did so in a “deviant” way—in a way that considerably departs from the way in which I predicted the causal chain would unfold. Given that it’s hard to defend the idea that a deviant causal connection exists in that case 4 , it arguably follows from this that no such causal connection exists between my failing to try to call the police and failing to call the police in that case. On reflection, this seems like a plausible result: it is plausible to think that the causal connection that would typically have existed between the two failures or non-doings is severed in this case, given that the phone lines were down. Although, normally, our failing to try to call someone results in our failing to call them, this doesn’t seem right when we couldn’t have called them even if we had tried. (Consider a different example in support of this: imagine that I think about calling my grandmother and I end up deciding against it, on an occasion in which, as it turns out, my grandmother was not home. Intuitively, in this case, my failing to call her doesn’t causally result in my failing to talk to her given that she was not home, and thus I wouldn’t have been able to reach her even if I had tried.) Note that the causal connection between my failing to try to call the police and my failing to call the police is intact in the Phones case, where there is nothing wrong with the phone lines. This means that there is a difference in the causal history of my failing to call the police between the two scenarios: the causal history includes my failure to try to call the police (or my failure to pick up the phone) in Phones, but not in No Phones. This difference in the causal history, in turn, adequately explains why my blameworthiness for failing to try to call the police extends to my failure to call them in Phones, but not in No Phones. This is because causation, the vehicle for the transmission of responsibility, is present in one case but not in the other. A similar argument applies to the other pair of cases: Not All Roads Lead to Rome and All Roads Lead to Rome. The argument can be reconstructed as follows. As I have explained, the following claims are true of All Roads Lead to Rome:



Actual-Sequence Freedom

101

(4) Ryder is blameworthy for steering Dobbin in a certain direction (in particular, because he could predict that doing so would result in some Romans being harmed). (5) Ryder is not blameworthy for the harm to the Romans. In conjunction with (4) and (5), Causation and Blameworthiness entails: (6) Ryder’s steering the horse in a certain direction didn’t causally result in the harm to the Romans, or, if it did, it did so in a “deviant” way. Again, given that it’s hard to defend the thought that there is a deviant causal connection here, this suggests that there is no causal connection whatsoever. And in this case too, on reflection, this is a plausible thing to say. It is plausible to claim that, when all roads would have equally led to the same outcome, steering the horse in one of those directions does not causally result in the outcome. But it does, when it is the only direction in which we could have steered the horse that would have resulted in the harm, as in Not All Roads Lead to Rome.5 So, here too, there seems to be a difference in the causal history between the two scenarios, and that difference explains why the agent’s blameworthiness does not extend to the harm in one of those scenarios (All Roads Lead to Rome). It does not extend to the harm in that scenario because the vehicle for the transmission of responsibility, causation, is not present in that scenario. As we can see, the apparent counterexamples to the supervenience claim are not genuine counterexamples: there is a difference in the causal histories that underlies the difference in freedom and responsibility. Not only is there a difference in the causal histories, but that difference can adequately explain why the agent is not responsible for the relevant outcome in one case although he is in the other: when he is not responsible, it’s because the vehicle for the transmission of responsibility is not present. As a result, far from undermining the view, the apparent counterexamples to supervenience can be used to support the view: they can be used to illustrate the way in which causation and responsibility interact, and the way in which freedom and responsibility are grounded in actual causal histories. Now, one might wonder: If the causal histories in the relevant scenarios are different in the ways I have suggested, why did they appear to be the same? In other words, why did the examples appear to be counterexamples to the supervenience claim?



102

Chapter Seven

The answer is: because in these cases the causal histories are different in a way that is not initially obvious. Take Phones and No Phones, for example. What I do, or fail to do, in those cases, is the same: I fail to pick up the phone, and I fail to call the police. But the causal connections between those omissions or non-doings is different in each case: my failure to try to call the police is causally connected to my failure to call them in one case, but not in the other. Similarly for the other pair of cases: Not All Roads Lead to Rome and All Roads Lead to Rome. What Ryder does is the same, in both cases: he steers the horse in the relevant direction, and what ultimately happens is also the same: some Romans are harmed. But, again, the causal connections between those events are different in each case, given certain facts of the circumstances in each case: Ryder’s act of steering the horse is causally connected to the harm in one case, but not in the other. The fact that the events that occur are the same, but the causal connections between them are not, helps explain why the fact that the causal histories are different was not obvious at first sight. We have seen that certain principles that connect causation with responsibility can help us rescue supervenience from an apparently serious threat. In the next section I argue that those principles can also motivate an account of responsibility for outcomes that is an especially good fit for an actual-sequence view of freedom.

4. Causal histories and responsibility for outcomes When are agents responsible for outcomes in the world? As we have seen, principles such as Causation and Responsibility, and Causation and Blameworthiness, express conditions (sufficient conditions) under which the responsibility of agents for certain things (say, their acts, or their choices) can extend to other things. The existence of these principles suggests a natural way of understanding responsibility for outcomes in the world: agents can be responsible for outcomes in the world to the extent that there exist such things from which their responsibility for those outcomes can be derived, given that the relevant conditions obtain. And, again, the relevant conditions that need to obtain will include a causal connection between those things and the outcomes in question. In other words, the principles that connect causation with responsibility motivate the following account of responsibility for outcomes in the world: Responsibility for Outcomes: An agent is responsible for an outcome in the world Y just in case there exists some X such that the agent is



Actual-Sequence Freedom

103

responsible for X, X causes Y, and the agent meets the relevant epistemic conditions for responsibility with respect to Y. This account of responsibility for outcomes has several important advantages. It is simple, natural, and well motivated. It is also a particularly good fit for an actual-sequence view of freedom. Actualsequence views focus on actual sequences. If we understand actual sequences as actual causal histories, as I have suggested we should do, then the natural thing to say is that agents can only be responsible for outcomes in the world when those outcomes have certain kinds of causal histories: causal histories that can be traced back to other things for which those agents are responsible, in a more basic kind of way. On the face of it, this account of responsibility for outcomes is preferable to any account that appeals to extra machinery to state the conditions under which agents can be responsible for outcomes. In particular, it is preferable to Fischer and Ravizza’s own actual-sequence account (the account developed in Fischer and Ravizza (1998), and then developed further by Fischer in several other places). Fischer and Ravizza’s account is notoriously complex, and it appeals to some substantial concepts (the concept of a “mechanism”, the concept of the “responsiveness” of a mechanism, the concept of a “triggering” event, and so on). I have suggested that none of this complex machinery is needed in order to extend the account of basic responsibility (whatever that account may be) to responsibility for outcomes. All we need is the concept of a causal history, which we already have built into the basic structure of an actual-sequence view (at least if we understand actual sequences in terms of causal histories, as I have suggested we should understand them).6

5. Conclusions I have argued that focusing on the concept of causation, or on the relevant concept of causation, can help actual-sequence views of freedom defuse an important objection and gain considerably more intuitive plausibility. Under the lens of the right concept of causation, what appeared to be counterexamples to actual-sequence views (and, in particular, to the supervenience claim that I think is central to such views) fail to be genuine counterexamples and, indeed, only help to make those views more plausible. There is a more general lesson to be learned from this. Given the focus of actual-sequence views of freedom on causal histories, we should expect the concept of causation to play a central role in such views. This is a role



104

Chapter Seven

that has been surprisingly overlooked by advocates of the actual-sequence model of freedom, as well as by its detractors. This is an important oversight, which can lead to significant misconceptions and misunderstandings about the virtues of actual-sequence views. My general conclusion, then, is that, in developing an actual-sequence view of freedom, one should play close attention to the important connections that exist between causation and the concepts of freedom and responsibility.7

References Fischer, J. and M. Ravizza 1998, Responsibility and Control: A Theory of Moral Responsibility, Cambridge: Cambridge University Press Frankfurt, H. 1969, “Alternate Possibilities and Moral Responsibility”, Journal of Philosophy 66 (23), 829 Ginet, C. 1990, On Action, Cambridge: Cambridge University Press Hall, N. 2004, “Two Concepts of Causation”, in Collins J., N. Hall and L.A. Paul (eds.) 2004, Causation and Counterfactuals, Cambridge, Mass: Massachusetts Institute of Technology Press, 225 —. 2007, “Structural Equations and Causation”, Philosophical Studies 132 (1), 109 Hitchcock, C. 2007, “Three Concepts of Causation”, Philosophy Compass 2 (3), 508 Sartorio, C. 2005a, “A New Asymmetry between Actions and Omissions”, Noûs 39 (3), 460 —. 2005b, “Causes as Difference-Makers”, Philosophical Studies 123 (12), 71 —. 2011, “Actuality and Responsibility”, Mind 120 (480), 1071 —. manuscript, Causation and Freedom Van Inwagen, P. 1983, An Essay on Free Will, New York: Oxford University Press Yablo, S. 2002, “De Facto Dependence”, Journal of Philosophy 99, 130

Notes  1

Frankfurt presents his argument against the principle of alternative possibilities in Frankfurt (1969). The principle of alternative possibilities had been widely accepted until then, by both compatibilists and incompatibilists about determinism and free will. Ginet (1990) and van Inwagen (1983) famously embrace it. Fischer and Ravizza developed the most sophisticated actual-sequence view of responsibility, based on the basic insight of Frankfurt cases, in Fischer and Ravizza (1998).



Actual-Sequence Freedom

105

 2

See Sartorio (2011 and manuscript). In Sartorio (manuscript) I examine, also, how the relevant concept of cause fits with other commitments of actual-sequence views. I argue, in effect, that actualsequence views of freedom can benefit greatly from a closer look at the underlying concept of cause and its key metaphysical features. 4 The concept of a deviant causal connection is typically taken to involve causal connections that include totally unexpected events or processes in the causal chain. For example, imagine that I fire my gun at a victim. Since I am not a very good shot, I miss; however, the bang happens to awaken an assassin who was secretly hired to shoot the victim, and who had fallen asleep; the assassin then walks toward the victim, shoots at him, and kills him. In this case I cause the death of my victim, but in a deviant way. Now, if there is a causal relation between my failure to try to call the police and my failure to call the police in No Phones, it doesn’t seem to be a deviant connection, in this kind of way. If my failure to try to call them causes my failure to call them, it’s, presumably, in the way I expected (my failure to try to call them simply results in my failure to call them, not by the intermediary of other unexpected events or processes). 5 Several recent theories of causation have this result. See, e.g., Hall (2006) and Yablo (2002). See also Sartorio (2005b). 6 In Sartorio (manuscript) I give an account of basic responsibility that is in some respects similar to Fischer and Ravizza’s own account, but in other ways it is importantly different. One main difference between our accounts is that my account respects the supervenience claim in a way that Fischer and Ravizza’s view does not. I argue that this is a main advantage of my actual-sequence view over theirs. 7 Part of the research resulting in this paper was made possible through the support of a grant from the John Templeton Foundation. The opinions expressed in this publication are those of the author and do not necessarily reflect the views of the John Templeton Foundation. I am particularly grateful to Al Mele, director of the Big Questions in Free Will project, for the support. I am also grateful to the participants in a discussion in the “Flickers of Freedom” blog in August of 2013, on an occasion where I presented some of the ideas in this paper, for their valuable comments and suggestions. Finally, I am very grateful to Fabio Bacchini for his generous invitation to contribute to this volume, as well as for inviting me to give a presentation on related topics at a meeting of the Italian Society for Analytic Philosophy that took place in Alghero in September of 2012. 3



CHAPTER EIGHT CAUSES AND BLAME: A PRAGMATIC ANALYSIS DANIELE SANTORO AND MARCELLO DI PAOLA

1. Premise One of the crucial aspects of the philosophical reflection on the foundation of the social sciences is the role that the intentional vocabulary should have in the explanation of human actions. Social sciences have made several attempts to cut off intentionality from the explanations of social phenomena. Psychological behaviorism was perhaps the most systematic reductionist program in this direction, but in general the very idea of providing a lawful account of human behavior in analogy with natural sciences boils down to excluding the intentional vocabulary from the social sciences, for there cannot be a law-like formulation of intentional phenomena. 1 In this paper, we contest this reductionist approach on pragmatic grounds. We defend the idea that the use of intentional vocabulary is not dispensable from social explanations because causal language is pragmatically dependent on the intentional vocabulary. In particular, we present and defend two claims. First is the idea that there is no distinction in principle between causal and intentional vocabularies, but that they rather presuppose a common linguistic practice in which we baptize someone responsible in the same way we baptize a cause. Examples of these practices are common in the ordinary language of social explanations (Section 2). Support for this view can be found in Wittgenstein’s remarks on the use of “cause” and “blame” in his later writings. We call this the “scapegoat conception of causality” (section 3). It maintains that the individuation of a cause is a reaction to an effect, and it is possible only against the background of a social practice of inquiring and blaming, thus tracing causes from effects. We hold that intentional and causal statements differ only in the role they play in the

107

Causes and Blame: A Pragmatic Analysis

same social practice of scapegoating causes, and thus that distinctions between the intentional and the causal domain should be traced back to their pragmatic dimension. We offer an account of how this social practice works, and, in the final part of the paper we propose to systematize the pragmatic relationship between cause and blame within a broadly inferentialist framework, arguing that intentional and causal statements differ only in their inferential role in the social practice of scapegoating causes (section 4).

2. Blaming causes and treating intentions as causes Attempts of explaining away the intentional vocabulary are very common in the social sciences, especially in the field of economics. In the domain of micro-explanations, the main argument for dispensation is given by the difficulty of providing a measurement of desires and preferences in which consumer’s behavior is commonly described. As a consequence, those attitudes are standardly replaced by quantifiable variables (paradigmatically, utility functions) that can be easily manipulated within idealized and abstract models. Likewise, the standard macro-economic analysis constructs these models as relationships between inputs and outputs that make no reference to the intentional vocabulary. However, despite this common attitude, in the ordinary language of folk psychology one finds also a reverse tendency to embed intentional expressions within the description of causal mechanisms. Ironically enough, this phenomenon is even manifest in the common parlance of economists, at least when they do not speak in the official idiom of textbooks. For instance, on a Wall Street Journal article from January 2011 on the explanation of the recent financial crisis, one could read a reprimend of the simplistic narratives offered by the Congress's inquiry commission about the causes of the 2008 financial crisis. The authors criticized the two accounts prevailing in the opinion of the commission: In a November 2009 article, Brookings Institution economists Martin Baily and Douglas Elliott describe the... common narratives about the financial crisis. The first argues that the primary cause was government intervention in the housing market. This intervention, principally through Fannie Mae and Freddie Mac, inflated a housing bubble that triggered the crisis. This is the view expressed by one of our co-commissioners in a separate dissent. The second narrative blames Wall Street and its influence in Washington. According to this narrative, greedy bankers knowingly manipulated the financial system and politicians in Washington to take advantage of homeowners and mortgage investors alike, intentionally jeopardizing the

108

Chapter Eight financial system while enjoying huge personal gains. That's the view of the six majority commissioners.2

While the first explanation is couched in causal terms, the second shifts to a psychological vocabulary: greed, manipulation, and intentional jeopardizing lead to the crisis. Yet—all things considered—both explanations appear to be legitimate, at least in this regard: both the failure in the pure mechanics of the system, as well as the intentional manipulation of the market, prompted a common reaction—blame. [The] two narratives have popular appeal: They each blame a clear entity, and thus outline a clear set of reform proposals. Had the government not supported housing subsidies (the first narrative) or had policy makers implemented more restrictive financial regulations (the second) there would have been no calamity [italics added].

Although one should not expect from a newspaper article the same accuracy of academic writing, the idea that certain entities such as governments and financial markets can be blamed for triggering crises does have its “its popular appeal”. Thus, unless one discards this language as metaphorical or incorrect altogether, some attention should be paid to the ordinary language of blaming causes. But what does it mean to blame a cause? The answer comes in two steps. The first consists in noticing that the ordinary explanation provided above takes the intentional vocabulary as perfectly meaningful, and draws a counterfactual inference replicating the reasoning one would make about causes: had not the first event taken place, the second wouldn’t have either. Likewise, the institutional intervention is directly blamed for its causal role: “had the government not supported housing subsidies, or had policy makers implemented more restrictive financial regulations, there would have been no calamity” (ibid). Thus, when the event figuring in the antecedent is an intentional action, the consequence can be blamed on the agent in virtue of its being counterfactually dependent on that action. This argumentative scheme, which we may call “blaming the cause”, is quite common within moral and legal reasoning, and yet often overlooked outside these fields. Along with this, we can also find a second aspect. According to the WSJ, the turning point in the contagion was that confidence and trust in the financial system evaporated, as the health of almost every large and midsize financial institution in the U.S. and Europe was questioned. The financial shock and panic caused a severe contraction in the real economy” [italics added].

109

Causes and Blame: A Pragmatic Analysis

Thus, not only institutions such as market and governments can be blamed, but also the lack of confidence and trust contributed to cause a contraction in the real economy. That is, intentional states are taken here as prompting causal chains. Not only we blame causes, but we also “take intentions as causes”. Both these linguistic practices are perfectly meaningful and offer counter-examples to the official idiom of social scientists. Yet, the problem with these usages is that they seem unfit to be modelled within a more regimented and precise language, particularly when the task is to assess what counts as a genuine explanation. Pure semantic analysis seems insufficient to disentangle this puzzle, for it would judge incorrect, or just metaphorical, expressions that seem perfectly meaningful. The main limit of semantic analysis in this case concerns the requirement of extensional equivalence for causal claims. According to this requirement, the truth of a causal statement like "x causes y" is preserved for all the substitution instances that refer to the same event. Yet, intentional expressions cannot guarantee the requirement of extensional equivalence, for it is not necessarily the case that a causal statement involving an action is preserved under every possible description of the action.3 Do we have an alternative? We think we do: our claim is that we should bite the bullet, and admit that these widespread linguistic phenomena are genuinely contentful. An explanation of their consistency can be found in some seminal ideas by Ludwig Wittgenstein, which the pragmatic analysis should systematize.

3. A scapegoat conception of causality In the Tractatus, Wittgenstein’s position about causation was radically anti-factualist: “the belief in a causal nexus is superstition” (5.1361). But Wittgenstein also seemed to embrace a different, less Humean insight: though there are no causal observables, the logical structure of the worldlanguage nexus is causal, since causation is the only form under which our descriptive systems can be conceived. Later on, in his remarks in Cause and Effect and On Certainty, he developed this view in a different direction, outlining a conception of causation as a retrospective account from effects to causes, deeply similar to the language game of holding someone responsible for a deed. Since the concepts of blame and cause share a common grammar of use, we baptize someone responsible in the same way in which we baptize a cause.

110

Chapter Eight

What did Wittgenstein mean? As a start, one should notice that his focal critique was not the concept of cause as such, but the surreptitious distinction between causes in the order of nature and causes in the order of action. An elucidation of the grammar of our language highlights common patterns of behaviour in causal happenings. This development begins with Cause and Effect, where he claims (contra Russell) that, in order to explain why we describe the world as causally structured, there is no need to postulate any direct intuitions of causal relations: it is enough to recognize that certain statements, describing a first event as the cause of a second, are simply never subjected to criticism. This is, for instance, the case of when I'm hit with a stick, experience pain, and “intuitively” know that the blow caused the pain. We may genuinely call this the “experience of a cause”, Wittgenstein says; but not because we are directly made aware of a specific cause, as if this sort of judgments were prompted by a some immediate acquaintance with relations among events. For there could be endless possible alternative causes for the pain: while the blow may only have the function of giving me the impression of touch, pain could actually be exploding inside me. Causal propositions are beyond doubt not because solidly grounded on intuitions or a priori categories—but because the question of whether they are grounded at all is never posed. Neither are they beyond doubt because they are unmistakably accurate: as in the case of pain, I cannot be certain about any specific cause. They are beyond doubt because one must be certain about there being a cause in general in order to make any statement about relations among events, or lack thereof. In Cause and Effect, Wittgenstein develops this idea through a Gedankexperiment: I look at two plants, a rose and a poppy. I am led to think that the macro-differences I see between them correspond to micro-differences in their seeds’ biological composition. Different seeds cause different plants: I doubt not that fine-grained genetic inspections would find the seeds to differ in some respect. Suppose now that the seeds are found to be identical. How to explain the rose and the poppy being two different plants? We would not know what to think. Suppose, instead, that we finally do find a difference, perhaps at the quark level. Wittgenstein still asks us to prove that the macro-difference between the two plants does not merely correlate to, but is causally determined by the micro-one; and of that, we can in no way be certain. We may keep on searching desperately; or we may simply baptize a cause: We also speak of ‘tracking the cause’: in a simple case we follow, so to speak, the rope, to see who’s pulling it.

111

Causes and Blame: A Pragmatic Analysis When we find such individual—how do we know that it is him, his pulling, the cause of the fact that the rope is moving? Do we establish that through a series of experiments? (Cause and Effect, 15.10)

We do not. When something happens, we look for something we call its cause, but at the root of the grammar of causation are not scientific observations, logical categories, or direct intuitions. There is action, acts of investigation. The search for causes is then a non-scientific, eminently practical activity. We react to the effect by searching the cause: He who follows the rope and finds who’s pulling can take a further step, and conclude: so this was the cause, —or rather, is it not the case that all he wanted to find was whether someone was pulling, and who? (Cause and Effect, 16.10).

The pragmatic reading of the grammar of causal statements highlights a peculiar and somewhat heterodox understanding of causality as a form of scapegoating. In paradigmatic cases, a scapegoat is pointed at as the cause of an otherwise mysterious misfortune that hits the group—but the individuation of scapegoats is not an experimental, much less a logical enterprise. In scapegoating, explaining coincides with blaming: the chain between the scapegoat and the misfortune it is blamed for , is never spelled out scientifically, and more than that, it need not be; all that matters is that someone is responsible for the misfortune—that “someone did it”: for if that is so, then something can be done back. In one case ‘he is the cause’ simply means: he pulled the rope. In other cases it means something like: these are the facts that I must change in order to eliminate this phenomenon (Cause and Effect, 16.10).4

The grammar of causation shows a family resemblance with that of blame and responsibility. In paradigmatic scapegoating cases, when a group is hit by misfortune, the language game of providing an explanation is enacted, in causal terms, with reference to some violation of social trust that broke the balance between the group and its environment. The causal chain runs from misfortune to the scapegoat, carrying an imputation of blame. Wittgenstein's remarks do not rule out the predictive use of causal statements characteristic of natural sciences, but they suggest that the grammar of causation is primarily backward-looking: a reconstructive or genealogical enterprise, not a predictive one. We call this a “scapegoat” conception of causality, suggesting that the way we use causal concepts—at least in ordinary causal talk—is part of a

112

Chapter Eight

more general practice of naming someone guilty. So understood, scapegoating appears as a two-ply conception in which the intelligibility of causal statements is possible only against a background of social practices of imputing blame and responsibility. This way, intentional and causal statements do not appear to differ in any metaphysical respect, and their role can be distinguished just by looking at their respective epistemic goals. Wittgenstein’s remarks on causal attributions can be formulated within a provisional theory. At least two aspects characterize this view: (i) the attribution of cause is a reaction to an effect, and its paradigmatic form is backward not forward looking;5 (ii) the primitive language game of causal attribution is the same language game of blame ascription; and it is a form of action, an act of investigation, not of speculation. In regards to the first aspect, it should be noticed that the attributive attitudes from effects to causes is not a mechanism, nor can it be captured within a formally valid inference. Rather, the attribution of causes and the attribution of responsibility reflect the same grammatical use, which does not depend on empirical evidence. As for the second aspect, it should be clear that the image of the cause as a scapegoat expresses an intentional reaction to the social facts of human existence. We do not observe causal relations; neither do we conceive what we observe through the mediation of a transcendental category of the intellect; nor do we intuit a cause. What we do is we proclaim—or baptize, or scapegoat—a cause.6

4. A pragmatics for scapegoating The primary concern of Wittgensteinian philosophers is to offer a view of meaning in terms of its use. According to a radical version of the semantics-pragmatics nexus, the meaning of a word or sentence is its use: this way, semantics simply collapses into pragmatics. According to a more moderate reading, semantic and pragmatic rules are descriptively autonomous, but meanings can be explained only in terms of a correlative practice of using linguistic expressions.7 We think that a correct analysis of the grammar of causation cannot be a reduction of semantics to pragmatics. Against this extreme reading, we believe that Wittgenstein’s remarks on causation should be understood in light of a moderate interpretation: meaning and use are descriptively autonomous and yet they cannot be analyzed apart from one another. An

113

Causes and Blame: A Pragmatic Analysis

account of this sort is what Robert Brandom calls “pragmatic analysis”. A pragmatic analysis of the meaning-use relations does not show what a speaker does when she says something, but it says what one needs to do in order to be able to say something (Brandom 2008). The goal is here to explain semantic competence in terms of the correlative practices or abilities of employing a given vocabulary, and then spelling out which conceptual capacities such practices or abilities presuppose.8 In order to be genuinely explanatory, the theory should say which relations between practices and vocabularies are required for full semantic competence. Our attempt is to systematize the scapegoating conception by looking at the specific meaning-use relationships that link the causal and intentional vocabularies. A pragmatic analysis of this kind should then explain what speakers mean when they say something in terms of what one needs to know how to do in order to say it. That is, it explains assertions in terms of what abilities or capacities one must possess in order to express those assertions. The original idea can be found in Brandom's Between Saying and Doing (2008: ch. 4), where he shows that “in using ordinary empirical vocabulary, one already knows how to do everything one needs to know how to do in order to introduce and deploy modal vocabulary” (Brandom 2008, p. 98). According to Brandom, the empirical vocabulary presupposes (pragmatically) a linguistic practice or set of abilities that are sufficient for a speaker to do whatever he/she needs to know to deploy the modal vocabulary. This is the Kant-Sellars thesis about modality. 9 Its goal is to show that the empiricist concern with the unintelligibility of modal concepts falls back on itself. For our purposes, we can modify the Kant-Sellars thesis in the following way: (1) in using ordinary causal vocabulary, one already knows how to do everything required in order to introduce and deploy intentional vocabulary.10 (2) The expressive role characteristic of intentional vocabulary is to make explicit semantic, conceptual connections and commitments that are already implicit in the use of causal vocabulary. We can employ the intentional vocabulary alone to express causal relations, but we cannot employ the causal vocabulary alone to express intentional relations. The asymmetry is due to the fact that the expressive power of intentional relations is stronger than the expressive power of causal relations. A corollary of the Kant-Sellars thesis is that the causal vocabulary is indeed unintelligible without presupposing a discursive practice employing the intentional vocabulary. Let’s start with claim (1), above.

114

Chapter Eight E

Considerr first the paraadigmatic anallysis of empirrical statementts such as “this is the ccause”. They are a not seman ntically autonoomous (though h they are descriptivelyy so) becausee, in order forr a speaker to be able to un nderstand “he is the caause”, he mustt have the com mpetence suffi ficient to infer from “he is the cause” that “he didd it” (or that “he is responnsible for it”), and vice versa. Such competence requires r that speaker s is ablle to master in ntentional concepts in two distinct ways: w in the modal m sense suuch that from “he is the cause” it cann be inferred that “it is imp possible that hhe is innocent”, and in the normativve sense such that he licensses as correct the inference from “he is the cause”” to “he is respponsible”. Claim (2), above, iss more straig ghtforward: iintentional vo ocabulary makes expliicit the semantic connection ns and commiitments associiated with causal conceepts. By sayinng “if he is th he cause, thenn he is responssible” the speaker is implicitly unndertaking th he commitmeent to the reject r the conclusion ““he could be innocent”. The praggmatic analyssis shows what a speaker must be ablee to do in order to usee the casual voocabulary. Following Branddom, we can represent this analysiss by means of a diagram:

Fig. 1. Meaaning-use relaationship betw ween the intenntional and the causal vocabularies. “PV-Suff” sttands for a Prractice that iss sufficient to deploy a Vocabulary; ““VP-suff” stands for a Vocab bulary that is su sufficient to speeficy what one needs to do in order to engage e in a Praatice; “Res VV”” stands for thee Resultant relationship aamong Vocabullaries. This relationship is prragmatically meediated by the practice.

115

Causes and Blame: A Pragmatic Analysis

The diagram represents the meaning-use relation between the intentional and the casual vocabulary, and shows that such relation is pragmatically mediated by the practice of employing counterfactuals. Consider the causal vocabulary that is the target of this pragmatic explanation. The question is what makes it possible for a speaker to master causal terms properly, i.e. to use them correctly within an inference. The answer is that it is sufficient for a speaker to engage in the inferential practice of counterfactual reasoning directed at singling out causes. Two features of these practices should be highlighted. The first is summarized below the rectangle of the diagram as the idea that practices deploying causal terms involve the use of counterfactuals. That is, there is no practice of causal attribution that can be meaningfully expressed without counterfactual reasoning. Singular causal expressions can be in the indicative mode, but a competent speaker can make those attributions only if he is also able to justify those attributions, say by reasoning that “c is the cause of e” because “e would have not occurred, had c not been the case”. The second feature is that the description of an inferential practice sufficient to deploy the causal vocabulary must contain terms of blame (and responsibility). In other words, possessing the vocabulary of blame and responsibility is sufficient to engage in the counterfactual practice that is, in turn, sufficient to deploy the causal vocabulary. According to this analysis, it follows that—akin the sum of vectors for physical forces—the causal vocabulary results from the intentional vocabulary. In this role of the practice consists the idea of a pragmatically mediated relationship between the intentional and the causal vocabularies. In conclusion, in using the causal vocabulary, one already knows how to do everything required in order to introduce and deploy the intentional vocabulary. In our “liberal” understanding of Wittgenstein, his remarks on the grammar of causation capture this idea of the relationship, and—more importantly—the asymmetry between the causal and intentional vocabularies: we can intelligibly phrase a causal relation by means of an intentional vocabulary (so, it makes sense to “blame a cause”), whereas we cannot intelligibly phrase an intentional relation by means of a purely causal vocabulary (so, it does not make sense to “cause a blame”). Such an asymmetry exemplifies as well the second claim of the Kant-Sellars thesis, that is the idea that the intentional vocabulary has the expressive role of making explicit the semantic properties of the causal vocabulary, while the opposite is not true. What also follows from the asymmetric relationship singled out by this analysis is an important corollary of the Kant-Sellars thesis: the causal vocabulary is indeed unintelligible without presupposing a discursive practice that already contains the use of intentional terms.

116

Chapter Eight

An example of the pragmatic explanation defended so far can be found in the language of economists we referred to in the first section. A salient feature of this discursive practice is that inferences made within this context are not formal, but material, i.e. they are inference whose correctness depends on the non-logical vocabulary contained therein. For instance, the inference from “Milan is North of Rome” to “Rome is South of Milan” is correct in virtue of the meaning of the non-logical terms “North” and “South”. No logical constants figure in it. The same is the case with “he did it” and “he is the cause”. It is in virtue of the meaning of the non-logical expressions “cause” and “he did it” (which stands for an attribution of blame), that one can make the inference Wittgenstein was pointing at. Likewise, within the social practice of imputing responsibility for the events that prompted the financial crisis, we can legitimately transpose the inferences containing intentional terms into causal inferences without losing semantic value. Here is an example: in the social practice of imputing responsibility, one can move from an intentionally loaded inference If “the government supported housing subsidies”, then “we blame the government for triggering the crisis”, to the causal inference If “the government supported housing subsidies”, then “the government intervention triggered the crisis”, in virtue of the shared premise, and by substituting the intentional term with a causal term (plus some linguistic adjustment). The substitution granted by the shared premise of the two material inferences is what we call a transposition. It is important to emphasize that the transposition can only happen among material inferences where the meaning of the terms involved is fixed by the discursive practice that treat those inferences as correct. After all, the terms “cause” and “blame” are not synonymous in every possible context of utterance, and therefore not every suitable redescription of a sentence containing “cause” with a sentence containing “blame” will have the same truth-value. Only in some suitable discursive practice the relationship among these vocabularies is salient. This is what happens in the informal talk of economists, in which knowing how to engage in the practice of imputing responsibility appears to be sufficient to know how to engage in the practice of individuating causes. Vice versa, in the social practice of individuating causes:

117

Causes and Blame: A Pragmatic Analysis

If “the government supported housing subsidies” then “the government intervention triggered the crisis” cannot be transposed into If “the government supported housing subsidies” then “we blame the government for triggering the crisis”. unless the game of blaming an agent is already in place. So, what we see here is that the social practice of individuating causes is not sufficient for the social practice of imputing responsibility. We can show the same result by making explicit the counterfactual reasoning underlying the material inferences: (a) the government supported housing subsidies, therefore we blame the government for triggering the crisis; is justified because (b) had the government not supported housing subsidies, there would have been no crisis. From a semantic point of view, we say that if (a) is true, then (b) is true, concluding that the truth of (a) depends on the truth of (b). But semantics does not reveal the commitments associated with (the proprieties of) the inferences involved. When we introduce the pragmatic dimension, we notice that being committed to the causal statement (b) presupposes that one is already able to make a commitment with regards to (a). That is, being committed to (a) is sufficient to know whether one is committed to (b) as well. In our example, committing oneself to the claim that the government should be blamed for the crisis is sufficient to understand and evaluate the corresponding claim that the government caused the crisis. Sufficiency relations of this sort are pragmatically, not semantically mediated: a competent speaker or theorist who knows how to use causal terms, shows already the abilities sufficient to engage in the practice of attributing responsibility and blame. One of such abilities is the capacity of counterfactual thinking.

118

Chapter Eight

5. Conclusion In this paper we have suggested that the causal and the intentional vocabularies are intrinsincally related, and we have provided an example of how pragmatic analysis can discern their relationships. A pragmatic analysis explains what people mean when they say something in terms of what they need to know how to do in order to say it. In other words, it explains assertions in terms of what abilities or capacities one should possess in order to make them. In our case, the pragmatic explanation shows that the vocabulary of causes can be deployed only if one knows how to use the vocabulary of blame. The semantic analysis fails to see how causal expressions can be explained by reference to intentional expressions. Yet, we do not suggest reducing the semantics of these terms to their pragmatic role, neither do we argue that the vocabulary of causes is explained away by the vocabulary of blaming. Causal and intentional vocabularies can be indeed described one apart from the other, and there is no surreptitious reenactment of teleologism in the domain of science in the argument we provided. What we rather pointed out is that the ordinary causal talk is not an autonomous practice that can be engaged in apart from the practice of attributing blame, intentions, and responsibility. If social scientists have grounds for vindicating a form of theoretical autonomy, it is not the ostensible peculiarity of their conception of causality.11

References Anscombe, E. 1957, Intention, Cambridge, Mass.: Harvard University Press Bach, K. 2002, “Semantic, Pragmatic”, in Campbell, J. K., M. O'Rourke and D. Shier (eds.), Meaning and Truth, New York: Seven Bridges Press, 284 Brandom, R. 2002, “Pragmatics and Pragmatisms”, in Conant J. and Urszula M. Zeglen (eds.), Hilary Putnam. Pragmatism and realism, London and New York: Routledge, 40 —. 2008, Between Saying and Doing. Towards an Analytic Pragmatism, Oxford: Oxford University Press Cappelen, H.W. and E. Lepore 2005, “Radical and Moderate Pragmatics: Does Meaning Determine Truth Conditions?” in Zoltan, Z. (ed.) Semantics versus Pragmatics, Oxford: Oxford University Press, 45 Davidson, D. 1980, Essays on Actions and Events, Oxford: Clarendon Press

119

Causes and Blame: A Pragmatic Analysis

Elster, J. 1989, Nuts and Bolts for the Social Sciences, Cambridge: Cambridge University Press Hedstrom, P. and R. Swedberg (eds.) 1998, Social Mechanisms. An Analytical Approach to Social Theory, Cambridge: Cambridge University Press Morris, Ch.W. 1938, “Foundations of the Theory of Signs”, in Neurath, O. (ed.) International Encyclopedia of Unified Science, 1 (2), Chicago: University of Chicago Press Rosenberg, A. 2008, Philosophy of Social Science, Philadelphia: Westview Press Strawson, P. 1962, “Freedom and Resentment’’, Proceedings of the British Academy 48, 1 Varzi, A. 2008, “Failures, Omissions, and Negative Descriptions”, in Korta K. and J. Garmendia (eds.) Meaning, Intentions, and Argumentation, Stanford: CSLI Publications Wittgenstein, L. 1969, On Certainty, edited by G.E.M. Anscombe and G. H. Von Wright, Oxford: Basil Blackwell —. 1976, “Cause and Effect: Intuitive Awareness”, Philosophia 6, (3-4), 409 Woodward, J. 2003, Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press —. 2007, “Causation with a Human Face” in Price, H. and R. Corry (eds.) Causation, Physics, and the Constitution of Reality, Oxford: Oxford University Press, 66 —. 2008, “Causation and Manipulability”, The Stanford Encyclopedia of Philosophy, Edward N. Zalta, ed., URL =

Von Wright, G.H. 1971, Explanation and Understanding, Itacha: Cornell University Press —. 1973, “On the Logic and Epistemology of Causal Relation”, in Sosa, E. and M. Tooley (eds.) 1993, Causation, Oxford: Oxford University Press, 105

120

Chapter Eight

Notes 1

Davidson’s Anomalous Monism is a sophisticated elaboration of this view. See Davidson (1980). For a recent overview of the lawlike statements in the social sciences, see Rosenberg (2008, chapters 1-2, 6). For a critical view of the idea of causal laws, see Elster (1989), and Hedstrom & Swedberg (1998). 2 See “Congress's inquiry commission is offering a simplistic narrative that could lead to the wrong policy reforms”, by Bill Thomas and Keith Hennessey, Wall Street Journal, January 27, 2011. 3 See Anscombe (1957) for a classical analysis of intentional statements. 4 This insight will be later framed in terms of manipulability, that is the idea that for relation to count as causal, it must be the case that by manipulating the antecedent will result in the manipulation of the effect. Thus, a genuinely causal relation is one which is potentially exploitable for purposes of control (Woodward 2008). The seminal work on causation as manipulability is Von Wright (1971, 1973). For a contemporary refinement of this view, see Woodward (2003, 2007). 5 Arguably, this idea has close similarities with Peter Strawson conception of “reactive attitude”. See Strawson (1962). 6 See Cause and Effect, 21.10. 7 The classical distinction between semantics and pragmatics is in Morris (1938). For an overview of the debate see Bach (2002). See also Cappelen, H., Lepore, E. (2005) for a somewhat similar distinction between moderate and radical pragmatics, and Brandom (2002), for a map of the different kinds of pragmatists views about the semantics-pragmatics nexus. 8 We follow Brandom in using the notion of vocabulary in a general way as a set of linguistic items, which are or contribute to the construction of meaningful expressions. According to this usage, general and singular terms, but also atomic sentences, as well as indexicals, are part of a vocabulary. See Brandom (2008, p. 1). 9 In Brandom's words: “One of Kant’s most basic ideas, revived by Sellars, is that this idea is mistaken. The ability to use ordinary empirical descriptive terms such as ‘green’, ‘rigid’, and ‘mass’ already presupposes grasp of the kinds of properties and relations made explicit by modal vocabulary” (Brandom 2008: 96-97). 10 The framework is modified this way: empirical vocabulary is replaced by empirical vocabulary, and modal vocabulary is replaced by intentional vocabulary. 11 A previous version of this paper was presented at the at the Tenth Conference of the Italian Society for Analytic Philosophy (SIFA, University of Sassari, Alghero, in September 2012) at the II Madrid Workshop on “Inferentialism in the Philosophy of Language, Mind, and Action” (UNED, Madrid, October 2012); and at the Seminar Series Lectiones Commandinianae (University of Urbino, April 2012). We thank the audience and the organizers for the discussion. Special thanks to Claudio Calosi, Enzo Fano, Pierluigi Graziani, Sven Rosenkranz, Filippo Santoni de Sio, and Achille Varzi for the helpful comments provided on those occasions.

CHAPTER NINE RESPONSIBILITY REGARDLESS OF CAUSATION FEDERICO L.G. FAROLDI

Felix qui potuit rerum cognoscere causas. —Publius Vergilius Maro, Georgicon 2. 490 In jure non remota causa sed proxima spectatur. —Francis Bacon, The Elements of the Common Law of England, 1630

This paper deals with the relationship between legal responsibility and causation. I argue that legal responsibility is not necessarily rooted in causation. My discussion deals with legal responsibility, except when I explicitly refer to other kinds of responsibility. The general claim I aim to disprove is that responsibility is descriptive because it is fundamentally rooted in causality, and causality is metaphysically real and founded. My strategy is twofold. First, I show (in §1) that there are significant and independent noncausal forms of responsibility that cannot be reduced to causal responsibility; second, in §2, I show that the very notion of causality is— lato sensu—not plainly descriptive. I will suggest that even causation is tied to evaluative elements, contrary to what is assumed by many theorists and practitioners working in normative domains. In §3, I give a brief account of the three most discussed contributions to the relationship between causation and responsibility (legal and moral) in the last 50 years: H. Kelsen’s Society and Nature (1943), Hart and Honoré’s Causation in the Law (1959) and finally M. Moore’s Causation and Responsibility (2009). I shall consider these contributions in light of the thesis I defend in this work.

122

Chapter Nine

Introduction: a bird’s eye view In this work causation will not be dealt with as an independent topic, i.e. causation tout court, but always in a certain normative system; thus, I will not deal with causation in the domain of natural sciences. “Cause” is a rich and polysemic term. Just using “cause” or “causation” does not explain anything—arguably, in fact, not only are there different concepts of causation, there are also different conceptions of cause and causation, in different disciplines, used to advance different aims.1 Dealing with causation in the law is a complex task due to two related terminological problems and not only for reasons intrinsic to the matter of causation. In law, there is no bijective relation between the words “cause” and “causation” and cause and causation themselves: each time the law speaks of “cause” and “causation”, it does not necessarily refer to cause and causation (a well-known example is, in Italian civil law, the use of “causa” to refer to the functions or reasons of contracts).2 Clearly, there is nothing special about the ambiguity of natural language employed in legal contexts, but we should be forewarned not to take this language at face value. We may be interested in causes with three different aims in mind. Those aims can significantly overlap sometimes, but can be kept separate for philosophical inquiries. They are: Causation1: forward-looking causation. This particular type is concerned with individuating future probable outcomes, given a certain state of affairs or a certain set of situations, i.e. “causes”. Causation2: backward-looking causation. This particular type is concerned with individuating those states of affairs or certain sets of situations whose prior existence necessarily (?) determine a given outcome. It has an explanatory aim. Causation3: causal responsibility. This particular type of causality is concerned with ascribing to an agent (not necessarily a living agent) a given outcome, “that his, her or its agency serves to explain and that can therefore plausibly be treated as part of the agency’s impact on the world (Honoré 2010).”3 Clearly, causation is relevant in all three declensions, but while causation1 is probably more considered in the legislative, law-making process, causation2 and causation3 are those which play the prime roles in (criminal) courts.

Responsibility Regardless of Causation

123

Let us now sketch a provisional map of the possible relationships between responsibility and causation.4 CN: Causation is necessary According to this view, causation is necessary for responsibility, but not sufficient. It means that one can be punished only for those offences one has caused, but not for all offences one has caused. This seems to be Moore’s view: he has to build up a complex theory of causation in order to be able to account for those threshold cases where apparently there are no traces of causation. Another vivid example of this view is Antony Duff’s jurisprudence, according to which causation is necessary (although not sufficient) for responsibility (cf. Duff 2008, 2009). CS: Causation is sufficient According to this rare view (possibly more common in tort law), one is liable for all offences one has caused, but not only for them. The fact that someone has caused an offence leads directly to responsibility, but this view does not exclude that there can be other foundations for responsibility. CNS: Causation is jointly necessary and sufficient According to this view, causation is jointly necessary and sufficient for responsibility. It means that one is criminally liable for all and only those offences he has caused. CnNnS: Causation is neither necessary nor sufficient According to this view, causation is neither necessary nor sufficient for responsibility. The basis for liability may be causation, but it may not be. I will defend this last view (that causation is neither necessary nor sufficient for responsibility) in the rest of this paper.

1. First argument: causation is not necessary for responsibility Betrachte einmal die Vorgänge, die wir “spiele” nennen. […] Denn, wenn du sie anschaust, wirst du zwar nicht etwas sehen, was allen gemeinsam wäre, aber du wirst Ähnlichkeiten, Verwandtschaften, sehen, und zwar eine

124

Chapter Nine ganze Reihe. [...] Wir sehen ein kompliziertes Netz von Ähnlichkeiten, die einander übergreifen und kreuzen. Ähnlichkeiten im Großen und Kleinen (Wittgenstein 2009, §66).5

This section argues that responsibility is not based on causation. In particular, I shall point out several kinds of non-causal responsibility. The phenomena—legal and moral—usually grouped under the umbrella term of “responsibility” are so conceptually diverse that they cannot admit a common metaphysical basis. The paradigmatic case of (criminal) responsibility is when the mens rea (the mental elements of the offence, depending on the jurisdiction) and actus reus (the “wrong deed”) pertain to the very same individual. There are, however, borderline cases, when either the mens rea or the actus reus are missing or are scattered across different people. When there is criminal responsibility just for actus reus regardless of mens rea for any material element of the offence, common law jurisdictions speak of strict liability.6 In this section, I focus instead on other kinds of “queer” responsibility, namely those I group under the term “collective responsibility”. 7 While in strict liability the actus reus and the possible sanctions pertain to the very same individual, collective responsibility places the various elements (actus reus, mens rea and possible sanctions) across several individuals.8 DEFINITION: Responsibility is collective when the actus reus, mens rea and possible liabilities do not pertain to the very same individual. &ollective responsibility has at least three different “realizations”: (i) group responsibility, (ii) shared responsibility, (iii) vicarious responsibility. Group responsibility is defined as when the responsibility for actions is ascribed to members of that group qua group members, regardless of any actus reus or mens rea they could have or not have had, done or exercised. A paradigmatic example is, I think, the responsibility for genocide imputed to Nazi party members (following Nuremberg) regardless of their intending or knowing the plan or putting it into practice. Shared responsibility is defined as when the responsibility for actions is equally imputed to members of a group qua members of that group, assuming that all of them need to do or intend to do the action in question; in the absence of this, everyone is held equally liable. Group responsibility differs from shared responsibility because the former requires a status

Responsibility Regardless of Causation

125

(being member of a group) regardless of any action or intention, whereas the latter assumes actions and intentions as prior. A paradigmatic example is, I think, the responsibility of cleaning a shared kitchen by members of an apartment block. All and individual members need to clean the kitchen, but it might be the case that, for special arrangements, only a part of the group is entrusted with this task. Were the kitchen unclean, all members would be held equally liable, regardless of previous arrangements among the parties. Vicarious responsibility is defined as when an individual is held responsible (and consequently presumably liable) for an actus reus someone else committed. Vicarious responsibility differs both from group responsibility and from shared responsibility because the individual held responsible or liable has—by definition—no bearing on the act in question, either factually or mentally (no actus reus nor mens rea)— although he can (and often ought to) exercise control over those for whom he is vicariously responsible. A paradigmatic example is, I think, the vicarious responsibility of an employer for his employees or that of an editor-in-chief for what appears in his publication.9 Causation is not necessary for responsibility: I have discussed three legitimate examples of responsibility where the ascription of responsibility is non-causal. In group responsibility, just being a member of a particular group is sufficient for being held responsibility for the (mis)deeds of the group, regardless of any personal causal contribution. In shared responsibility, there is the mens rea but not necessarily the caused actus reus. In vicarious responsibility, the responsible individual has no causal bearing on the act in question, which was committed by someone else, but responsibility is derived from a particular relationship or status one has. Despite their diversity, none of the forms of collective responsibility discussed in this section thus far are based on a traditional notion of causality. This seems the case for two reasons: first, the source of responsibility is not necessarily or solely individuated in agency, but rather in a given status of an individual or a group, a status more often than not resulting from normative relations; second, the causal contribution (or omission) of a particular individual cannot be directly linked to the actions for which one is punished. What one has (or has not) caused is severed from his responsibility. We have two ways out this impasse: first, to disown collective responsibility as an actual form of responsibility (because it is not causal-

126

Chapter Nine

based)—but this would be question-begging; second, we may acknowledge that there are forms of responsibility de facto non causal, and therefore conclude that causation is not necessary for responsibility. Another argument shows the conceivability of non-causal forms of responsibility. Within a strict consequentialist view, one can do away with causation—exactly as with desert: what matters is not (only) past deeds, but what happens next. If the state of affairs justifies placing responsibility (or blame, or punishment) on someone completely unrelated to the action in question, an act-consequentialist is prima facie compelled to accept that causation is not necessarily linked to responsibility (although it might contingently be). And since (act-)consequentialism is a legitimate (even if not necessarily true or correct) moral theory that can account for responsibility, then it is not even a conceptual truth that responsibility is necessarily based on causation: this would ultimately depend on the (meta–)normative theory of values one endorsed (his Wertanschauung). In this section, I have suggested that there are many substantially different phenomena grouped under the umbrella term of “responsibility”— many of which are even non-causal. None of these phenomena have anything descriptive (factual) in common, and thus, at least in these important cases, responsibility cannot be explained away by pointing to a (common) metaphysical, descriptive trait of the world, as causality is often interpreted.

2. Second argument: causation in law and morals is not descriptive Wie ist denn der Begriff des Spiels abgeschlossen? Was ist noch ein Spiel und was ist keines mehr? Kannst du die Grenzen angeben? Nein. Du kannst welche ziehen: denn es sind noch keine gezogen (Wittgenstein 2009, §68).10

The scope of this section is to hold the following thesis: even if causation were jointly necessary and sufficient to determine responsibility, this would not show that responsibility is non-normative, that is, grounded in something ultimately descriptive.11 The general argument can be summarized as follows: (i) causation is necessary for responsibility; (ii) causation is descriptive, or non-normative (because it is scientifically provable, etc.); therefore (iii) responsibility is necessarily descriptive, or non-normative.12 I will

Responsibility Regardless of Causation

127

reject this conclusion. Even assuming (i)—which has been discussed and rejected in §1, showing that causation is not necessary for responsibility— I shall show that (ii) is false, and that causation is neither descriptive nor non-normative. Now I shall put forward three arguments I think summarize the problems with premise (ii), that is, the consideration of causation (in law and morals) as descriptive: first, the problem of remoteness; second, what I call the “causal sorites”; third, the heterogeneity of actions with regard to omissions. The non-descriptive aspects of causation in natural sciences are well-known in philosophy of science. I shall assume the reader is familiar with this shared background and the three arguments I shall discuss are to provide further evidence to the thesis that causation is also non-descriptive in legal and moral domains. (i) Regressus ad infinitum/remoteness This problem is pretty familiar (also) to lawyers: how far should one go before causes become irrelevant? Several acts are jointly necessary for an action to happen, and these acts can be remote. Now, are these remote acts (still necessary for further necessary things to obtain) of legal relevance for the action in question?13 Take a murder. If the killer’s mother and father had not met, this particular homicide would not have occurred, because—it seems, on an intuitive reading—this particular killer would not have existed (and so on, back in time). From a purely descriptive point of view, the encounter of the killer’s mother and father is a necessary condition for the killer’s being there (and eventually committing the murder) so it should be considered a cause, or at least part of the relevant causal chain. In a counterfactual analysis, if the killer’s mother and father had not met, then this killer would not have existed and this homicide would not have been committed. Now, the law (and morals) is usually not interested in this kind of “causal” chains. Instead, judges and lawyers are more concerned with a specifically legal cause, also called proximate, that is “near and immediate, or directly traceable, or foreseeable (Feinberg and Coleman 2008, p. 603). The legal or proximate cause is called also cause-in-fact. Now, it seems very hard to justify such a choice on purely descriptive grounds. The only real difference I can think of in the murder example is a temporal factor—but this still does not clearly point out why more recent or more remote causes are to be preferred or discarded without a choice already made on our part. Please keep in mind that this argument has a specific diachronic dimension.

128

Chapter Nine

(ii) The causal sorites paradox A lesser problem—quite distinct from the “regressus ad infinitum”—is the causal version of the Sorites paradox: not how far back in time we should go to pick the relevant cause, but how much our cause has to contribute to (say) the offence. Let us grant that causation is metaphysically primitive (i.e. not reducible to other physical things or forces). Still, it is plausible that causation is a “scalar relation” (Moore 2009, p. 105) and therefore a matter of degree.14 This might be a problem with accomplice liability, if we, counterintuitively, admit that the accomplice has had a causal (albeit maybe non-physical) role in the offence. It is for the law (or for morals) to “draw the line”: to decide whether a certain degree of contribution is going to count as a cause or not. And it is apparent that this line is not descriptively determined, but always based on some sort of evaluation. Note that the “causal sorites” is quite different from the “regressus ad infinitum”: the former is a synchronic problem, while the latter is diachronic. (iii) Actions vs. omissions A notion of responsibility that necessarily requires a descriptive conception of causation has a problem with omissions. In fact, whereas omissions are of some importance at least for criminal responsibility, descriptive criteria for causation cannot easily track the intuitive distinction between actions and omissions. Either omissions cannot cause anything (and one cannot be responsible for an omission, contra the evidence from criminal law), or—in order to assign responsibility for omissions—omissions are causes: this either makes them coincide with actions, or escalates into counterintuitive conclusions (for example, that roughly the whole humanity should be held responsible for the effects of any omission whatsoever). Consider two situations where a person dies. Now, let us stipulate that situation (a), where a robber shoots a passerby, may be prima facie described as “killing”; another situation (b), where nourishment for a terminally ill patient is not provided, may be described as “letting die”. Do (a) and (b) differ? Moral philosophers do not agree on this problem, so let us keep our discussion to the legal domain. It seems that they may be considered different even in finer-grained cases, for instance in the distinction between active and passive euthanasia, and in radically different charges you can incur: in Italian criminal law, (a) would possibly get you a sentence for “omicidio volontario” [approximately: murder] (up

Responsibility Regardless of Causation

129

to 21 years in prison), while (b) would possibly get you a sentence for “omissione di soccorso” [approximately: duty to rescue] (1 year sentence or a 2.500-euro fine).15 If (a) and (b) were two different things (as criminal law seems to recognize), then a possible grounding would be identifying (a)—killing—with an action, and (b)—letting die—with an omission: actions would be causes, whereas omissions would not. On these grounds killing could be punished because someone actively caused it—in a legal (moral) framework where causation is necessary for responsibility. Given that, it is easy to see why in (a)—killing—one is responsible. But (b)—letting die—is rather puzzling. Either omissions are causes, in which case there should be no difference between responsibility for killing and letting die because the effects are the same, and the responsibility should be the same; or omissions are not causes, and since one has not caused anything with one’s omissions, no responsibility is warranted.16 Carolina Sartorio (e.g. 2009; 2012) argues that accepting this distinction between causal action and non-causal omission is an untenable position because it is metaphysically unfounded and because it would force us to accept absurd consequences. This is exemplified by the socalled “Queen of England’s problem” in terms of omissions: my gardener is responsible for my plants’ death, because if my gardener had watered my plants, they wouldn’t have died; therefore, the Queen of England is also responsible for my plants’ death, because, had she watered my plants, they wouldn’t have died. If we consider someone responsible for an omission, then (without further qualifications based on specific roles or requirements) we should also consider responsible all those who have not undertaken that action— because the descriptive criteria are the same—and literally all those imputable according to the the relevant criteria (age, (in)sanity, and so on). Thus, descriptive criteria are not enough for responsibility, which needs non-descriptive elements to discriminate between finer alternatives. An independent, “descriptive” view of causation cannot convincingly account for the traditional (and legally relevant!) distinction between killing and letting die, or more generally between actions and omissions: the law “treats omissions both as causes and yet not as causes (Moore 2009, p. 82).” In this section I have suggested that the very notion of causality is not plainly descriptive, but tied to evaluative elements.

130

Chapter Nine

3. Three views on responsibility and causation: Kelsen, Hart and Honoré, Moore In this section, I give an account of the three most discussed contributions to the relationship between causation and responsibility (legal and moral) in the last 50 years: H. Kelsen’s Society and Nature (1943), Hart and Honoré’s Causation in the Law (1959) and finally, M. Moore’s Causation and Responsibility (2009). I will engage with their contributions with regard to the position I have developed in this work.

3.1 Kelsen: Kausalität und Zurechnung Hans Kelsen (1886 – 1973), one of the major legal philosophers of the XX century, grappled with the problem of causation and responsibility in several works (cf. Kelsen 1939, 1943, 1960, 1973). In brief, Kelsen held that (i) imputation [Zurechnung] is the principle used to ascribe sanctions in normative domains; (ii) imputation, while being analogous of causality [Kausalität], is not based on causality; (iii) quite on the contrary, it is causality that, as a cultural category, stemmed from the principle of imputation, as the Ancient Greek word “aitía” indirectly shows. Thus, imputation (and thence responsibility) is not necessarily allocated on a causal basis: it may well be, but this does not have to be the case. This conclusion flows well with the rest of Kelsen’s pure theory of law: once we have the principle of imputation, it is for the actual laws to fill it with practical criteria. There is no need for responsibility to be based on causality. Kelsen’s position is comparable to mine, although his is based on other arguments and on a different perspective: arguably he has a Kantian understanding of causation and imputation as pertaining to two sharply distinct domains. We shall shortly see that not all legal theorists were of the same opinion.

3.2 Hart and Honoré: Causation in the Law Published in 1959, Hart and Honoré’s (hereafter HH) monumental study Causation in the Law sought to found (normative) principles for attributing moral and legal responsibility upon the (descriptive) principle of causation.

Responsibility Regardless of Causation

131

The principles for moral and legal responsibility are not “inventions of the law” but, rather, are “common-sense principles of causation” that are “part of the ordinary man’s stock of general notions” and based on questions of fact “similar to the conventional view of the law’s use of other highly general notions such as those of temporal or spatial location” (Hart and Honoré 1959, pp. 91-92). HH were concerned with the concept of causation commonly used by ordinary people (and therefore—for them—reflected in law via ordinary language) rather than those used by philosophers or physicists (cf. pp. xxxiii-xxxiv, 1-3). In an extreme synthesis, HH’s argument can be summarized in the following way: (1) legal responsibility (the criteria for) is justified if it tracks (the criteria for) moral responsibility;17 (2) causation is a necessary condition for moral responsibility;18 (3) the relevant concept of causation here is the concept commonly used by ordinary people in speech. I shall rhapsodically note three problems (also problematic within HH’s view). Linguistic Analysis Linguistic analysis (the method chosen by HH to tackle the issue of causation) does not exhaust empirical and conceptual aspects of the problem. It is thus a dubious method to investigate such a complex task. Which event is a cause? We have seen that not all necessary conditions (for the sufficiency of the set) are causes. Why is that the case? Because what can be a cause is not natural or descriptive, but depends on the aim we have to ascribe an action to that agent, and is therefore the evaluative part of the process. Which cause is relevant? Once we have a list of causes for our event, we still need to choose those which are relevant. Think of a so-called overdetermination case. A fire, deliberately started by John, reached Rachel’s house and was about to burn it down. Suddenly, a violent earthquake made the house collapse. Now, both events were independently sufficient to destroy the house. There is an important difference, though: while the fire was lit by John with the purpose of destroying Rachel’s house (let us take this for granted), the earthquake was a natural, “extraordinary” event. According to HH, both the fire and the earthquake are causal relevant factors, because each was independently necessary for the sufficiency of the set of factors that destroyed Rachel’s house. Now, is John to pay for reparation, even though the event that

Chapter Nine

132

destroyed Rachel’s house was the earthquake? It seems we must choose which cause is relevant for our purposes. If we adopt a mere chronological criterion, the earthquake was the most recent event and therefore, perhaps, causally responsible. But again, this sort of relevance decision seems based not on purely arbitrary factors, but at least on evaluative premises. In the end, HH overlapped and conflated the issue of natural causation with the issue of responsibility-attribution. They failed to see that the attribution of responsibility can be non-causal, and therefore that natural causation and the attribution of responsibility cannot be accounted for in the same (causal) way.19 Natural causation and responsibility are different—though not unrelated—“things”.20 In this respect, HH’s position is quite different from the argument I adopted earlier in this paper.

3.3 Moore: Causation and Responsibility I shall briefly consider the position of Michael S. Moore, expressed in particular in his latest book, Causation and Responsibility. I cannot consider Moore’s rich and complex position fully here. Thus, I shall try to underline and discuss the points most relevant to my thesis. Moore thinks, for instance, that legal responsibility should closely track moral responsibility,21 and that moral responsibility is based on natural, empirical properties such as causation, which is necessary for it and purely descriptive.22 I shall now quote two brief passages and then try to formalize and dismantle his argument. The metaethical postulate is that moral responsibility [...] supervenes on natural properties like causation, intention, and the like. The postulate of legal theory is that legal liability (in torts and criminal law) falls only on those who are morally responsible (Moore 2009, p. vii; cf. also Moore 1997).

In other words, [A]ll law, on my view of it, must be based on policy [...]. This policy would be to attach legal liability to morally blameworthy actions. It is morality, not legal policy, that tells us that actions that cause harm are more blameworthy than those that merely attempt or risk such harm. It is metaphysics, not legal policy, that tells us when an action causes a certain harm (Moore 2009, p. 230).

Moore’s argument can be roughly summarized as follows:

Responsibility Regardless of Causation

133

(i) moral responsibility depends (necessarily but partially) on causation, and since (ii.a) criminal liability (and lato sensu criminal law) is based on moral responsibility, and since (ii.b) criminal liability (and lato sensu criminal law) ought to be based on moral responsibility, then (iii) criminal liability substantially depends on causation, and͒ (iv) criminal liability ought to (i.e. it is justified to) depend on causation. I do not think Moore’s reasoning sound, and I reject both conclusion (iii) and (iv), for two key reasons. First, even granting both premises (ii.a) and (ii.b) for argument’s sake,23 I have argued (in §1) that premise (i) is false: responsibility in general is not necessarily based on causation—even if it may be contingently based on causation. Second, Moore’s understanding of causation (in law and morals) as naturalistically justified is fundamentally wrong— as I have argued in §2. To compare HH’s argument, here is Moore’s version: (1) legal responsibility (the criteria for) is justified if it tracks moral responsibility (the criteria for); (2) causation is a necessary condition for moral responsibility;24 (3) the relevant concept of causation here is NOT the concept commonly used by ordinary people in speech. A careful metaphysics theory of causation is needed to justify moral and legal doctrines. For Moore, all responsibilities must be causal. Therefore, he has almost no choice: either his theory is factually disproved, or those noncausal responsibilities (vicarious responsibility, accomplice liability) must not be considered proper cases of responsibility. The latter is precisely his strategy. The following paragraph is devoted to showing how Moore ignores the facts (the reality of accomplice, non-causal liability) to fit his theory (no liability/responsibility without causation). Accomplice Liability The key test for the notion of causation, both in the law and in metaphysics, is to account for accomplice liability. Obviously, an accomplice is a person who, it is presumed, helps or instigates the wrongdoer to commit the crime. However, the extent to which the accomplice’s contribution is a causal contribution is open to question. It is plausible to think this instigation requires both an actus reus and a mens rea to be considered accomplice liability.25 Leaving aside the actus reus requirement, is mens rea limited to knowledge (of possible

134

Chapter Nine

consequences) or could it possibly require purpose (the purpose of helping the future wrongdoer)? And must there be an intention to merely help the future wrongdoer, or an intention to bring about the criminal offence, directly or indirectly via the wrongdoer? Unfortunately, Moore limits his discussion only to the actus reus requirement. Moore seeks to abandon the accomplice liability doctrine. Why? I try to summarize (and, alas, somewhat simplify) his argument here: (i) The attribution of responsibility (here, liability) must be strictly causal;26 (ii) An accomplice has a causal role only in HH’s intervening cause sense; (iii) HH’s idea of intervening causes is metaphysically unfounded and should be abandoned; (iv) The role of an accomplice cannot be causal (with regards to the offence) in any sound metaphysical sense. Therefore: (v) There must be no accomplice liability.27 But accomplice liability is a perfectly accepted form of (at least legal) responsibility, and since there are great difficulties in interpreting accomplice liability in causal terms, it seems advisable to forgo the causal requirement, as I argue in this paper. Causation in law (and morals) In the last part of his book, Moore finally tackles the “beast”: the metaphysical notion of cause (causality). His analysis is, roughly, split into two parts: (a) an analysis of causal relata and (b) an analysis of the causal relation. In other words, accounting for what causation is (“counterfactual dependence, nomic sufficiency, probabilistic dependence, regular concurrence, something else or nothing at all (p. 327)”) is quite different from (yet related to) accounting for the entities causation links (“events, aspects of events, facts, negative events (p. 327)”). (a) Causal relata: a deontic fallacy As for (a): causal relata, we start to notice a crack in his mechanism. He loosely argues for a distinction between metaphysics and the law. In metaphysics, the true causal relata are “fine-grained” things: states of affairs. Instead, the relation most desirable for use in law is different: (coarse-grained) events are the relata on which legal liability should turn, recognizing that such relata will be constructions based on the true relata of the causal relations, which are states of affairs (p. ix, emphasis added).

Responsibility Regardless of Causation

135

Now, let me just put forward two informal objections: First, from the basic descriptive level we creep into the domain of the normative. He said that we need an objective, descriptive account of the metaphysics of causation and causal relata (as fine-grained states of affairs) and then he changes his mind and builds up new causal relata especially for the law, as the above quotation shows. Why? Because this relation is “most desirable for use in law”. Those two different theories seem neither equivalent nor interchangeable. He simply appears to pick the most convenient, regardless of which is true. Second and more generally, he commits a deontic fallacy:28 the fact that something is desirable or should obtain (for instance, responsibility ought to be based on causation) determines our thinking that the world is as it should be (we select the relata so that responsibility is based on causation, and all other non-causal forms of responsibility—such as accomplice liability—do not exist). (b) The nature of causation As for (b): the nature of causation, (see pp. x-xi and part IV) the confusion continues. Moore identifies law’s causal theory as counterfactual dependence.29 For the law, counterfactual dependence (roughly, the thesis that c is a cause of offence y iff y would not have occurred if c had not occurred) is both necessary and sufficient for causation. Moore adopts a twofold (and inconsistent) strategy: first, he provides us with a series of reasons and argument against both the sufficiency and the necessity of counterfactual dependence for causation (citing, among other things, the existence of non-causal counterfactuals, the need to consider omissions as causes, overdetermination, etc.): counterfactual dependence and causation are not the same thing; but second, he confidently argues that counterfactual dependence is a legal and moral “desert-determiner independent of causation (p. 426).” This is to say that blameworthiness (and mutatis mutandis liability) depends not (only) on causation but (also?) on counterfactual dependence. How and when? Moore presents us with counterintuitive cases, such as symmetrical and asymmetrical overdetermination (see Sartorio, 2012) and blameworthiness for omissions and preventions—cases that can hardly be accountable in terms of non-counterfactual causation. There, counterfactual dependence occurs without causation, and Moore wants to assign liability without causation. For this last section on causation and counterfactual dependence, I shall adopt the so-called charity principle in reading Moore’s theory.

136

Chapter Nine

Above, we have seen that Moore’s idea is that causation is necessary for responsibility (liability). Then, we saw how desert is often determinable only by using a criterion of counterfactual dependence; but we have seen how—according to Moore—counterfactual dependence neither is, nor implies, causation. Now, I would prima facie say that there is a non-sequitur in Moore’s argumentation. But on a more charitable reading, we might try to apply the familiar, legal distinction between conviction and sentence. In this way, causation is necessary for conviction (i.e., the attribution of responsibility/liability); counterfactual dependence determines, instead, the (severity of the) sentence—even if I cannot say the extent to which counterfactual dependence would be either sufficient or necessary for the sentence. But in the end, I think Moore’s arguments are simply untenable. Just one example is as follows. We have seen that: (a) He considers causation necessary for responsibility/liability; (b) He repeatedly states that “omissions cannot be causes” (p. 444); (c) He thinks that there must be liability for omissions (pp. 444ff.) And these three premises are simply inconsistent. If (b) holds, than we must discard either (a), so that causation is not necessary for responsibility/liability, or (c), that is, we cannot attribute liability for omissions, since (b), they cannot be causes. I do not think Moore’s reasoning sound and I have rejected both conclusion (iii) (liability necessarily depends on causation) and (iv) (liability must depend on causation). Besides the two main reasons given above (responsibility in general is not necessarily based on causation; and causation is not simply “out there”), in this last section I have argued that Moore’s reasoning is, at the very least, inconsistent, as the case of omissions shows.

4. To sum up I have argued that causation is neither necessary for responsibility, nor convincingly descriptive. I have suggested that, although based on physical, empirical evidence, both law and morals need to draw a line based on evaluative considerations. To sum up, I hope to have disproved the claim that responsibility is ultimately descriptive because it is fundamentally rooted in causality, and causality is descriptive. My strategy was twofold. I put forward two arguments: First, I showed that there are significant and independent noncausal forms of responsibility that cannot be reduced to causal

Responsibility Regardless of Causation

137

responsibility; second, I showed that the very notion of causality (in law and morals) is—lato sensu—normative, or at least non plainly descriptive. Empirical research might tell us where to look to find causes, but it will not indicate which causes are relevant in order to ascribe responsibility—or, in other words, where to draw the line.30

References Arendt, J. 1987, “Collective Responsibility”, in Bernhauer, J. (ed.), Amor mundi, Dordrecht: M. Nijhoff Azzoni, G.M. 2004, “Religioni aziendali”, Sociologia del diritto 31 (2), 181 —. 2012, “Etica e comunicazione della Corporate Social Responsibility”, in Invernizzi E. and S. Romenti (ed.), Relazioni pubbliche e Corporate Communication. La gestione dei servizi specializzati, 2, Milan: McGraw-Hill, 29 Bacon, F. 1630, The Elements of the Common Law of England. Beebee, H. 2013, “Hume’s Two Definitions: The Procedural Interpretation”, Hume Studies 37 (2), 243 Beebee, H., C. Hitchcock and P. Menzies (eds.) 2009, The Oxford Handbook of Causation, Oxford: Oxford University Press Benjamin, M. 1976, “Can Moral Responsibility Be Collective and Nondistributive?”, Social Theory and Practice 4 (1), 93 —. 1998, “Why Blame the Organization? A Pragmatic Analysis of Collective Moral Responsibility”, Teaching Philosophy 21 (2), 201 Bobzien, S. 2006, “Moral Responsibility and Moral Development in Epicurus”, in The Virtuous Life in Greek Ethics, Cambridge: Cambridge University Press Braham, M. and M. van Hees 2012, “An Anatomy of Moral Responsibility”, Mind 121 (483), 601 Caruso, G. forthcoming, Exploring the Illusion of Free Will and Moral Responsibility, Lanham (Maryland): Lexington Books Corlett, J.A. 2001, “Collective Moral Responsibility”, Journal of Social Philosophy 32 (4), 573 Dubbink, W. and J. Smith 2011, “A Political Account of Corporate Moral Responsibility”, Ethical Theory and Moral Practice 14 (2), 223 Duff, A. 2008, “Responsibility and Liability in Criminal Law”, in Kramer M. et al. (ed.) The Legacy of H.L.A. Hart, Oxford: Oxford University Press —. 2009, Answering for Crime, Responsibility and Liability in the Criminal Law, Oxford: Hart Publishing

138

Chapter Nine

Epstein, R.A. 1973, “A Theory of Strict Liability”, The Journal of Legal Studies 2 (1), 151 Faroldi, F.L.G. 2012 “Fallacia deontica. From "ought" to "is"”, Rivista Internazionale di Filosofia del Diritto 89 (3), 413 —. 2013, “Hart e la responsabilità oggettiva” ms. under review. —. 2014, The Normative Structure of Responsibility, London: College Publications. Feinberg, J. 1965, “Action and Responsibility”, in Black, M. (ed.) Philosophy in America, London: Allen & Unwin, 134 —. 1968, “Collective Responsibility”, Journal of Philosophy 65, 674 —. 1970, Doing & Deserving: Essays in the Theory of Responsibility, Princeton: Princeton University Press Feinberg, J. and J. Coleman (Eds.). (2008), Philosophy of Law, 8th edition, Belmont (CA): Wadsworth Publishing Fischer, J.M. 2006, My Way: Essays on Moral Responsibility, Oxford: Oxford University Press —. 2012, Deep Control: A Theory of Moral Responsibility, Oxford: Oxford University Press Frankfurt, H.G. 1969, “Alternate Possibilities and Moral Responsibility”, The Journal of Philosophy 66 (23), 829 French, P.A. 1984, “The Principle of Responsive Adjustment in Corporate Moral Responsibility: The Crash on Mount Erebus”, Journal of Business Ethics 3 (2), 101 Garrett, J.E. 1989, “Unredistributable Corporate Moral Responsibility”, Journal of Business Ethics 8 (7), 535 Gilbert, M. 2006, “Who’s to Blame? Collective Moral Responsibility and its Implications for Group Members”, Midwest Studies in Philosophy 30 (1), 94 González, E. 2002, “Defining a Post-Conventional Corporate Moral Responsibility”, Journal of Business Ethics 39 (1-2), 101 Graham, K. 2001, “The Moral Significance of Collective Entities”, Inquiry 44 (1), 21͒  —. 2006, “Imposing and Embracing Collective Responsibility: Why the Moral Difference?”, Midwest Studies in Philosophy 30 (1), 256 Hart, H.L.A. 1982, Essays on Bentham: Studies in Jurisprudence and Political Theory, Oxford: Clarendon Press —. 2008, Punishment and Responsibility, 2nd ed., Oxford: Oxford University Press Hart, H.L.A. and T. Honoré 1959, Causation in the Law, Oxford: Clarendon Press Honoré, T. 2010, “Causation in the Law”, in The Stanford Encyclopedia of

Responsibility Regardless of Causation

139

Philosophy, ed. by E. N. Zalta, Winter 2010 Hume, D. 1748, Enquiry Concerning Human Understanding, Oxford: Clarendon Press, 1902 Isaacs, T.L. 2006, “Collective Moral Responsibility and Collective Intention”, Midwest Studies in Philosophy 30 (1), 59 —. 2011, Moral Responsibility in Collective Contexts, Oxford: Oxford University Press Kelsen, H. 1939, “The Emergence of the Causal Law”, Erkenntnis 8 (8),69 —. 1943, Society and Nature: A Sociological Inquiry, Chicago: University of Chicago Press͒ —. 1960, Reine Rechtslehre, 2nd ed., Vienna: Franz Deuticke͒ —. 1967, Pure Theory of Law, trans. by Max Knight, Berkeley: University of California Press —. 1973, “Causality and Accounting”, in Weinberger, O. (ed.) Essays in Legal and Moral Philosophy, Dordrecht: Reidel, 154 Mäkelä, P. 2007, “Collective Agents and Moral Responsibility”, Journal of Social Philosophy 38 (3), 456 Miller, S. 2001a, “Collective Moral Responsibility for Omissions”, Business andProfessional Ethics Journal 20 (1), 5 —. 2001b, Social Action: A Teleological Account, Cambridge: Cambridge University Press —. 2006, “Collective Moral Responsibility: An Individualist Account”, Midwest Studies in Philosophy 30 (1), 176 Miller, S. and P. Mäkelä 2005, “The Collectivist Approach to Collective Moral Responsibility”, Metaphilosophy 36 (5), 634͒ Moore, G. 1999, “Corporate Moral Agency: Review and Implications”, Journal of Business Ethics 21 (4), 329 Moore, M.S. 1997, Placing Blame: A General Theory of the Criminal Law, Oxford: Oxford University Press —. 2009, Causation and Responsibility, Oxford: Oxford University Press Nahmias, E.A. et al. 2005, “Surveying Freedom: Folk Intuitions About Free Will and Moral Responsibility”, Philosophical Psychology 18 (5), 561 Nelkin, D.K. 2007, “Do We Have a Coherent Set of Intuitions About Moral Responsibility?”, Midwest Studies in Philosophy 31 (1), 243 Posner, R.A. 1972, “A theory of negligence”, The Journal of Legal Studies 1 (1), 29 —. 1973, “Strict Liability: a Comment”, The Journal of Legal Studies 2 (1), 205 Risser, D.T. 1985, Corporate Collective Responsibility, PhD thesis, Temple University

140

Chapter Nine

—. 2009, “Collective Moral Responsibility”, in Internet Encyclopedia of Philosophy Russell, P. 1995, Freedom and Moral Sentiment: Hume’s Way of Naturalizing Responsibility, Oxford: Oxford University Press —. 2008 “Hume on Free Will”, in The Stanford Encyclopedia of Philosophy, ed. by E. N. Zalta, Fall 2008 Sartorio, C. 2009, “Causation and Ethics”, in Beebee et al. 2009 —. 2012, “Two Wrongs Do Not Make A Right: Responsibility and Overdetermination”, Legal Theory 18, Special Issue 04, 473 Sheehy, P. 2006, “Holding Them Responsible”, Midwest Studies in Philosophy 30 (1), 74 Shoemaker, D. 2009, “Responsibility and Disability”, Metaphilosophy 40 (3-4), 438 Silver, D. 2006, “Collective Responsibility, Corporate Responsibility and Moral Taint”, Midwest Studies in Philosophy 30 (1), 269 Smiley, M. 2011, “Collective Responsibility”, in The Stanford Encyclopedia of Philosophy, ed. by E. N. Zalta, Fall 2011 Smith, J. forthcoming, “A Political Account of Corporate Moral Responsibility”, Ethical Theory and Moral Practice Soares, C. 2003, “Corporate Versus Individual Moral Responsibility”, Journal of Business Ethics 46 (2), 143 Stapleton, J. 2008, “Choosing What We Mean by 'Causation' in the Law”, Missouri Law Review 73 (2), 433 —. 2009, “Causation in the Law”, in Beebee et al. 2009 Tollefsen, D. 2003, “Participant Reactive Attitudes and Collective Responsibility”, Philosophical Explorations 6 (3), 218 Velasquez, M. 2003, “Debunking Corporate Moral Responsibility”, Business Ethics Quarterly 13 (4), 531 Welch, J.R. 1992, “Responsabilidad Colectiva y Reduccionismo”, Pensamiento 48, 49 Williams, G. 2006, “Responsibility”, Internet Encyclopedia of Philosophy, URL = < http://www.iep.utm.edu/responsi/ > Williamson, T. 1994, Vagueness, London: Routledge Wilmot, S. 2001, “Corporate Moral Responsibility: What Can We Infer From Our Understanding of Organisations?”, Journal of Business Ethics 30 (2), 161 Wittgenstein, L.J.J. 2009, Philosophical Investigations/Philosophische Untersuchungen, Hacker P.M.S. and J. Schulte (ed.), 4th ed., London: Wiley-Blackwell Wright, R.W. 2003, “The Grounds and Extent of Legal Responsibility”, San Diego Law Review 40, 1425

Responsibility Regardless of Causation

141

—. 2008, “The Nightmare and the Noble Dream: Hart and Honoré on Causation and Responsibility”, in Kramer M. et al (ed.) The Legacy of H. L. A. Hart: Legal, Political and Moral Philosophy, Oxford: Oxford University Press

Notes 1

For evidence bearing on this point, see my discussion of “aitía”—meaning both “guilt” and “cause”—in Faroldi (2014, p. 4–5). I will not go into the metaphysical discussion on causality here, but will only be concerned with the “zoom-level” of inquiry employed in legal and moral philosophy. 2 Cf. Italian Civil Code, artt. 1325, 1343–45. 3 For similar remarks, see Hart (2008) and Feinberg (1965, 1970). 4 The matter is complicated because we have (at least) three unknowns, so to speak: moral responsibility, criminal responsibility and causation. A general and convincing theory must, of course, fully take into account the relationships between all three unknowns. 5 “Consider the activities we call “games”. For if you look at them, you won’t see something that is common to all, but similarities, affinities, and a whole series of them at that. We see a complicated network of similarities overlapping and crisscrossing. Similarities in the large and in the small” (Wittgenstein 2009, §66). 6 ‘Regardless’ here means that (i) there are no mens rea elements required for the offense to take place; (ii) mental elements are irrelevant for the offence. Of course mens rea and fault are not the same thing, as I pointed out in (Faroldi 2014, p. 92 and passim). 7 For the concept (and consequences) of collective moral responsibility, see for instance Arendt (1987); Feinberg (1968); Smiley (2011); Benjamin (1976, 1998); Bobzien (2006); Braham and van Hees (2012); Caruso (forthcoming); Corlett (2001); Fischer (2006, 2012); Gilbert (2006); Graham (2006); Isaacs (2006, 2011); Mäkelä (2007); Miller (2001a, 2001b, 2006); Miller and Mäkelä (2005); Nahmias et al. (2005); Nelkin (2007); Risser (2009); Sheehy (2006); Shoemaker (2009); Silver (2006); Soares (2003); Tollefsen (2003); Velasquez (2003); Williams (2006). 8 For a systematic analysis and classification of various forms of non-standard responsibility, see Faroldi (2013). 9 Accomplice and corporate responsibility are examples of non-personal responsibility. On the moral significance of corporate responsibility, see Dubbink and Smith (2011); French (1984); Garrett (1989); González (2002); Graham (2001); G. Moore (1999); Risser (1985); Silver (2006); Smith (forthcoming); Soares (2003); Velasquez (2003); Welch (1992); Wilmot (2001). Corporate social responsibility seems to me a very complex and interesting form of responsibility (if it is a case of responsibility stricto sensu, of course), but I will not deal with it in this work; cf. for instance Azzoni (2004, 2012). Accomplice responsibility will be dealt with in §3.3. 10 For how is the concept of game bounded? What still count as a game, and what

142

Chapter Nine

no longer does? Can you say where the boundaries are? No. You can draw some, for there aren’t any drawn yet (Wittgenstein 2009, §68). 11 For both subjective and objective limitations, I shall not be concerned here with a critique of the general, metaphysical notion of “causation”. This section’s title might also read as “Moral and Legal Causation as Normative”. 12 A similar argument is made by Moore (2009). On this, see infra at §3.3. 13 Here, I am using the terms “act”, “action” and “event” in non-technical senses. 14 For the general problem of vagueness, see at least Williamson (1994). 15 The Italian penal code (art. 40.2) states (stipulates, prescribes?) that omissions are (to be considered/count as?) causes: “[...] Non impedire un evento, che si ha l’obbligo giuridico di impedire, equivale a cagionarlo.” 16 I thank Fabio Bacchini for specific discussion on this point. 17 Needless to say, they need not be coincident: “we must bear in mind the many factors which must differentiate moral from legal responsibility in spite of their partial correspondence [...]. [T]he fact that the individuals have a type of [causal] connection with harm which is adequate for moral censure or claims for compensation is only one of the factors which the law must consider” (p. 66). 18 “[D]oing or causing harm constitutes not only the most usual but the primary type of ground for holding persons responsible in [this] sense.” (p. 65, emphasis added). 19 For a parallel reading of HH, cf. Wright (2008, p. 177): “They insisted that the principles of attributable responsibility should be treated as causal rather than noncausal principles. They seem to assume that in order to avoid ad hoc, policydriven determinations of attributable responsibility, the principles of attributable responsibility (beyond the basic natural causation principle) must be ‘causal’ principles”. 20 The classical debate in (at least common) law was dominated by minimalist works (holding that the criteria for the attribution of responsibility are neither objective nor causal) such as Posner (1972, 1973) and by maximalists (holding that there are factual causal criteria for attributing responsibility) such as Epstein (1973). 21 The postulate of legal theory is that legal liability (in torts and criminal law) falls only on those who are morally responsible Moore (2009, p. vii). 22 “The nature of causation—what causation is—is a matter of fact, inviting theoretical speculation”. Causation is “a real relationship in the world”—cf. (Moore 2009), passim. For a position different from both mine and Moore’s, see Stapleton (2008, 2009): for her, the notion of cause in legal settings must be “untainted by normative controversies”. 23 I have shown elsewhere (cf. Faroldi 2014) that premises (ii.a) and (ii.b) are to be rejected. Briefly, I considered an argument from formalization (if criminal liabilities were to be based ultimately (or informally) on moral responsibility, and moral responsibility depends on moral theories that are usually hard to formalize and codify (because they are subject to inter-community negotiation, emotionbased judgments and other peculiar traits), then there would be problems in drawing legal provisions—however indirectly—from moral theories); an argument

Responsibility Regardless of Causation

143

from disagreement (on which moral theory should our concept of criminal responsibility be based?); and an argument from the judges’ discretionality (if there was no agreed theory of moral responsibility, criminal legal systems could hardly be based upon a varying and inconsistent set of norms. The attribution of criminal liability would depend eventually on the moral system endorsed by the judge). 24 “[A]bsence of causation eliminates responsibility (by licensing consequentialist justifications), rather than merely reducing it (when justifications are not in issue)” (p. 77). 25 As many criminal statuses do—cf. Model Penal Code. 26 Vide supra for my account of Moore’s justification of this argument. 27 Please note that I cannot expand on Moore’s follow-ups on the topic here. 28 On which, see Faroldi (2012). A deontic fallacy is deriving in some way an “is” from an “ought”, for instance by acknowledging the reality of something not as it is, but how it should be. 29 I cannot go into details about counterfactual dependence here. In sum, this view descends from this much quoted passage in Hume (1748, p. 87): First definition: “We may define a cause to be an object followed by another, and where all the objects, similar to the first, are followed by objects similar to the second.” Second definition: “Or, in other words where, if the first object had not been, the second never had existed” (emphasis added for the definition of counterfactual dependence). In this quoted passage, Hume equated a regularist and a counterfactual (in italics) view of causation: these views are not extensionally equivalent. This equation is problematic, as noted, for instance, by Beebee (2013). For a general interpretation of causality and responsibility in Hume, see Russell (1995, 2008). Counterfactual dependence has recently been widely questioned in philosophy in general and ethics in particular, following Frankfurt (1969), who argued for a non-counterfactual-based attribution of responsibility. 30 For helpful discussion and comments on previous drafts of this work, I thank Fabio Bacchini, Amedeo G. Conte, Guglielmo Feis, Sergio Filippo Magni and a reviewer. Part of this work was done during a research stay at Trinity College, Dublin, on an Erasmus EU grant.

CHAPTER TEN SKEPTICISM AND CONTROL SOFIA BONICALZI

1. Introduction In Philosophical Explanations, Robert Nozick pointed out that the problem of free will is interesting in itself—simply because we all want to perceive ourselves as free agents—independently of its relationship with moral responsibility (1981, p. 291). However, one of the reasons why the debate about the existence of free will is so intriguing is exactly its connection with the problem of moral responsibility. It is almost impossible to discuss the issue of moral responsibility without touching on the question of the existence of free will. Among the various definitions of moral responsibility that have been formulated (see Fischer & Tognazzini 2011), the most problematic is also the one that seems to concern us more closely, being inseparably paired with the problem of human freedom. According to Derk Pereboom, the sense of moral responsibility that is at stake with relation to the problem of free will is the idea that we deserve to be blamed for our bad actions while we deserve to be praised for our good actions. Indeed, since the essential condition for moral blame (or praise) is that the offender (or the benefactor) acted under a free choice (1995, 2001, 2013), only the existence of this form of desert—which is “basic” in the sense “that the agent, to be morally responsible, would deserve the blame or credit just by virtue of having performed the action, and not, for example, by way of consequentialist considerations” (2001, p. xx)—could support the idea that feelings such as resentment or anger might be justified and, maybe, promoted (Zimmerman 1988, McKenna 2012). Despite being inclined to avoid an asymmetrical treatment of the criteria for blame and praise, for the sake of simplicity my focus would be primarily on the plausibility of traditional moral blame. The possibility of justifying moral blame seems especially challenging for those who, though

Skepticism and Control

145

accepting the concept of “desert” as a plausible one, don’t find traditional retributivism appealing. Before considering the core issues, let’s start with some coarse-grained distinctions. Believing that an individual deserves moral blame for a bad action does not necessarily coincide with experiencing angry feelings towards her. We could think that some individuals who violated a rule are blameworthy, at the same time experiencing a kind of sympathy towards them. Considering one as an apt target for moral blame is not just a matter of feelings and emotions, but implies a rational judgment about her past behaviour. Obviously, anger and resentment might also represent immediate and unreflective reactions towards one’s bad action but, when we are in the position of justifying these feelings, we usually appeal to specific and more structured moral reasons and judgments. Even though we could not be able to avoid feelings of anger and disappointment towards the offender (Nichols 2007), it is still the case that we have to consider whether the corresponding judgment is really consistent. In moral theories inspired by utilitarian principles, judgments of moral blame might be justified in virtue of their practical consequences. Nevertheless, we often formulate judgments of moral blame that cannot be understood or justified in the light of considerations of this sort: for example, one might blame an individual who violates a promise made to a dead man, considering her objectionable for the violation itself. In this case, the bond that generated that promise might no longer exists, and the trespasser might be sure, for some reasons, that such violation would not produce any real consequence or that no one (apart from her) could be aware of the infringement. What generates the judgment is rather the consideration of the connection between a choice with certain characteristics and a specific action. Which kind of traits should be considered necessary in order to formulate a legitimate judgment of moral blame? The acknowledgment of an action resulted in damages is not enough to legitimate a judgment of this kind: it might be possible that the causal responsible is not also morally responsible for the action performed. If the cat drops a crystal ball and the ball breaks, the cat is to be considered causally responsible but not morally responsible for the damage (see Fischer & Ravizza 1998, p. 1). The same would seem to be true for young children or for those who, for some specific characteristics, are considered as unsuitable targets of moral objections. The exemption from moral responsibility attributions may be permanent or only temporary, as in the case of the individual who acts under the influence of chemical substances capable of causing an alteration of waking consciousness: although we

146

Chapter Ten

might consider this agent responsible for having taken these substances, we could hardly consider her morally responsible for the specific action performed (excluding cases of derivative responsibility). In some situations, we could also separate the causal responsible from the moral responsible of the very same action. If Bryan deceives Mark, persuading him to perform the apparently (morally) neutral action p without revealing him the real negative effect of doing p intuitively, while Mark is causally responsible for p, Bryan is the one morally responsible. We will probably admit the same if Mark performs violations under the influence of a form of compulsion, such as a serious threat to his safety or a physical constraint. In this case moral responsibility is transferred, as a whole or at least in part, to the individual—if there is one—who exercises the compulsion. Consider an adult person in normal conditions, not subjected to evident constraints and quite aware of the possible consequences of her action. What are the conditions that justify us in considering her morally responsible, in the sense of desert, for her own acts? What is missing— more or less irremediably—in the previous cases of exemption is the opportunity to consider the agent as the real author of the action, in a way that allows her to exercise an effective control over it, regardless of her being also causally responsible of it. This is not a definition of what moral responsibility is, but something as a minimal requirement for considering people as suitable targets of considerations of moral blame. If we accept this minimal requirement as necessary for having moral responsibility in the desert sense, we can move to a different plan, wondering whether it is legitimate to attribute such a capacity to normal human beings, and if we are in the position to formulate moral judgments that imply that agents have the ability to perform actions over which they can exercise an effective form of control. The problem I want to consider does not concern the plausibility of some sort of moral responsibility, but the specific form of moral responsibility that is at stake when we are trying to define the problem of control. The question is related to what should happen in the physical world to make such a condition possible so that the minimal requirements could be satisfied. And it is at this point that the question of free will comes back into play and that the possible solutions diverge.

2. Some traditional libertarian and compatibilist solutions According to the libertarian program, the truth of moral responsibility in the desert sense is guaranteed by the existence of a form of free will that

Skepticism and Control

147

is possible only in an indeterministic context. Given the absence— necessary but not sufficient—of the conditions that would exclude or limit by themselves the possibility of moral responsibility attributions, something else is required. To be responsible for an action in the desert sense, the agent must be in the position of making a real free choice, usually selecting among genuinely available alternatives. The final choice must be in some way autonomous—or independent—from previous factors and antecedents, in a way that permits us to identify at least one fragment of true indeterminism in the decision-making process. We could say that our actions are really up to us because the decision-making process is in some way distinct from the mechanism of cause and effect which is supposed to regulate the physical world (general and local determinism) (van Inwagen 1983, Clarke 1997, Kane 2002a). Compatibilists disagree (Smart 1961, Bok 1998, Fischer & Ravizza 1998, Sartorio 2013), stating that indeterministic freedom is not a necessary ingredient to save moral responsibility, which can be grounded in utilitarian considerations or in a different kind of desert-based view. In particular, giving up the notion of “desert”, there is still a utilitarian sense in which moral responsibility attributions could be important, even though determinism is true: if determinism obtains, moral responsibility attributions can be justified for social purposes (Dennett 1984). As mentioned earlier in the chapter, this sense of moral responsibility does not seem to be really at stake, since it is quite independent from the truth of determinism or indeterminism. Indeed, the notion of “utilitarian responsibility” does not seem to capture all the intuitions about what moral responsibility is. The main reason why compatibilists reject libertarian views is that their solutions seem to represent an implausible departure— an “obscure and panicky metaphysics” (1962, p. 25), to use P. F. Strawson’s famous words—from the outcomes of our best scientific theories. However, even if determinism were true, this does not mean that one must resign to the complete lack of freedom and responsibility. Compatibilism proposes that the absence of certain excusing conditions is not only the necessary but also the sufficient condition for freedom—often interpreted not as the opportunity to choose between genuinely available alternatives, but as the faculty to act according to the determinations of the will, no matter what is their origin (e.g. Frankfurt 1988, Watson 1975, 2004). The difference between a kleptomaniac and a thief is defined on this basis: while the kleptomaniac acts under the influence of forces that she lacks the power to control (and therefore she could not be considered morally responsible for the crime), the thief acts in perfect agreement with

148

Chapter Ten

her own will (regardless of the determining factors) and thus can be considered morally responsible for the theft. Traditional libertarian and compatibilist views are the victims of two opposing arguments (van Inwagen 1983, Russell 1995, Haji 1998, Nelkin 2001, Finch 2013). Against libertarian incompatibilism, the Mind Argument states that if our choices are not derived from a cause, then they occur simply by chance. And, if our choices happen by chance, then we could not be considered morally responsible for the actions we perform. The argument has different versions, but synthesizes an intuition we already find in the Stoics and in David Hume: if the choice is not determined by anything, it is simply a matter of luck. The most famous anti-compatibilist argument, the Consequence Argument, tries to show that determinism is not a good alternative ally to build a desert-based view: if all our actions are determined by the conjunction of the laws of nature and the events of the past (both assumed as unchangeable), then our actions are not dependent on us, and therefore we are not responsible. So, if determinism obtains, then we would not be responsible for our actions. Also in this case, even though the formalized versions of the argument have been discussed in relation to the validity of its premises (Fischer 1986), it crystallizes an older and plausible intuition. If none of our actions depend on us in an effective way, then we could not be responsible for them. If the thief, like the kleptomaniac, has no power over her own will, how might she be said to be responsible for her actions? The juxtaposition of these two arguments is the so-called dilemma of determinism. According to these two arguments, there are no cases in which human beings would be suitable targets for responsibility attributions in the desert sense, since in both cases what is missing is the opportunity to exercise a significant form of control over our conduct. If we accept that the legitimacy of responsibility attributions implies that there must be a sense in which the actions performed by an agent depend on her, the problem of how the agent can control her actions is decisive. There are several different ways in which libertarianism and compatibilism try to face the issue.

3. A libertarian proposal. Ultimate responsibility One of the main problems for libertarianism is to show how it is possible to develop an indeterministic conception of human agency that could escape the luck objection. Some libertarian theories (Agent-Causal Theories) articulate a causal presentation of the mechanism of choice,

Skepticism and Control

149

conceived as an indeterministic process. Human beings would possess a specific causal power, not reducible to the mechanical causation that rules the physical world. This specific agential power is not conceived as something situated completely outside of the net of causal antecedent factors, but it is an additional factor, which could be decisive for the occurrence or the non-occurrence of an event. The agent has the (intrinsic) ability to originate new causal chains, producing the occurrence of an event in an indeterministic way (Clarke 2003, O’Connor 2008). Again, the alleged obscurities of such a conception of human agency—which seems to involve the reference to strange entities not included in the ontological horizon defined by the scientific view of the world—represents the main obstacle for the viability of this position. Robert Kane proposes a different solution (1985, 2002a, 2002b), according to which the relationship between free choice and moral responsibility is captured by the concept of ultimate responsibility: The idea is this: to be ultimately responsible for an action, an agent must be responsible for anything that is a sufficient reason (condition, cause, or motive) for the occurrence of the action (2002b, p. 407).

According to this model, individual choices can be explained by reference to the character, motives, or previous experience of the agent who, for being (ultimately) responsible, must be (at least partially) responsible for having this character and these motives, by virtue of past choices and actions. This approach has the advantage to shift the attention from the principle of alternative possibilities—which is in any case implied—to the actual sequence, focusing on the decision-making process. Nevertheless, if we are not responsible for our single choices, but for having formed the character that produces these choices, how could we be responsible for the choices that built our character? Kane stops the regress to the so-called self-forming actions (SFAS): the original, self-determined deeds from which our character is built and where that fragment of indeterminism, which seems necessary for libertarian free will and moral responsibility, is situated. The whole set of our actions is not indeterministically produced, but only the choices related to those moments in life when we oscillate between conflicting perspectives and motivations. It is in the very moment of the decision that a complex of motivations prevails over the other one, after an indeterministic phase preceding the deliberation. Then, if one is able to perform the actions she chose, she can be considered as an apt target for moral responsibility attributions in terms of praise and blame. If compatibilism requires only a form of antecedent determining control,

150

Chapter Ten what libertarians must require for undetermined SFAs, I believe, is another kind of control altogether—(that compatibilist cannot obtain)—namely, ultimate control—the originative control exercised by agents when it is “up to them” which of a set of possible choices or actions will now occur, and up to no one and nothing else over which the agents themselves do not also have control (Kane 2002b, p. 431).

This type of control permits ultimate responsibility (UR) and implies indeterminism, introducing a precisely indeterministic stage within the decision-making process, which would save the role of subjects as creators of their own ends and purposes. Nevertheless, as Daniel Dennett (2003) pointed out, if the thesis is consistent it is necessary not only that a number of alternative possibilities are present in the interval preceding the decision, as Kane stated, but also (and crucially) in the instant in which we “push the button”. However, in Kane’s account, the decision seems to come from nothing or, more simply, is the result of the motivations that were dominant at that moment. As observed by Clarke, this model could offer a metaphysically adequate account of what free will can be, but it is not adequate to solve the problem of control: An event-causal libertarian view secures ultimate control, which no compatibilist account provides. But the secured ultimacy is wholly negative: it is just (on a centered view) a matter of the absence of any determining cause of a directly free action. The active control that is exercised on such a view is just the same as that exercised on an eventcausal compatibilist account (Clarke 2003, p. 71).

Then, one is likely to fall in the labyrinth of justifying how our choices are not simply the result of chance. Eventually, it seems that, in a libertarian perspective, the only possible way to bypass the problem of luck rests, again, on the problematic idea of the agential freedom.

4. A (semi)compatibilist proposal. Guidance control The main problem for a compatibilist who wants to avoid utilitarian solutions and embrace a desert-based view is the need to show that, even in a deterministic context, there is a sense in which our actions are significantly up to us, so that we can be suitable targets for responsibility attributions involving a reference to desert. Accepting the validity of Frankfurt cases—according to which the presence of alternative possibilities is not a necessary condition of moral responsibility—is not enough to save moral responsibility in the desert sense (1969). Indeed,

Skepticism and Control

151

Frankfurt cases show that the absence of alternative possibilities is not a sufficient condition for eliminating moral responsibility, but don’t provide a positive account of which are the necessary conditions for it. Once we accepted the (controversial) claim that we do not need alternative responsibilities in order to be morally responsible, there is still a lot to do. According to Frankfurt’s results, the attribution of responsibility does not depend on the actual opportunity to make alternative choices, but on the structure of the actual sequence of action. If the outcome does not derive from the presence of a coercive force, it is still possible to hypothesize the existence of a form of moral responsibility. Indeed, the presence or the absence of alternative possibilities can be a minor factor also in some cases where, intuitively, moral responsibility is obviously diminished by the existence of excusing condition. It seems that even where the coercion is clear and the victim is perfectly aware of her unfortunate condition, something identifiable with freedom of choice may potentially survive. A blackmailed father, who is asked to pay a very large sum of money for the release of his daughter, might choose to pay, commenting: “I could not do otherwise”. In this case, even though the choice derives from a form of constraint (a serious menace to his daughter’s life), he could possibly decide not to pay, in spite of the seriousness of the consequences. The reason why his responsibility is diminished is not that he does not have alternative possibilities, but it depends on the way (coercion) in which the specific sequence of action occurs. In other cases, there might be self-imposed constraints that limit one’s possibility to do otherwise, without necessarily threatening moral responsibility. For instance, Dennett presents the case of Martin Luther, who was not trying to deny the responsibility for what he did when, defying the Roman Church, said that, according to the imperatives of his conscience, he could not have done otherwise. We can be free and responsible even if the action we perform can be seen as the only possible one: Those who hold dear the principle of “could have done otherwise” are always insisting that we should look at whether one could have done otherwise exactly in the same circumstances. I claim something stronger; I claim that I could not do otherwise even in any roughly similar case. I would never agree to torture an innocent person for a thousand dollars (Dennett 1984, p. 133).

Therefore, what is really crucial is not the actual presence of alternative possibilities but, once again, the possibility that the action is up to us: the

152

Chapter Ten

decisive challenge is related to the ability to exercise a form of control over our actions. How can compatibilism respond to the problem of control? Accepting the validity of Frankfurt cases, John Martin Fischer formulated one of the most intriguing proposals—according to which determinism might be compatible with moral responsibility and not with free will—for addressing the issue of control (1995, 2011). In Fischer’s example, a driver is considered in two different situations: (1) in the first case, if he decides to turn right as he actually does, he achieves the goal. Had he decided to turn left, he would have normally turned left. In this case the driver possesses two different types of control over the vehicle, named guidance and regulative control. The first term indicates that he is able to do what he currently does, while the latter describes his power to act in a different way; (2) in the second case, something breaks down in the guidance system of his car while he is driving, and he is completely unaware of it. If he turns right everything works in the normal manner but, if he turns left, the car would move to the right. In this case the driver would lose the second kind of control, but not the first, the one that, in agreement with Frankfurt’s results, is useful for the determination of moral responsibility (1995, p. 131). Focusing on the level of current events, according to Fischer, it is possible to formulate a notion of “control”—and then of “moral responsibility”—which involves only the ability of the agent to “do what he is doing”, while the actual opportunity to make alternative choices, understood as a different kind of power agency (regulative control), is not crucial for defining the moral responsibility we need. My claim is that guidance control is the freedom-relevant or control component of moral responsibility. Thus, an agent can legitimately be held morally responsible for his behavior, even though he lacks regulative control (or freedom to choose and do otherwise) (2011, p. 6).

What we need in order to be morally responsible in the desert sense is not a complete libertarian power of control over our action but, more modestly, the ability to do what we are doing. Fischer proposes a subjectivist representation of moral responsibility, according to which being morally responsible partly rests on recognizing ourselves as legitimate targets of responsibility attributions, being also able to respond, through our behaviour, to the reasons (among which we find the moral reasons) that are in place in the context of choice. Secondly, it rests on being the holders of the mechanism of choice from which the action flows (we are not the owners of the mechanism in case of coercion).

Skepticism and Control

153

Nevertheless, one might object that guidance control is not enough for structuring a desert-based theory of moral responsibility. Indeed, sympathizing with Fischer’s proposal seems to imply an a priori adherence to compatibilism. The idea of “being owners of the mechanism” would seem to have important different implications depending on whether we accept a compatibilist or a libertarian point of view: in the first case the “property” refers to the absence of external coercion, while in the second it translates the idea that it is the will that is autonomous. Only in the compatibilist perspective it is possible to speak of freedom or moral responsibility distinguishing merely between (not free) actions imposed through coercive mechanisms and (free) actions dictated by a desire generated by the concatenation of preceding causes. The idea that the agent is responsible because she is the owner of the mechanism of choice from which the action flows (independently of which determining factors produced the choice)—which seems a powerful intuition in a compatibilist perspective—could not satisfy the libertarian requirement of the autonomy of the will in an important sense. With his Four-case Argument, Pereboom has effectively argued, through a generalization strategy, that there are no real differences between manipulative scenarios in which some external circumstances directly or indirectly determine Plum to commit a murder (case 1 and 2), and scenarios in which physical determinism obtains and Plum—an often rationally egoistic man—decides “autonomously” to kill White as a consequence of some rigorous training practices (case 3), or merely in virtue of the character he possesses (case 4). In all these cases, Plum doesn't have the ability to control his conduct and, so, the opportunity of blaming him for his bad action seems again to lose its consistency (1995, 2001, 2014).

5. The skeptical outcome In a different way, the forms of libertarianism and compatibilism examined try to defend a vision of moral responsibility that incorporates an explanation of why praise and blame, when reasonably formulated, are deserved. The reason why defending an adequate theory of desert is seen as so important is of a moral kind. Responsibility attributions play an important role in our interpersonal life and in the survival of the moral community. The presumption is that giving up the idea that moral praise and blame are deserved would lead to the bankruptcy of the moral structure. Treating people as apt targets of moral responsibility attributions is, in a basic sense, considering them as praiseworthy and blameworthy. This does not mean that we could not have number of different

154

Chapter Ten

conceptions of moral responsibility, the majority of which do not have anything to do with the problem of metaphysical freedom. For instance, relational theories of moral responsibility, such those proposed by R. Jay Wallace (1994) or Thomas Scanlon (1998, 2008),—where moral responsibility more or less coincides with the correct attribution of moral responsibility—do not lose their force if the physical world turns out to be deterministic. As already mentioned, also utilitarian conceptions of moral responsibility would survive, but the dispute between libertarianism and compatibilism about the problem of desert cannot be adequately solved on that basis. In general terms, we might conceive ourselves as responsible beings because, by virtue of the fact that we are inhabitants of the world in constant relationship with other individuals, we are part of an implicit community that we consider important to preserve. In this case, moral responsibility is connected with the ability to do what is reasonable to do in relation to certain constraints. A reasonable and well-educated moral agent will respect some general or more specific rules of conduct as a part of her duty towards the society. She might perceive herself as a responsible agent in virtue of the fact that she will consider in her power— given her general ability to understand some general principles of behaviour and to grasp moral reasons—to accommodate her conduct to some implicit or explicit rules. In a similar way, individuals occupying certain positions might be held morally responsible in relation to duties imposed by their specific role. Thus, for example, the CEO of a multinational corporation may be rightly considered responsible for the fraud committed by the firm she leads. In this case, moral responsibility can be interpreted as the idea that rational people who take on some commitment have to maintain certain standards of behaviour. As for the cheating CEO, personal responsibility comes from having violated specific rules of conduct and might be connected with a form of legal responsibility. Also the practice of legal punishment—though questionable from other points of view—is not really threatened by the idea of a lack of control, since it can be justified in virtue of its role in the maintenance of a wellordered society. Being part of a community where moral norms are respected is a decisive part of personal and collective flourishing. Consistent moral responsibility attributions play the part of a connective glue that structures our moral practices, no matter the specific conception of morality we endorse. However, if we affirm that the CEO has not only violated the constraints imposed by her specific position but, because of

Skepticism and Control

155

her act, is also blameworthy, then the connection between free will and moral responsibility in the desert sense seems inevitable. One might object that skepticism about basic desert would cover some intuitive distinctions that our legal systems are able to incorporate, those between (blameworthy) people who, in normal conditions, commit a crime and (not blameworthy) people who act badly, but are victims of—for instance—a form of compulsion. The distinction seems intuitive and fair, but I think there would be other possible ways to justify our treatment of criminal behaviours avoiding a direct reference to basic desert. If an individual commits a crime under the influence of a nervous disease, curing (where possible) her mental illness seems the best way to help her in restoring her damaged relationship with the moral community. The socalled excusing conditions have an important role in the definition of the kind of rehabilitation that is more adequate. In the case of an agent who, without any internal or external coercion, displays a criminal behaviour, punishment can play a double function. If applied reasonably, it could have the same rehabilitative role that, in other cases, is the proper function of medical care. That is the reason why punishment should be reasonably commensurate to the crime and tend to rehabilitative ends. In extreme cases, where the success of rehabilitative process seems very unlikely, punishment can maintain its utilitarian role by helping the majority of people to preserve themselves from the dangerous conduct of a single one (Pereboom 2001). If compatibilism derives the idea of desert from a theoretical construction that seems too ad hoc, while libertarianism is accused of violating the outcomes of physical and biological sciences, the foundation of moral responsibility on the roots of the concept of desert seems precarious. Giving up the idea of justified desert, there might be still room for discussing about moral responsibility, and what seems to remain is a form of justified skepticism.

6. Managing skepticism In the history of the debate, there have been some great challenges to the idea that the controversy about free will and moral responsibility could be solved from the theoretical point of view. In his powerful essay, Freedom and Resentment (1962), P. F. Strawson proposed the irrelevance of the question of determinism with respect to the problem of proper attributions of moral responsibility. Regardless of the structure of the physical world, we are naturally prone to adopt reactive attitudes towards other people, so that we could not avoid responsibility attributions as

156

Chapter Ten

natural accompaniments of an interpersonal relationship. The difficulties in assuming an objective attitude (which “cannot include the range of reactive feelings and attitude which belong to involvement or participation with others in inter-personal human relationship” (1962, p. 9)) in normal relationships should not be interpreted as the reason for the falsity of determinism, but it rather reveals the practical impossibility of abandoning an established pattern, since it is not in our power to avoid feelings such as anger, resentment or gratitude, as they rationally rest on the role that they play in our lives. Also without considering the criticism to the Strawsonian approach—how is it possible to distinguish between the appropriate reactive feelings and the inappropriate ones if they are not anticipated by a rational judgment? (See Fischer & Ravizza, 1993)—why, if determinism were true, not try to modify our reactions towards other people? Even though we are not able to refrain from experiencing impulsive reactions towards other people’s behaviour, we could commit ourselves to modifying our less immediate attitudes, maybe in the sense of a forwardlooking attitude towards the people we usually consider as blameworthy. One might object that, if determinism obtains, no one is able to decide to modify her conduct. Nevertheless, even if determinism obtained, moral attitudes and feelings—together with some other factors—could contribute to orient our behaviour. As stated by Pereboom (2001, 2002, 2013, 2014), to accept skepticism as a coherent solution, one should be prepared to assume a revisionist attitude, questioning the legitimacy of her reactive attitudes as originating from the idea that human beings really deserve traditional blame and praise. At the same time, there are still reasons to behave in a good way since good deeds do not lose their inherent value only because their authors are not praiseworthy or blameworthy in the desert sense (and, at the same time, bad actions do not became bad only when we judge their authors as blameworthy). Against claims like these, Ishtiyaque Haji argues that determinism subverts deontic notions of rightness and wrongness. If the OIC thesis (“ought implies can”) is valid—and if it is valid that “ought” implies “could have refrained”—, a world without free will (freedom to do otherwise) is a world with no obligations and, consequently, no moral duty and moral right: if I cannot do x, I have not a moral obligation to do x and failing to do x seems not to be wrong for me (Haji 2002). Thus, in a deterministic world, even if there were something inherently good or evil, no one could be coherently pushed to pursue it. The truth of determinism would undermine the roots of morality and destroy the rationality of reactive attitudes and interpersonal feelings, which presuppose:

Skepticism and Control

157

correct deontic moral judgments; they presuppose, for example, that certain acts are indeed right, or wrong, or obligatory, and not just intrinsically or overall good or bad. But then as determinism undermines the truth of such judgments, it will also undermine the rational grounds for these very practices or attitudes (Haji 2003, p. 258).

The suggestion is that, if determinism is true, we are forced to recognize the global irrationality of reactive attitudes linked to the correctness of deontic judgments. The price might be extremely high: If the very stance of holding responsible, though, is to be understood in terms of susceptibility to the moral sentiments such as guilt, indignation, and resentment, determinism would call into question the rationality of this stance because it imperils the rational grounds for such moral sentiments. Moreover, insofar as the reactive emotions underpin valuable social practices in our lives, the rationality of these practices themselves would be imperiled by determinism (Haji 2003, p. 249).

Adopting a revisionist attitude towards reactive attitudes seems not to be enough since it is the same structure of deontic morality that is at stake. While—accepting Frankfurt’s theses—moral responsibility might not need the possibility to do otherwise, deontic morality has different control requirements: “If one ought to do something from the perspective of objective reasons, then one can do it; the ‘ought’ of objective reasons implies ‘can’” (Haji 2012, p. 19). Moreover Haji claims that determinism undermines the bond between obligations and objective pro tanto reasons (reasons which can be outweighed by other reasons, not “reasons allthings considered”): if we have a moral obligation to do x, then we have a pro tanto reason (not outweighed by other reasons) to do x, but “no one can have such a reason to do something unless one could have done otherwise” (2012, p. 235). Obligations are connected to (pro tanto objective) reasons, which in turn require having alternatives: If you have an objective overall moral obligation to do something, then you have a pro tanto reason to do that thing. Moreover, if you have a pro tanto reason to do something, then you could have done otherwise. Thus, if you have an objective overall moral obligation to do something, then you could have done otherwise—you had a pertinent alternative (Haji 2011, p. 303).

The menace seems insidious both for compatibilism (or semicompatibilism) and skepticism, as it states the incompatibility between determinism and deontic morality. Without entering into the

158

Chapter Ten

details of such a complex debate (for replies and discussions on the topic, see Fischer 2003, Pereboom 2001, 2014 p. 138, Copp 2008, Vranas 2007), in the remaining part of the chapter I would briefly consider how it might be possible to resist—at least partially—the threat. A first point is that, if we accept that determinism is incompatible with moral responsibility in a desert sense, then the fact that determinism might be incompatible also with deontic morality does not constitute, by itself, a proof of the falsity of determinism. We can start by asking if such agnosticism is a coherent outcome and, then, we have to deal with the problem of how our morality can survive this possibility. A large number of studies about the functioning of brain mechanisms, following Libet’s pioneering experiment about the choice-making process (Libet 2002. For a new version of the experiment, see Soon, Brass, Heinze & Haynes 2008. For criticisms about the interpretation of Libet’s experiments, see Mele 2006, De Caro 2009, 2013), have often been interpreted in the direction of a progressive reduction of our ability to exercise a conscious control over our actions. However, this line of research—despite its potential contribution to the analysis of specific criminal cases—could hardly give an answer to our questions about the existence of something called free will, or directly affect our view of moral responsibility attributions (Roskies 2006). Free will, or the mechanism that regulates our agency, is not something that could be simply discovered or revealed, even though it is likely that, in the next future, we will know more about the mechanisms that regulate our decision-making processes. Free will needs not to be conceived as an unsolvable metaphysical problem, as claimed by Colin McGinn (1993), neither moral responsibility has to be seen as something incompatible with both the truth of determinism and with the truth of indeterminism, as stated by Galen Strawson (1986). Pereboom (2001) suggests that remaining agnostic about free will can be rather seen as a provisional empirical standpoint, given the credibility of our best scientific theories but in absence of a real disproof of agential theories. If human beings were agents capable of causing events in a partly autonomous way, there would be room for the attribution of traditional blame and praise in the desert sense, but what prevents us from embracing this hypothesis is the relative unlikelihood that we can enjoy this special status. The skepticism or agnosticism that many philosophers currently advocate could be considered as the consistent outcome of the lack of solutions for such a complex problem. What about morality? While Saul Smilansky (2000) stated that the dissolution of the illusion of free will would be dangerous for the morality

Skepticism and Control

159

system, Pereboom (2001, 2013, 2014) effectively defends a revisionist forward-looking conception of moral judgments that, rejecting the idea of basic desert, promises to survive the collapse. Ought-judgments are often formulated in order to express requirements that fall outside the domain of deontic morality strictly considered, as in the case of prudential suggestions (“you ought to be careful when you work in the kitchen”), or do not imply something that can be actually realized in the agent’s lifetime (“people ought to stop using violence”). In some cases we use ought-judgments also for giving suggestions about how one should modify her attitudes towards something or someone, even though we know that that particular attitude does not depend on something that the agent can effectively decide to do (“you ought to stop loving that man”). In these circumstances the link between “ought” and “can” seems particularly weak and yet, at the same time, the correspondent oughtjudgment is not incoherent. In the case of a moral conversation between two individuals, oughtjudgments are usually motivated by a double task: (1) overtly expressing a plausible moral opinion and (2) affecting the agent’s behaviour. Are these two functions impaired if determinism obtains? Consider (2): when we formulate a moral judgment about one’s bad deed, we overtly express the desire that, in similar future circumstances, the agent would try to modify her conduct or, more generally, we think that the world would be a better place if the agent would behave better. In normal situations, moral admonishments usually affect people’s conduct, being substantial factors among a set of determining conditions. While fatalism claims that everything is fixed since the beginning of time in an unchangeable way— so that, whatever action we perform, we could not do anything to change the course of the events, which is completely fixed—determinism, as I understand it, is a more limited thesis, a principle of regularity according to which every fact is the result of a set of previous determining causes (Björnsson and Pereboom 2014). If determinism is not confused with fatalism, there are no plausible reasons to exclude moral judgments from the range of determining conditions. Even if we are not free agents, there seem to be still opportunities to save the value of morality but, while (2) depends on considerations about the chance to affect future behaviour, (1) seems to be important by itself, independently of the actual opportunity to change people’s conducts. This is the role of deontic moral considerations that seems more in danger if determinism obtains. Nevertheless, apart from cases of unreflective reactions, (1) is justified only in those cases where we consider the agent as an apt target of moral admonishments, given her general ability to grasp

160

Chapter Ten

moral reasons and, eventually, to change her behaviour according to moral concerns. The actions of animals and children are not usually subjected to deontic considerations of any sort, just because we realize that they could not be responsive to moral considerations. Someone might object that when we are judging that someone ought to do something, what we want to say is not just that, in a future similar circumstance, she should behave in a different way but that, in that particular occasion she ought to have behaved in a different way because of a specific moral obligation. In some way the old requisite of alternative possibilities seems to play a substantial role in our intuitions about deontic morality. If Laura steals James’s wallet and she could not have done otherwise, saying that she had a moral obligation not to steal the wallet seems unfair. I want to suggest that, while a desert-based compatibilism has to deal with this puzzle, skepticism about traditional moral blame could be a better candidate to bypass the problem. Saying that Laura ought not to steal James’s wallet (and then, that she ought to do otherwise) implies only that, due to her general ability to grasp moral reasons, she was in the right position (before acting) to understand that stealing a wallet is a bad deed—I prefer to leave open the question about the presence of a real moral obligation not to steal the wallet or, better, about the compatibility between the skeptical perspective and the existence of real moral obligations, but I am inclined to say that the presence of objective moral reasons not to steal the wallet might be interpreted as a moral obligation not to steal the wallet or, maybe, that, to appreciate the force of moral reasons, we do not necessarily need to understand them in terms of moral obligations—, even though, from a skeptical point of view, blaming her just because she acted in that way remains unfair. From this perspective, also in a case where there is no actual opportunity to influence the agent’s behaviour, the ought-judgment would not lose its consistency. Consider the expression: “Nero ought to have been gentler with Christians”. If we consider emperor Nero as an insane person, totally unable to grasp moral reasons, the expression simply means that, from a general perspective, it would have been better if he was gentler with Christians, even if we realize that, for his particular mental condition, he was not able to behave in a different way. In this case his incapacity to behave in a different way turns out to be an excuse for his bad behaviour, but does not undermine the idea that the same act of persecuting Christians was bad from a moral point view. If, instead, we are prone to judge Nero as a cruel man with a normal capacity to grasp moral reasons, the expression “Nero ought to have been gentler with Christians” implies both that it would have been better if he were gentler with

Skepticism and Control

161

Christians, and that it was legitimate to pretend a different behaviour from him, just because he was supposed not to be insensitive to moral concerns. Thus skepticism does not necessary make us unable to distinguish between a man who is not able to grasp moral reasons (because of some impairment) and an individual with a normal developed rationality. Nevertheless, skepticism undermines the idea that people deserve to be blamed just because they did what they decided to do, even if we can still maintain that, being insensitive to moral claims, they behaved badly. Supposing that what we said about (2) is consistent, responsiveness to reasons is not impaired just because determinism obtains. What we are doing when we formulate a judgment of moral wrongness is expressing a consideration about the moral quality of a certain act—both in a utilitarian and in a non-utilitarian framework, a single act can still be considered bad—joint with a consideration of the appropriateness of blame. A coherent moral judgment does not necessarily require the actual opportunity for the agent to do otherwise, but can be grounded in her general ability (due to her responsiveness to moral reasons) to modify her conduct and, eventually, to behave in a different way. Everything, including feelings and moral reasons, can become part of the chain of causes that determine the course of our actions and so, with a typical Spinozian spirit, the promotion of morality—whatever is the sense that we attribute to it—could still be a valuable goal.1

References Björnsson G. and D. Pereboom 2014, “Free Will Skepticism and Bypassing. Comments on Eddy Nahmias, ‘Is Free Will an Illusion?’”, in W. Sinnott-Armstrong (Ed.), Moral Psychology, Vol. 4, Boston: MIT Press Bok, H. 1998, Freedom and Responsibility, Princeton: Princeton University Press Clarke, R. 1997, “On the Possibility of Rational Free Action”, Philosophical Studies 88 (1), 37 —. 2003, Libertarian Accounts of Free Will, New York: Oxford University Press Copp, D. 2008, “‘Ought’ Implies ‘Can’, and the Derivation of the Principle of Alternate Possibilities”, Analysis 68, 67 De Caro, M. 2009, Il Libero Arbitrio. Una Introduzione, Roma-Bari: Laterza

162

Chapter Ten

—. 2013, “Il Problema Filosofico della Responsabilità”, in M. De Caro, A. Lavazza, and G. Sartori (Eds.), Quanto Siamo Responsabili? Filosofia, Neuroscienze E Società, Torino: Codice Edizioni, 25 Dennett, D.C. 1984, Elbow Room. The Varieties of Free Will Worth Wanting, Cambridge: MIT Press —. 2003, Freedom Evolves, London: Penguin Books Finch, A. 2013, “On Behalf of the Consequence Argument: Time, Modality, and the Nature of Free Action”, Philosophical Studies, 163, 151 Fischer J.M. 1986, “Van Inwagen on Free Will”, in L. Stevenson, R. Squires and J. Haldane (Eds.) Mind, Causation and Action, Oxford: Basil Blackwell —. 1995, The Metaphysics of Free Will: an Essay on Control, Oxford: Blackwell —. 2003, “‘Ought-Implies-Can’, Causal Determinism and Moral Responsibility”, Analysis 63 (3), 244 —. 2011, Deep Control: Essays on Free Will and Value, New York: Oxford University Press Fischer, J.M. and M. Ravizza (Eds.) 1993, Perspectives on Moral Responsibility, Ithaca: Cornell University Press Fischer, J.M. and M. Ravizza 1998, Responsibility and Control: A Theory of Moral Responsibility, New York: Cambridge University Press Fischer, J.M. and N. A. Tognazzini 2011, “The Physiognomy of Responsibility”, Philosophy and Phenomenological Research 82 (2), 381 Frankfurt, H. 1969, “Alternate Possibilities and Moral Responsibility”, Journal of Philosophy 66, 828 —. 1988, The Importance of What We Care about: Philosophical Essays, New York: Cambridge University Press Haji, I. 1998, Moral Appraisability, New York: Oxford University Press —. 2002, Deontic Morality and Control, Cambridge: Cambridge University Press —. 2003, “Determinism and Its Threat to the Moral Sentiments”, The Monist 86 (2), 242 —. 2011, “Obligation, Reason, and Frankfurt Examples”, in R. Kane (Ed.), The Oxford Handbook of Free Will, 2nd ed., New York: Oxford University Press, 288 —. 2012, Reason’s Debt to Freedom: Normative Appraisals, Reasons, and Free Will, New York: Oxford University Press Kane, R. 1985, Free Will and Values, New York: SUNY Press —. 2002a, Free Will, Hoboken: Wiley-Blackwell

Skepticism and Control

163

—. 2002b, “Some Neglected Pathways in the Free Will Labyrinth”, in R. Kane (Ed.), The Oxford Handbook of Free Will, 1st ed., New York: Oxford University Press, 406 Libet, B. 2002, “Do We Have Free Will?”, in R. Kane (Ed.), The Oxford Handbook of Free Will, 1st ed., New York: Oxford University Press, 551 McGinn, C. 1993, Problems in Philosophy: The Limits of Inquiry, Hoboken: Wiley-Blackwell McKenna, M. 2012, Conversation and Responsibility, New York: Oxford University Press Mele, A. 2006, Free Will and Luck, New York: Oxford University Press Nelkin, D. 2001, “The Consequence Argument and the Mind Argument”, Analysis 61 (2), 107 Nichols, S. 2007, “After Incompatibilism: A Naturalistic Defense Of The Reactive Attitudes”, Philosophical Perspectives 21 (1), 405 Nozick, R. 1981, Philosophical Explanations, Cambridge: Harvard University Press O’Connor, T. 2008, “Agent-Causal Power”, in H. Handfield (Ed.), Dispositions and Causes, Oxford: Oxford University Press Pereboom, D. 1995, “Determinism Al Dente”, Noûs 29, 21 —. 2001, Living without Free Will, Cambridge: Cambridge University Press —. 2002, “Meaning in Life Without Free Will”, Philosophic Exchange 33, 19 —. 2013, “Optimistic Skepticism about Free Will”, in P. Russell and O. Deery (Eds.), The Philosophy of Free Will: Selected Contemporary Readings, New York: Oxford University Press, 421 —. 2014, Free Will, Agency and Meaning in Life, Oxford: Oxford University Press Roskies, A.L. 2006, “Neuroscientific Challenges to Free Will and Responsibility”, Trends in Cognitive Sciences 10, 419 Russell, P. 1995, Freedom and Moral Sentiment, New York: Oxford University Press Sartorio, C. 2013, “Making a Difference in a Deterministic World”, Philosophical Review, 122 (2), 189 Scanlon, T. 1998, What We Owe to Each Other, Cambridge: Harvard University Press —. 2008, Moral Dimensions: Permissibility, Meaning and Blame, Cambridge: Harvard University Press Smart, J.J.C. 1961, “Praise and Blame”, Mind, 70 (279), 291

164

Chapter Ten

Smilansky, S. 2000. Free Will and Illusion, Oxford: Oxford University Press Soon, C.S., M. Brass, H.J. Heinze and J.D. Haynes 2008, “Unconscious Determinants of Free Decisions in the Human Brain”, Nature Neuroscience 11, 543 Strawson, G. 1986, Freedom and Belief, New York: Oxford University Press Strawson, P.F. 1962, “Freedom and Resentment”, in Id., 1974, Freedom and Resentment and Other Essays, London: Methuen van Inwagen, P. 1983, An Essay on Free Will, New York: Oxford University Press Vranas, P.B.M. 2007, “I Ought, Therefore I Can”, Philosophical Studies 136 (2), 167 Wallace, R.J. 1994, Responsibility and the Moral Sentiments, Cambridge: Harvard University Press Watson, G. 1975, “Free Agency”, Journal of Philosophy 72 (8), 205 —. 2004, Agency and Answerability: Selected Essays, Oxford: Oxford University Press Zimmerman, M.J. 1988, An Essay on Moral Responsibility, Totowa, NJ: Rowman & Littlefield

Notes 1 I am grateful to Derk Pereboom, who read different drafts of the present paper and supplied highly useful comments and suggestions. I wish to thank Mario De Caro, Luca Fonnesu and Guglielmo Feis for their invaluable help and support.

CHAPTER ELEVEN UNDERSTANDING STRENGTH OF WILL MICHAEL BRENT

Richard Holton has presented an important criticism of two prominent accounts of action, a criticism that employs a notion of strength of will. Holton claims that these well-known accounts of action cannot explain cases in which an agent adheres to the dictates of a previous resolution in spite of a persistent desire to the contrary. In this chapter, I present an explanation and defense of Holton’s criticism of these accounts of action, and then I argue that while Holton highlights a crucial deficiency in both, his own explanation of strength of will is problematic.

1. Strength of will as adherence to a resolution How do you succeed in persisting with a resolution in the face of a compelling desire to the contrary? For example, imagine that you have recently decided to give up espresso and that you now desire to refrain from doing so. Imagine further that unaware of your decision I present you with the opportunity to drink a freshly brewed espresso from your favorite café and you find yourself with a compelling desire to accept my offer, one whose influence is felt more powerfully than your previous desire. In spite of this strong desire, though, you refrain from accepting my offer, thereby adhering to your previous decision and displaying what shall here be called strength of will. According to Richard Holton, two prominent accounts of action cannot explain cases in which you manage to do this sort of thing.1 He describes the two accounts2 as follows: The Humean Account: All action is explained in terms of your beliefs and desires, where you act on whichever of your desires are strongest.3 On this account, when you adhere to the dictates of a resolution, the resolution itself must be understood either as the strongest desire or the strongest combination of beliefs and desires.

166

Chapter Eleven

The Modified Humean Account: All action is explained in terms of your beliefs, desires, and intentions, where intentions are a sui generis kind of motivational factor, irreducible to the other two, and where you act on whichever motivational factor is strongest.4 On this account, when you adhere to the dictates of a resolution, the resolution itself is the strongest motivational factor. How might a proponent of the Humean Account of action explain such cases? According to Holton, the most promising way to explain strength of will on the Humean Account is as follows. Using the above example, after deciding that you should give up espresso, you thereby come to desire that you refrain from drinking espresso rather than doing so. When presented with the opportunity to drink a freshly brewed cup from your favorite café, though, you are confronted with a compelling desire to drink it, a desire that is felt more powerfully than the previous desire to refrain from drinking. If the Humean Account of action is correct and you must act on your strongest desire, then you will succumb to the compelling desire to drink espresso and thereby fail to display strength of will. So how might people who are confronted with this pattern of desires show strength of will? What might the proponent of the Humean Account say? Holton identifies two options, one in which further desires are added to the mix, and another that involves adding further beliefs. For the first option, Holton suggests that the proponent of the Humean Account could add a strong desire to adhere to your previous decision, thereby understanding a resolution as special kind of desire that is designed to block compelling desires to the contrary.5 Here, a resolution is a secondorder desire to be unmoved by particular first-order desires. Thus, at the moment in time when you are confronted with a compelling desire to drink an espresso, provided that your second-order desire to be unmoved by precisely this kind of first-order desire is stronger than the first-order desire itself, you have a desire-driven way in which you can resist the temptation to break with your previous decision. For the second option, Holton suggests that the proponent of the Humean Account could add a further belief rather than an additional desire.6 The belief in question involves accepting two propositions: (a) if you resist the next espresso, you will give up drinking espresso for good; (b) if you fail to resist the next cup of espresso, you will fail to give up drinking espresso for good. The first proposition expresses the idea that resisting the next cup of espresso will be an effective means of giving up espresso for good, so that accepting the proposition will enable you to

Understanding Strength of Will

167

believe that resisting the desire to drink the next cup of espresso will play an instrumental role in realizing your desire to give up espresso for good. The second proposition expresses the idea that resisting the next cup of espresso is necessary in order for you to be successful in giving up espresso for good, so that accepting the proposition will enable you to avoid believing that you can both drink the next cup of espresso and be successful in giving up espresso for good. Here, a resolution is a twopronged belief that is designed to reinforce the motivational power of your decision in the presence of strong inclinations to do otherwise. Thus, at the moment when you are confronted with a compelling desire to drink a cup of espresso, provided that you believe both propositions to be true, you have a belief-driven way in which you can strengthen your desire to resist the temptation to break with your previous decision. Now, according to Holton, neither option saves the Humean Account of action from the charge of implausibility. Why? Even if we incorporate the above responses into the Humean Account, it maintains that adhering to a resolution consists in the triumph of a stronger desire over a weaker one, a victory that occurs in any situation in which you act in light of your strongest desire. Thus, if the Humean Account were correct, then we would expect that the experience of acting in light of your strongest desire in mundane cases would be just like or identical to the experience of acting in accordance with a resolution in the face of a compelling desire to the contrary, since in both cases your action results simply from the triumph of a stronger desire over a weaker one. But this is not correct, says Holton, for it is often the case that you must struggle to maintain a resolution in the face of a desire to do otherwise. That is, it is often the case that you adhere to a resolution only by way of struggling to resist or overcome a compelling desire to the contrary. The struggle to maintain a resolution in such circumstances is importantly different from what occurs when you make a decision amongst a variety of mundane options or simply act in light of your strongest desire, and Holton is correct to point out that such struggle is omitted by the Humean Account of action precisely because the account explains action only in terms of beliefs and desires, where you act on whichever desire is strongest. How might we explain this kind of struggle? Holton suggests that we might augment the Humean Account along the lines suggested by proponents of the Modified Humean Account of action. In so doing, intentions are understood to be neither desires nor a conjunction of desires and beliefs, but as a unique kind of mental state, irreducible to the other two. Intentions are motivating states that can move one to act and that can preserve the motivational force of an earlier belief or desire, even if the

168

Chapter Eleven

earlier belief or desire is no longer present to mind, and even if there are contrary desires urging one to do otherwise. On this account, a resolution can be seen as a particular kind of intention that one forms precisely so as to defeat any contrary desires that might emerge at a later point in time. It is worth emphasizing the ways in which this Modified Account of a resolution differs from that offered by proponents of the Humean Account. On the Modified Account, the number of motivational factors is enlarged, so that to be motivated to act, one need not require a desire and a belief, and action need not be the result of whichever desire is the strongest. An intention can exert its own kind of motivational force, so that even in the absence of the desire or belief that gave rise to it in the first place, the intention can overcome whatever desires are present at the time of action. Rather than saying that one acts always to satisfy one’s strongest desire, the Modified Account claims that one acts always in light of one’s strongest motivational factor, which includes one’s beliefs, desires, and intentions. Thus, when an agent adheres to the dictates of a resolution, the resolution itself is the strongest motivational factor. Holton thinks that the Modified Humean Account of action is in trouble for the same reason that he thinks that the original Humean Account is problematic. The trouble is that both accounts omit a crucial element that is present in many cases where one adheres to the dictates of a resolution: namely, the struggle required when forcing oneself to remain resolute in the face of a desire to the contrary. The Modified Account of a resolution construes success in such cases in terms of the strength of a particular motivational factor, so that strength of will occurs when one lets the strongest of one’s motivational factors have its way. Here, the struggle that can occur in situations where one displays of strength of will is not accounted for. I shall say more about Holton’s criticism below, after introducing his account of strength of will.

2. Holton’s account of strength of will Holton thinks that the best way to explain strength of will is by introducing another motivational factor into the equation, namely, that of willpower. Introducing the notion of willpower enables us to explain strength of will in terms of your beliefs, desires, intentions and the strength of your willpower, understood as a separate factor. According to Holton, the notion of willpower is that of a cognitive capacity that you actively employ as such. It can be likened to a muscle, insofar as it requires a distinctive kind of effort to use, it can tire in the short term, and it can be strengthened over time. Explaining strength of will in terms of a

Understanding Strength of Will

169

distinctive notion of willpower incorporates the fact that often times you must struggle in order to resist the temptation of a compelling desire that threatens to subvert your resolution. The struggle itself is something that you do or do not exert, and Holton claims that it is required because you are actively employing your willpower in the attempt to overcome a desire to the contrary. That is, the struggle is evidence that you are using a distinctive cognitive capacity to remain resolute in the presence of a desire to do otherwise. Crucially, for Holton the struggle to resist a compelling desire is not straightforwardly physical, such as that involved when lifting a heavy object or walking uphill, since it can be present whether the resolution is to perform an action or to refrain from performing an action. Holton claims that no matter how strong the desire to drink espresso might be, it is not the case that the struggle to resist it consists in actually preventing muscles that are straining to reach for the cup. Rather, the struggle involved in resisting a desire that threatens to thwart a resolution is best understood as mental. In particular, it is the mental act of refusing to revise a resolution by not reconsidering it, in spite of the presence of a powerful inclination to do just that.7 The state of mind in question is one in which you are aware of the resolution, and perhaps even the consideration(s) for which it was originally adopted, but it is not reconsidered or reevaluated. You merely call it to mind in a kind of passive rehearsal, and you do not allow yourself to embark on a procedure that would be involved in revising it. Here, you must struggle in order to call to mind the resolution at precisely the moment at which it is being threatened by a competing inclination to do otherwise. When all goes well, you are able to resist the tempting course of action by refusing to revise and reconsider a resolution designed to prohibit that very course of action.8 Thus, strength of will is the ability to retain a firm and unwavering commitment to your resolution by calling the resolution to mind at the moment in time when it is needed and refusing to reconsider or alter it in any way. What evidence is there that such a distinct capacity exists? Holton provides three sources of evidence that the capacity is distinct, each from recent research in social psychology. First, the ability to abide by a resolution looks to be affected by factors that are distinct from the beliefs, desires, intentions, and resolutions themselves. For example, reformed alcoholics, dieters, and people who are trying to quit smoking are more likely to forgo their commitment to abstaining from alcohol, food, or cigarettes when they are depressed, anxious, or tired.9 States of this kind systematically affect your ability to act in line with all of your resolutions, be they resolutions not to drink, not to over-eat, not to smoke, or whatever.

170

Chapter Eleven

According to Holton, the most likely explanation of this fact is that such states do not systematically strengthen your desires to perform the prohibited actions, but rather they inhibit your ability to follow any resolution that you might have formed. Second, it appears that willpower is limited. For example, forcing yourself to eat radishes rather than chocolate makes you less likely to persist later on in solving a difficult puzzle, and suppressing your emotional reactions to a film makes you less likely to persist later on in maintaining your grip on a handle.10 According to Holton, the most likely explanation of such facts is that the ability to sustain a resolution is affected by the strength of your willpower at that moment in time. That is, it seems that the ability to persist in a course of action is determined not simply by the strength of your desires, beliefs, intentions, and resolutions, but also by the strength of your willpower, precisely that motivational factor that appears to be depleted by repeated or earlier use. Third, it seems that your willpower can be developed and strengthened by repeated practice. Experimental subjects who undergo a regime of selfregulatory exercises, such as working on the improvement of posture, display a significantly reduced tendency to suffer from depletion of willpower.11 Much like Aristotle’s claim that you can become virtuous by acting virtuously, it appears that you can become strong-willed by acting in precisely that way. As Holton himself admits, such evidence does not conclusively prove that such a capacity exists, but I think he is correct to suggest that it provides additional and compelling grounds to think that there is a distinct cognitive capacity that is employed when actively recalling and refusing to reconsider a resolution in the face of compelling desires to do just that. Indeed, postulating the existence of such a capacity does seem to be a promising way of explaining the distinctive kind of struggle that is displayed in cases where you act in this kind of strong-willed manner. As they currently stand, both the original Humean Account and the Modified version do not have the resources for explaining the characteristic struggle that you exert when overcoming a strong desire that threatens to undermine a resolution. Both depict the strong-willed agent in an impoverished way, omitting a crucial feature of action that seems present in many different situations. However, although Holton has raised an important and successful criticism of both accounts of action, his own explanation of strength of will is not without its difficulties, as we shall see in the next section.

Understanding Strength of Will

171

3. Problems for Holton There are two related reasons why Holton’s account of strength of will is problematic. The first concerns the causal necessity of the mental action of recalling the relevant resolution and refusing to revise it, and the second concerns its relative causal strength. Holton has not offered an explanation of the causal process by which strength of will occurs in the kind of case introduced in §1, nor has he specified why displaying strength of will requires that you recall a resolution and refuse to revise it, rather than, say, simply refusing to perform the pertinent action. The second problem is that, rather surprisingly, Holton’s account of strength of will lacks an explication of the notion of strength. The result is that it is unclear whether his account is applicable in cases where an action that is already underway is threatened by a pernicious desire. Let’s examine each in turn.12 First, Holton says very little about how we are to understand the mental action of refusal. Specifically, he does not offer an explanation of its causal role in enabling you to overcome the powerful desire that threatens to undermine your resolution and thereby refrain from performing the relevant action. We are thus left wondering why as a strong-willed agent it is necessary that you recall the resolution and actively refuse to revise it as part of the process by which you overcome the pernicious desire. For instance, Holton does not specify whether refusing to alter the resolution thereby increases its relative motivational strength, or whether it diminishes or altogether blocks the motivational force of the problematic desire, or whether it functions in another way altogether. As a result, it is unclear whether, for the strong-willed agent, the act of recalling the resolution and refusing to alter it makes that resolution stronger than the threatening desire, or whether that act suppresses the motivational strength of the threatening desire without affecting that of the resolution, or something else entirely. Moreover, even if we assume on Holton’s behalf that the act of recall and refusal functions in one of these ways, how does doing so enable you to refrain from performing an action that would satisfy the threatening desire? After all, it seems possible that two otherwise indistinguishable agents, both of whom have previously resolved to refrain from drinking espresso and are presently confronted with an equally potent desire to drink a freshly brewed cup, might successfully recall that resolution and refuse to alter it in any way, and yet one such agent drinks the espresso whereas the other agent does not. From a causal perspective, what might explain the difference between such otherwise identical agents? The account presented by Holton suggests only that strong-willed agents tend to be capable of

172

Chapter Eleven

refusing to alter the relevant resolution, but this does not explain the causal process by which refusing to alter a resolution can enable one agent but not the other to refrain from performing the relevant action. As a result, we are left wanting an explanation of the difference between such otherwise identical agents, from the perspective of the causal factors at work. I shall return to this point below. In addition, Holton has not explained why the action of recalling and refusing to revise your resolution is a causally necessary feature of the process by which you display strength of will. Thus, his account does not rule out the possibility that you can overcome a potentially threatening desire not by refusing to revise your resolution, but by refusing to perform the relevant action. Holton assumes that in cases where you display strength of will, the problematic desire is a desire that threatens your prior resolution. This, however, is not obviously correct. When confronted with the opportunity to drink an espresso and the very potent desire to do so, that desire seems to threaten your resolution only indirectly. In order for the newly acquired desire to threaten your resolution, it seems that you must be aware of the conflict between this new desire and your resolution, and that succumbing to this new desire would thereby undermine the latter. But even granting you an awareness of this, why must you also refuse to revise your resolution so as to avoid drinking the espresso? Why not refuse to drink it outright, as it were, without refusing to revise your resolution? The mental action of recalling and refusing to revise your resolution is a potentially unnecessary step in the process by which you overcome a pernicious desire, so Holton owes an explanation of why it should play this role. Now, the second reason why the account of strength of will offered by Holton is problematic is that, rather surprisingly, it lacks an exposition of the notion of strength. Willpower, says Holton, is a cognitive capacity the direct exercise of which consists in the mental action of recalling a resolution and refusing to reconsider it. If this cognitive capacity is sufficiently robust, then doing so will enable you to succeed in adhering to your resolution. But what does it mean for this cognitive capacity to be sufficiently robust? In what does its relative strength consist? Holton does not say, and this is problematic. This is problematic because there are cases in which recalling a resolution and refusing to reconsider it seems to be causally insufficient, yet you nevertheless display strength of will. For example, consider a scenario in which you display strength of will when persisting in the performance of an action that is already underway. Imagine that after a lengthy process of deliberation you recently decided to run a marathon, formed a resolution to do so, and embarked on an

Understanding Strength of Will

173

ambitious training plan. Imagine further that today is the day of the race, conditions are normal, and you are in the midst of running the twenty-third mile. As you are nearing the end of that mile, you are suddenly confronted with a very strong desire to give up: you are in pain and near exhaustion, you no longer believe that you can finish the race, and you desire to stop running more so than you desire to finish. Finally, imagine that in spite of all this, you manage to overcome that desire, pain, and near exhaustion and complete the marathon, thereby adhering to your resolution and displaying strength of will. Does Holton’s account apply in such cases? That is, does his account explain how it is that you are able to force yourself to continue running in spite of your newly acquired desire, the pain and exhaustion, and lack of self-confidence, thereby adhering to your resolution and displaying strength of will? The example suggests that, in cases where an action is already underway, adhering to your original resolution requires more than calling it to mind and refusing to revise it. In addition, while you are in the midst of exercising what Holton describes as willpower, you must also force yourself to remain in motion while overcoming the new desire to stop, the looming self-doubt, and the pain and near exhaustion. That is, you must also exert a great amount of effort so as to force your legs to remain in motion in spite of the potent desire to stop running the race. Crucially, the exertion of effort required to sustain and control your bodily capacities in such demanding scenarios is different from that required to recall your resolution and refuse to alter it, insofar as it is deployed in a distinctively bodily manner to sustain the ongoing activity of the requisite bodily capacities. Such effort is not accounted for by Holton’s description of what happens when you display strength of will, since his account restricts the applicability of the notion of effort to the use of the relevant cognitive capacity when refusing to alter a resolution. Displaying strength of will in cases of this kind requires that you continue to perform the relevant bodily action the moment that your newly acquired desire to stop running is felt more powerfully than your resolution to continue, as well as during your recall of the resolution, and even as you refuse to alter your resolution. In each moment during this process of recall and refusal, you remain in motion as a result of your ongoing exertion of effort, which sustains the movement of your legs and body, and which is distinct from the effort that you exert while using the cognitive capacity that figures in Holton’s account. Thus, since Holton restricts the applicability of the notion of willpower by construing it as a cognitive capacity that you employ specifically when refusing to alter a resolution in light of a pernicious desire, his account does not apply in situations where the

174

Chapter Eleven

relevant action is already underway and the effort that you exert so as to continue performing that action is more than merely cognitive.

4. Beyond mere adherence Once we acknowledge this wider causal role of your exertion of effort during the performance of your action, we can provide a more persuasive and comprehensive account of strength of will. Recall the abovementioned agents who are otherwise indistinguishable and who have both resolved to avoid drinking espresso and yet only one manages to adhere to the relevant resolution when confronted with a potent desire to drink an espresso. From the perspective of the relevant causal factors at work, the difference between these agents is not explained merely in terms of their awareness of the relevant resolution, their recalling it in the way that Holton describes, and their refusal to alter it. That is, when confronted with such a compelling desire, the act of recalling and refusing to revise that resolution can be a part of the process by which a potent desire to the contrary is overcome, and this can require a distinctive kind of effort, namely, that of remaining steadfast in your thoughts. But just as in scenarios where the relevant action is already underway, the activity of bringing the resolution to awareness and refusing to alter it is not sufficient for causally initiating, sustaining, and controlling the requisite bodily capacities in the manner demanded by the resolution. In fact, the resolution itself, just like the relevant desires, beliefs, and intentions, is not causally sufficient for this. It seems that you are causally responsible for this, inasmuch as through your exertion of effort you are initiating, sustaining, and controlling the activity of your bodily capacities and the relevant action, in light of the resolution that you have called to mind, and in addition to your act of refusing to alter it. Notice that for the proponent of the Humean and Modified Humean Accounts of action, the explanation of action requires that we refer only to your states of mind in which your desires, beliefs, or intentions are present, rather than to you, the agent. Indeed, the connection between both Humean Accounts and the philosophical commitments of David Hume himself are particularly relevant here.13 Famously, Hume denied that there was any such thing as you qua agent (or “the self”), in addition to the states of mind (or “perceptions of the mind”) that are present to awareness at any given moment in time, connected by what he described as the Principles of Association. It seems that contemporary proponents of both Humean Accounts retain an inexplicit commitment to such a picture of the agent. By highlighting your active and causal role in cases where you display

Understanding Strength of Will

175

strength of will in overcoming an urge that threatens to undermine the action that you are in the midst of performing, the assumption that we need not refer to your causal contribution in our explanation of action is problematic. Unfortunately, the account of strength of will offered by Holton does not fare much better. Although I disagree with the details of his account, Holton nevertheless recognizes a distinctive and active causal contribution for you, the agent, a role that is made explicit by cases in which you display strength of will, as Holton describes that notion. For Holton, by recalling a resolution without revising or reevaluating it, you become aware of the resolution at precisely the moment when doing so is required and, when all goes well, are thereby able to overcome a compelling desire to the contrary. But this places severe limitations on the active causal role that you play during the performance of your bodily actions, especially in cases where what Holton describes as strength of will is not required. It is only in cases where you must intervene, so to speak, and overcome the force of a potent desire, that we see a distinctively active and causal role for you during your performance of an action. For Holton, when strength of will is not required, the strongest of your desires, beliefs, and intentions cause the action that you perform. This can be understood as a commitment to a kind of psychological determinism, in which bodily actions are causally determined by your desires, beliefs, intentions, and other such motivational factors, rather than you, the agent.14 It is the underlying commitment to this claim that is the most problematic aspect of Holton’s account of strength of will.

5. Concluding remarks In this chapter, I have presented and further supported Richard Holton’s novel criticism of both the Humean and the Modified Humean Accounts of action, and claimed that, although headed in the right direction, the positive account of strength of will offered by Holton requires emendation. In particular, I argued that, on Holton’s account of strength of will, we are left wanting an explanation of how, exactly, you increase the motivational efficacy of a resolution simply by bringing it to awareness without revision or reevaluation, and that Holton’s account does not appear applicable to cases where you resist a strong desire that threatens to undermine the action that you are in the midst of performing and have resolved to complete. When we limit our conception of willpower to the kinds of cases that Holton considers, we limit ourselves to thinking of willpower as a cognitive capacity limited to the mental

176

Chapter Eleven

action of refusing to revise or alter a resolution. In response to these difficulties, I suggested that what’s missing from the account of strength of will presented by Holton is acknowledgement of the wider role of your exertion of effort, as evinced by cases where you force yourself to continue performing an action that is already underway as you resist a potent desire to the contrary, in addition to those sorts of scenarios described by Holton, where you display strength of will in overcoming a potent desire that threatens to undermine a prior resolution.15 .

References Baumeister, R., T.F. Heatherton, and D.M. Tice 1994, Losing Control: How and Why People Fail at Self-Regulation, San Diego: Academic Press Baumeister, R., E. Bratslavsky, M. Muraven, and D.M. Tice 1998, “EgoDepletion: Is the Active Self a Limited Resource?”, Journal of Personality and Social Psychology 74 (5), 1252 Brandt, R. 1988, “The Structure of Virtue”, Midwest Studies in Philosophy 13 (1), 64 Bratman, M. 1989, Intention, Plans and Practical Reason, Cambridge, Mass.: Harvard University Press Davidson, D. 2001, “Actions, Reasons, and Causes”, Essays on Actions and Events, New York: Oxford University Press Holton, R. 1999, “Intention and Weakness of Will”, Journal of Philosophy 96 (5), 241 —. 2003, “How Is Strength of Will Possible?” in Stroud S. and C. Tappolet (eds.), Weakness of Will and Practical Irrationality, New York: Oxford University Press —. 2009, Willing, Wanting, Waiting, New York: Oxford University Press Hume, D. (1739-1740) 1975, A Treatise of Human Nature, edited by L. A. Selby-Bigge, 2nd ed. revised by P. H. Nidditch, Oxford: Clarendon Press Kennett, J. and M. Smith 1996, “Frog and Toad Lose Control”, Analysis 56 (2), 63 Kennett, J. and M. Smith 1997, “Synchronic Self-Control is Always NonActional”, Analysis 57 (2), 123 Kennett, J. 2001, Agency and Responsibility: A Common-Sense Moral Psychology, New York: Oxford University Press Mele, A. 1987, Irrationality: An Essay on Akrasia, Self-Deception, and Self-Control, New York: Oxford University Press

Understanding Strength of Will

177

—. 1998, “Synchronic Self-Control Revisited: Frog and Toad Shape Up”, Analysis 58 (4), 305 —. 2003, Motivation and Agency, New York: Oxford University Press Muraven, M., R. Baumeister, and D.M. Tice 1999, “Longitudinal Improvement of Self-Regulation Through Practice: Building SelfControl Strength Through Repeated Exercise”, The Journal of Social Psychology 139 (4), 446 Pettit, P. and M. Smith 1993, “Brandt on Self-Control”, in Hooker B. (ed.), Rationality, Rules, and Utility: New Essays on the Moral Philosophy of Richard Brandt, Boulder: Westview Press, 33 Velleman, J.D. 1992, “What Happens When Someone Acts?”, Mind 101 (403), 461

Notes 1

See Holton (1999, 2003, 2009). Note that the accounts of action in question are an instance of what J. David Velleman (1992) has dubbed the standard story of action. Both accounts depict bodily actions as events that are caused by the onset of those of your beliefs, desires, intentions, and other motivational factors that make intelligible your performance of the action in question. Part of Holton’s interest in these accounts is whether they can explain your ability to adhere to a resolution in the face of a strong desire to the contrary, given the way that they account for the causation of action. See Holton (2003, p. 40). 3 Proponents of this account include Davidson (2001). 4 Holton cites the work of Michael Bratman (1989). Note that when Holton presents both accounts of action, he does not specify what it means to say of a belief, desire, intention, or other motivational factor that it is “strongest”. Very roughly put, we can assume that all else being equal, for one desire, A, to be stronger than another desire, B, is for you to be disposed to act upon A rather than B, where you believe that each desire can be satisfied by performing a specific action. 5 Defenders of a desire-based option include Brandt (1988) and Mele (1987, 1998, 2003). 6 Defenders of a belief-based account include Pettit and Smith (1993); Kennett and Smith (1996, 1997) and Kennett (2001). 7 For Holton, the difference between the reconsideration of a resolution and the revision of a resolution is this: to revise a resolution is to alter it in some way; to reconsider a resolution is to suspend it and thereby open oneself to the possibility of revising it. In order to refuse to reconsider a resolution and thereby to abide by its dictates in the presence of a compelling desire to do otherwise, Holton thinks that one must actively exercise this distinct cognitive capacity. 2

178

8

Chapter Eleven

Holton does not put the point precisely in this way, but I think it is the most perspicuous way to do so. As I shall argue below, this aspect of Holton’s account of strength of will is problematic. 9 Holton cites Baumeister, Heatherton and Tice (1994, pp. 151ff). 10 Holton cites empirical literature on what is called “ego depletion”. See Baumeister, Bratslavsky, Muraven and Tice (1998, pp. 1252-65). 11 Holton cites Muraven, Baumeister and Tice (1999). 12 It is important to note that there are interesting normative issues related to strength of will, such as whether there are conditions in which it would be inappropriate for you to adhere to a resolution, say, that I shall here set aside. My worries with the account that Holton offers concern only its causal dimension. 13 See Hume (1739-40, p. 252). 14 Note that “psychological determinism” is distinct from “physical determinism”, the latter of which is a claim that some physicists and philosophers are in the business of investigating. 15 For helpful discussion of earlier versions of this paper, I thank Akeel Bilgrami, Carol Rovane, Taylor Carman, Janet Metcalfe, Mario De Caro, Anubav Vasudevan, Marco J. Nathan, Brian H. Kim, Alex Madva, Katie Gasdaglis, Katherine Rickus, and Andrei Buckareff.



CHAPTER TWELVE RESPONSIBILITY AND NARRATIVE AGENCY IRENE BUCELLI

A large portion of the existing work in action theory shares the intuition that normal adult human agents are agents in a rather unique way. Human agency entitles us to ask normative questions such as “Should she have done this under these circumstances?”, and it seems to be distinctive of persons that we accord them a status as moral agents. Moreover, it is most natural to include responsibility as a characteristic peculiar to normal adult human agency, thus making normal adult human agents the paradigm of responsible agents. Undoubtedly, the work done in action theory to spell out the conditions for agency seems to have important consequences for the debate around the sort of creature that can properly be held responsible for her actions. In this paper I will consider theories of agency that ascribe a crucial role to the agents’ reflective endorsement. These theories,1 although presenting different models, urge that agential authority is required for our behaviour to count as a full-blown action and identify the condition of such authority in the agent’s reflective endorsement: in the commitment to his own doings by means of his reflective capacities. On the one hand this position points at the human capacity for self-consciousness as the feature that puts humans in a special relation to their behaviour. Because of this capacity humans can reflect: they can step back and question their actions as well as their own motives and attitudes. On the other hand, the agent commits to some of these motives, regarding them as reasons, and thus performs the relevant behaviour in light of them. Reason-responsiveness makes our actions something different to behaviour springing from whatever state we are in. The model of agency that results from these considerations is one that assigns two dimensions to agency: on the one hand there are full-blown actions, resulting from the agent’s selfgovernance; on the other hand there are the kind of activities that even



180

Chapter Twelve

some non-human animals can perform, despite lacking our cognitive reflective capacities. I will explore some ways in which the enterprise of establishing the conditions for agency in these terms can bear on responsibility. My very general question will hence be revolving “what implications do reflective endorsement theories of agency have for responsibility”? As ambitious as the project might sound, my aims will actually be quite modest. For example, I limit my analysis to David Velleman’s narrative model of agency. Additionally, I will discuss a simple argument for a direct connection between agency and responsibility. By dismissing it I hope to point at the kind of questions that a reflective endorsement theory has to answer in order to have relevant consequences for the work done in responsibility. My modest results will clear the way for future research. In doing this, I take it that endorsement theories seem promising in having consequences the debate on responsibility, and, within this framework, I believe that Velleman’s theory has significant advantages. In the first part of this paper, I present Velleman’s narrative model. In the second, I discuss the most direct way in which we could conceive of the relationship between agency and responsibility, showing why it is not a workable hypothesis. In doing so I am not claiming to dismiss other possible ways to understand the connection between the proposed view and responsibility, I rather wish to hint at possible future directions of research. Moreover, much of this paper is devoted to tying up loose ends in Velleman’s proposal in action theory. This means that in presenting the narrative model, I am accommodating material that is spread in different collections of articles. I consider that the work on narrativity completes his project on practical reasoning.2 Moreover, in drawing the possible consequences for responsibility, I do not claim to be arguing against Velleman’s own position on the topic,3 but rather I want to discuss what could seem a possible upshot of his model, as I will present it.

1. The narrative model of agency Velleman’s starting point is the idea that mere talk of a belief-desire structure, explaining how our actions are caused, does not warrant our treating such causally relevant motives as reasons. Velleman contrasts actions with slips or unconsciously motivated movements, in which my behaviour is not guided by my motive as a reason (and indeed in the performance I am not even aware of this desire of mine).4 His conception of reflective endorsement argues that motives can count as reasons only



Responsibility and Narrative Agency

181

when the subject reflects upon them and reinforces them by selfunderstanding. This means that within our psychological make-up there are intellectual motives that inhibit or reinforce others, and these motives are the ones operative when we “stand back and reflect”. When we reflect on our motives their causal force has no coercive hold over us, but this is so not because the agent is somehow over and above his desires as some sort of pure reasoner: in reality when “standing back and reflecting” we are doing what our predominant motives push us to do, our predominant motives at this stage being intellectual ones, striving for self-understanding. According to this picture we are “merely a system of passion in which some motives—intellectual ones—can hold back others”.5 Velleman thus talks of an “intelligibility drive”: a tendency to make sense of what we do and what goes on around us but which, unlike our reasons, does not need to figure among our conscious motives. We strive to perform behaviour that “makes sense”, which is reinforced by the motivation associated with this drive and counts as an action I have reasons for. This drive for selfknowledge reinforces certain courses of action instead of others because people have a motive to do what makes sense to them. In the balance of my overall motivation, my belief that ij-ing makes sense strengthens the desire that motivated me to ij in the first place and as such can be causally efficacious. Ultimately then, reasons can be causes inasmuch as the intelligibility drive reinforces the original motivation to perform those actions associated with motives that count as reasons. Defining this as a “narrative model” characterizes the terms of endorsement in Velleman’s picture: narratives provide the interpretative framework in the light of which we can establish “what makes sense” and hence our reasons. The condition for something to be endorsed is that it passes a coherence test, and the background on to which something can be established as coherent is my own self-conception.6 In testing his motives against the background of his narrative self-conception, the agent can identify his competing motives and throw his weight behind a certain motivation, reinforcing it via his endorsement. Coherence constrains the range of actions available to us. Only when acting through one’s narrative self-conception is one truly acting, hence making narrativity constitutive of what we can properly call agency.7 When saying “acting through” a narrative self-conception, we need to distinguish between fundamental drives that, according to Velleman, characterize human psychology. We have already seen how self-understanding plays a role: in exercising his capacity for self-consciousness the agent places his motives and his actions within a narrative framework that is necessarily associated with a



182

Chapter Twelve

certain self-description. A second drive, a drive for self-consistency, preserves coherence between behaviour and self-knowledge. This is the tendency to act in accordance with our self-representation in a way that makes our behaviour cohere with our narrative. It is because of this drive that we can conceive of action as fitting behaviour into a selfinterpretation. This is the process of enacting a coherent narrative and thus actions are the fulfilment of the agent’s narrative self-conception. For example, I can find myself in the position of leaving earlier from a birthday party because some other friends invited me somewhere else. Only inasmuch as I can coherently fit my leaving (the honest excuses I am to give or the lies I decide to tell instead) within my self-conception, will I be willing to leave (and tell the relevant excuses or lies). The way in which agential authority is connected to consciousness is based on interpreting the agent as essentially a narrator: the author of his own autobiography, enacting the narrative he himself makes up. In considering acting as enacting an implicit narrative Velleman attributes to narrativity a constitutive role in agency making it distinctive of full-blown actions, with the result that establishing the limits and the conditions of narrative practice defines the realm of one’s agential authority. There are a few reasons to privilege this narrative model when discussing reflective endorsement theories. Firstly, the reference to narratives is appealing inasmuch as it is intuitive that story telling does play a significant role in our lives. Narrative practice seems pervasive: 8 we are all familiar with phrasing our reasons in terms of stories. Secondly, conceiving the test for reflective endorsement in terms of narrative coherence easily accommodates conflicting motives and changes within the agent’s preferences. It only requires one’s motives to fit within one’s narrative in order for them to count as reasons. This avoids constraints to the agent’s wholeheartedness, since narrativity does not impose any such condition on the agent.9 Most importantly though, the narrative model attempts to understand reflective endorsement, avoiding some overly intellectualistic10 tendencies that we can trace in other projects.11 The narrative model can account for an extensive range of actions, because it rejects the idea that all full-blown action necessarily requires deliberation. The reflective intelligibility of practical reasoning is not conveyed by connections between the content of explicit thoughts, but it relies on background self-ascriptions that implicitly register the attitudes that the explicit thoughts express. Though unarticulated, these self-attributions serve as reasons for acting inasmuch



Responsibility and Narrative Agency

183

as they provide the logical structure of the agent’s thinking, but the structure need not be explicit.12 Reflection need not to be understood as a discursive, time-consuming procedure that articulates the different steps of a practical syllogism and that our motives need to undergo in order to come out as reasons for action. Instead, we can conceive our reasoning as running in parallel with behaviour and we can posit two kinds of reflection that can be at work in full-blown actions: one that can be associated with deliberation and one that would mainly be supervising the stream of our behaviour already underway. I will call this kind of reflection ‘perceptual’ because it does not involve the agent as explicitly reflecting prior to the performance of his behaviour, but rather it just involves the agent’s perception of the circumstances. This conception of reflection makes it possible to allow the most ordinary acts, which we perform regularly, seemingly automatically and mindlessly, to be accounted for as fullblooded action, although they do not involve instances of deliberation.

2. The narrative model and responsibility Reflective endorsement theories of agency might seem good candidates to highlight necessary features of responsibility. After all, only beings that can act for reasons can qualify as responsible, and they aim at isolating a portion of behaviour that only normal adult human agents can perform. Moreover, there is another reason why the narrative model seems to be particularly promising. The standard understanding of responsibility that follows the lead of Aristotle’s analysis in the Nicomachean Ethics,13 sees agents as responsible only for those things they can do voluntarily and establishes two conditions for their behaviour to qualify as such. One condition is a volitional condition,14 which refers to the “origin in the agent”15 of the relevant behaviour. This condition ends up excluding anything that was not up to the agent to bring about. The second is an epistemic condition by which the agent must know what he is bringing about. In the narrative model these two conditions are nicely tied together since the agent’s very motivational setup includes those cognitive motives that reinforce him to perform only what makes sense to him (and normally inhibit him otherwise): behaviour can hence be up to him only inasmuch as he exercises this form of cognitive control. So it seems that what the narrative model ends up identifying as action already has the two standard conditions for responsibility readily built in. I will now try to highlight a



Chapter Twelve

184

possible way in which this view of agency can have implications for our theory of responsibility. A simple argument would state that, assuming a)

we are responsible for our actions,

b) some condition C is constitutive of agency, and hence conclude that c) the very same condition C would be constitutive of responsibility. This would establish a direct connection between agency and responsibility. If true, reflective theories of agency that stress the two dimensions that necessarily characterize our behaviour (opposing actions in their full blown sense to mere activities) would seem to have an incredible explanatory significance. Indeed, if defensible, this would make the work done on the necessary conditions for full-blown agency essential for the responsibility debate. In this light, the narrative model, which identifies narrativity as the necessary condition for agency, would also result in the claim that narrativity is necessary for responsibility. However, this simple argument seems to be a non-starter. Even in this very general formulation, assumption a) has no reference to the other very important class of things we are standardly considered responsible for: omissions. The fact that omissions are not mentioned is particularly relevant because it would seem that reflective endorsement strategies have special problems accommodating them. These problems can concern either the reflective component of these theories of action or their reference to the agent’s endorsement; I will discuss the former in detail and I will just offer some remarks on the latter. I will run the discussion by considering reflective endorsement in general and then turn to the narrative model to see how it can face the objections.16 The reflective character of agency raises a first concern inasmuch as the picture of the agent’s standing back and evaluating seems to be very close to a deliberative process (explicit or implicit), which just does not occur in cases of negligence/omissions.17 Consider the following scenario: My partner has put me in charge of doing the shopping for the dinner, which we have organized tonight for a group of friends. When just about to leave work I receive some unexpected news that causes me to stay much longer at the office, busy dealing with the sudden task that I need to



Responsibility and Narrative Agency

185

solve in a hurry. I am distressed and forget completely about dinner. When I get home I see my partner and realize I have not gone shopping. It is too late to go now and we will have to either call the dinner off or suggest eating out instead. I take it that it is a common intuition that in this scenario I am responsible, maybe blameworthy. It would be a mistake to see at any point in this story (under described as it might be) my omission to buy dinner as the result of some reflective process in the form of deliberation. Deliberation is a form of voluntary activity, one that aims at a resolution (about what to believe in the case of theoretical deliberation; about what to do in the case of practical deliberation). However, deliberation is also different from other ways of determining what to do: it is different from deciding that does not involve any conscious reasoning. This means that I cannot unintentionally deliberate18 but also that deliberation is a conscious activity that occupies the agent’s attention.19 Cases of negligence such as the ones described present the agent as thoughtless or careless: they do not present the resulting behaviour we account them responsible or blameworthy for as the result of a conscious process that would make such events the focus of attention. If the subject we are considering is, in this sense, “thoughtless”, it seems that these kinds of omissions clash with any reflective endorsement theory that is committed to see reflection as a deliberative process. However, as we have briefly seen above, the narrative model does not equate reflection with deliberation: indeed, we have seen how Velleman stresses that there are two kinds of reflection, one of which is nondeliberative. I want now to look at this claim in more detail and eventually argue that there are reasons to doubt that this conception of reflection is really compatible with the narrative model. This point will question the extent to which the metaphor of narration can succeed in accounting for agency, once we have ascribed two distinct functional roles to practical reasoning (one as deliberator and one as supervisor). We have seen how Velleman has considered that much of our behaviour results from a nondeliberative kind of reflection, and this claim has intuitive appeal since automaticity, not deliberative thinking, seems to be the rule rather than the exception. The self conceived as a narrator must be able to enact his autobiography through both. Inasmuch as narrativity is constitutive of agency, both of these functional roles must be ascribed to the narrator. In order to get to this conclusion, I will highlight a distinction between deliberative and perceptual reflection that so far has been only implicit: a distinction between active and passive in this specific context. I define



186

Chapter Twelve

deliberative reflection as active inasmuch as the deliberator actively intervenes through his decision in leading his behaviour some way or another. On the other hand, when it comes to perceptual reflection, the agent need not intervene: behaviour, as Velleman phrased it, is “under way” and as long as it “makes sense”, the agent needs not to actively play a role in directing it. In this sense, the supervisor is passive. We could say that it is theoretically active, but practically inactive, since it is a system of monitoring behaviour that is unfolding without intervention.20 Indeed, the very idea of perceptual reflection was introduced to make sense of the intuition that automaticity, not deliberative thinking, is the rule and not the exception. In this non-active position, the supervisor shapes the intelligence of the sub-personal driver. As pictured above, in its supervisory role, practical reasoning stays in the background (both of action and thought). Identifying the agent with the supervising intelligence means that it was there all along, just “letting” his body move. To sum up, a core characteristic of Velleman’s supervisor is its being passive (and claiming that there are different degrees of passivity or activity within perceptual reflection is not an objection to my choice of case, since the supervisor is passive in at least some relevant conditions). Given this distinction between active and passive, I want to argue that one core feature of narration is its being a form of activity. Pace Velleman, narration is not a theoretical stance that the author has towards some events. It implies the practical task of producing a story, engaging in some mental composition of a story. As an author, the narrator is active: he is not told, but rather he tells the story. There would be no story at all without at least this intervention. This is not to say that a narrative needs to contain all the details (given that narratives allow a certain degree of vagueness).21 Nevertheless, if a scenario or a framework of meaning needs to be built, the narrator is active in its composition. It really is the practical dimension of narrativity that allows the metaphor to present the agent as an author, grounding Velleman’s conception of action. Is saying that narration is always an activity merely an ad hoc stipulation? The discussion above served to consider how the metaphor of narration carries important implications in terms of the functions ascribed to the self. The resultant view, by assuming a parallel between the twofold model of agency and the twofold model of the self, and hence equating full agency and narrativity, gave us a picture of practical reasoning in which you need to find yourself intelligible as the character you take yourself to be. The way you succeed at satisfying this need is by giving expression to your thoughts, feelings, and personality through the things you do:



Responsibility and Narrative Agency

187

Here the verb giving expression describes a substantive activity. In ordinary actions you do not let your thoughts, feelings and personality come out through your body, as if they were seeping through your pores. Instead of acting out in this manner, you perform intelligible actions by selecting a coherent subset of your thoughts and feelings, composing a coherent expression of them and thus producing words and movements with a unified message.22

Velleman’s conception of action requires this characteristic of narration, in order to avoid making the self purely fictional and casting the agent in the merely contemplative position of a commentator. As we have seen, Velleman’s idea is that the agent’s self-conception thus feeds back into behaviour and it does so by means of actual agential control. It doesn’t seem, then, that the narrator can really take up both functional roles and this would compel us to conceive it as a deliberator. Considering only deliberated actions to be actions tout court is problematic. If so, the range of actions that the model is capable of accounting for is significantly restricted, in a way that makes the simple argument implausible: if our actions are just deliberated ones, claim a) that we are responsible for our actions seems just unsound. It certainly seems to be unable to include the wide range of cases of non-deliberated omissions, and it also clashes with our most common practices of praise and blame. Even accepting that reflection need not be identified with deliberation, a second point of concern is the role played by the idea of endorsement. I won’t be able to address this point in very much detail but I hope to at least sketch some reasons for concern. Negligent agents are careless and it would be contradictory to attribute endorsement in these cases, which by definition are failures of concern resulting in some harm. Endorsement seems to imply the commitment of the agent to some action via some form of conscious choice or judgement. These cases of omissions seem problematic because there is simply no conscious choosing involved that results in their happening. I believe that if we posited some antecedent point in which the conscious volition took place we would be misunderstanding the case.23 At no point do I take the decision to neglect my promise to my partner. Does the narrative model, in its conceiving of endorsement in terms of narrative coherence, reply to this objection? I don’t think it does. After receiving the unexpected news, I can be enacting a narrative coherent with my self-conception of a caring employee concerned with his job, but at no point am I judging that, all things considered, doing what I promised to do just wouldn’t fit within my



188

Chapter Twelve

narrative (because, for example, it would oblige me to abandon my task and hence fail to behave like a committed employee would). If we understand actions as behaviour that makes sense to me to perform, in the light of some self-conception, I must know what I am doing and why in acting. In the cases I am considering the agent is too self-absorbed, forgetful, distracted or overwhelmed by emotion, with the result that the omission we account him responsible for can be understood as a failure of attention. Failures of attention just do not qualify as actions, in a fullblown sense. The fact that I can, post hoc, build up a narrative that explains my behaviour and makes sense of it, does not mean that I was in fact enacting such narrative. There seems to be no room to interpret myself, while performing such behaviour, as a narrator living out his own autobiography. I believe that these considerations suffice as reasons not to accept the simple argument that sees the conditions for agency determining the conditions for responsibility. Let me stress how this would be particularly problematic for the narrative model and notice a further difficulty that seems to emerge from this discussion. As we have just seen, there seem to be just no way for cases of negligence to qualify as actions in the narrative model. Moreover it doesn’t seem that any thing we do which is “out of narrative” would qualify as action. If we were to restrict responsibility to actions, in the full-blown sense, we would have to say that we are not responsible for anything we cannot fit within our narratives. The idea seems to be reminiscent of Hume’s claim that we are not responsible for things we do “out of character”. Hume’s idea is largely considered a mistake24 and intuitively we consider the lie told by the usually honest person, the cruel remark proffered by the usually kind person etc. as actions that these agents are responsible for. Because cases of good people acting badly elude Hume’s model, the theory seems to account too few people as subjects of responsibility. This is even more so, if we consider the possibility that people behaving “out of narrative” are not responsible, inasmuch as they are not truly acting. Narrativity, as the narrative model conceives it, requires conscious control. I have argued elsewhere25 about the difficulties that this notion of conscious control brings along with it. For my present purposes, it is sufficient to stress that the range of actions possessing narrative coherence (and that indeed are performed in virtue of it), and over which we appear to have conscious control as narrators is rather limited.26 To sum up, I have presented a simple argument that established direct connections between agency and responsibility. I have not claimed that



Responsibility and Narrative Agency

189

Velleman himself has argued along these lines but I have tried to examine whether or not it is possible for the narrative model to follow this path. If true, it would make theories striving to understand conditions of agency extremely explanatory, since conditions for agency would determine the conditions for responsibility. However, simply assuming that we are “responsible for our actions” excludes the large category of omissions, in which we can identify acts of negligence that do not seem to be possibly explained in terms of reflective endorsement. In the whole previous discussion I tried to show why omissions of this sort cannot be considered as instances of full agency, according to the narrative model and to reflective endorsement theories in general. If the narrative model cannot account for omissions, the simple argument is not going to work. I have argued that it cannot and I also argued that the model seems bound to a deliberative conception of reflection. One could revise the simple argument and claim that we have at least identified the condition C, constitutive of full agency, and that, at least where full actions are concerned, C is also constitutive of responsibility. However, this conclusion seems unwarranted: once we have established that we are not just responsible for our actions, but also for behaviour that does not qualify as such (we could say for our “activities” following Velleman’s jargon), we cannot guarantee that element C plays any constitutive role in responsibility in general. We would need to argue for the need of two concepts of responsibility: responsibility as it is attributed to actions, and responsibility as it applies to behaviour that is not a full-blown instance of agency. Otherwise it is not clear that condition C is playing any significant role: it might as well be the case that some other feature shared by both these categories is in fact the key to responsibility.27 It does not seem that defining full-blown agency can actually have any direct bearing on responsibility. With this I don’t want to claim that the debate around the specifics of human agency is entirely isolated from the one on responsibility. If the simple argument is not defensible, in which direction should the reflective endorsement theorist look to account for responsibility? Investigating the conditions for agency does not by itself result in giving us the conditions for responsibility. However, one could argue that identifying such conditions can provide us with normative standards through which we can then consider the question of responsibility. Indeed it would seem that this view is implied in much work on reflective endorsement.28 I don’t have space here to devote any attention to exploring this Kantian background and its implications,29 but I can hint at the direction for further work that this move implies. The starting point is acknowledging that the agent’s failures of attention or



190

Chapter Twelve

control are not actions in the full-blown sense but they lead us to consider the agent as responsible nonetheless. How can we be responsible for activities? In the narrative model in particular it is clear that the motives that “play the role of the agent” are absent or inoperative. How can we account the agent responsible or blame him if he doesn’t figure or play a role in these acts? When we consider an agent responsible for the action we need to find a reference that connects the agent to the relevant behaviour. Any theory that wants to hold on to the relevant distinction between action and activities, while acknowledging the limits of a direct connection between agency and responsibility, needs to answer this sort of questions.

References Aristotle 1999, Nicomachean Ethics, Terence Irwin (ed.), Indianapolis: Hackett Publishing Bratman, M. 2007, Structures of Agency: Essays, Oxford: Oxford University Press Bucelli, Irene (2014), Conditions of Agency and the Role of Reflective Endorsement: Does it make sense to distinguish between “kinds” of agency?, PhD Thesis Dennett, D. 1991, Consciousness Explained, London: Penguin Books Fischer, M. and M. Ravizza 1993, Perspectives in Moral Responsibility, New York: Cornell University Press Frankfurt, H. 1988, “Freedom of the Will and the Concept of a Person”, in The Importance of What we Care About, Cambridge: Cambridge University Press Gallagher, S. 2012, “Multiple Aspects of Agency”, New Ideas in Psychology 30, 15 Hornsby, J. 2004, “Agency and Actions”, in Steward H. and J. Hyman (eds.), Agency and Action, Cambridge: Cambridge University Press Hutto, D. 2007, Narrative and Understanding Persons, Cambridge: Cambridge University Press Korsgaard, C. 1996, The Sources of Normativity, Cambridge: Cambridge University Press —. 2009, Self-Constitution: Agency, Identity, and Integrity, Oxford: Oxford University Press Peacocke, C. 1998, Conscious Attitudes, Attention and Self-Knowledge, in Wright C., B. Smith and C. MacDonald (eds.), Knowing Our Own Minds, Oxford: Oxford University Press Ricoeur, P. 1988, Time and Narrative, Volumes 1-3, trans. B. Kathleen



Responsibility and Narrative Agency

191

and D. Pellauer, Chicago and London: University of Chicago Press Sher, G. 2006, In Praise of Blame, Oxford: Oxford University Press —. 2009, Who Knew?, Oxford: Oxford University Press Shoemaker, S. 1996, The First-Person Perspective and Other Essays, Cambridge: Cambridge University Press Velleman, D. 2000, The Possibility of Practical Reason, University of Michigan: Oxford University Press —. 2006, Self to Self: Selected Essays, Cambridge: Cambridge University Press —. 2007a, Practical Reflection, Stanford: CSLI Publications, (1989) —. 2007b, “Reply to Catriona Mackenzie”, Philosophical Explorations: An International Journal for the Philosophy of Mind and Action 10 (3), 283 —. 2008, “Bodies, Selves”, American Imago 65 (3), 405

Notes  1

For important more or less recent examples see Frankfurt (1988); Bratman (2007); Korsgaard (2009); Velleman (2007a). 2 Velleman himself recognized gaps still in need of filling in his reply to Catriona McKenzie (cf. Velleman 2007b, p, 284) and hence the attempt to provide a unified picture is welcome. 3 Responsibility is not an issue that Velleman’s work has focused on much, his primary concerns revolving around the structure of practical reasoning and the phenomenology of action. That the theory has important consequences for responsibility and could indeed be considered foundational for it is however often implied. cf. Velleman (2007a, p. 144): “I think that the nonmoral action concepts and the phenomenology of action deserve attention in their own right. My view is that explaining what is like to be an agent, how it’s possible to make a decision, are just as interesting as finding room in the world for punishment and blame. And even those who disagree with me on this score should consider that the problem of moral responsibility owes its very survival to these other, related problems”. 4 Velleman’s can be considered an attempt to improve on Davidson’s causal model of action. While it is possible to see that Velleman would share the worry that Davidson’s model “leaves the agent out” (Frankfurt 1988, Hornsby 2004), his idea is to hold on to the reductive enterprise while improving on it, rather than totally dismiss it. Hence Velleman’s explanation stresses the interplay of the same psychological elements considered by Davidson and strives to ascribe to some of it the role of the agent. 5 Velleman (2007a, p. 168). 6 Velleman (2006, p. 218).



192

Chapter Twelve



7 Velleman considers someone who is governed by his own self-conception as selfgoverned and autonomous, cf. Velleman (2007a, p. 173). I won’t touch on autonomy and its place within the model, nor on its relation to responsibility. 8 Hence the extensive literature on the topic. The range of philosophers interested in the role of narratives crosses the boundaries of continental and analytic philosophy, being an interest of people working in phenomenology as well as action theory of theory of mind. Cf. Hutto (2007), Dennett (1991), Gallagher (2012), Ricoeur (1988). 9 For an insightful criticism of Frankfurt’s position and its relation to Velleman’s model see “Identification and Identity” in Velleman (2006). 10 Velleman (2007b, p. 284). 11 For example, in Korsgaard’s, to mention one. I discuss her position in detail in Bucelli (2014). Coherence seems at least less demanding a test for the endorsement than Korsgaard’s own proposal, which ascribes a key role to the Kantian categorical imperative as a test for agency, cf. Korsgaard (2009). 12 cf. “The Centered Self” in Velleman (2006, p. 281). For the distinction between the actual occurring thoughts the agent engages in and the logic of practical reasoning cf. Velleman (2007a, pp. 88-89; 106-107). For the characteristics of narratives that allow the model to function through these two different kinds of reflection cf. Bucelli (2014). 13 1110a – 1113b3. 14 There are different formulations of this condition and so we can find it as “control condition”, as a “freedom-relative condition”, “voluntariness condition” etc. In defining it as a “volitional condition” I am not going to refer to any in particular and I just want to stress its concern with the agent’s will rather than knowledge. Cf. Fischer and Ravizza (1993), Sher (2009). 15 Aristotle, ivi. 16 An objection which I will not discuss would stress how the commitment of the narrative model to the standard causal account of action makes it impossible for the theory to account for actions that are not ‘positive performances’. For a thorough discussion of the limits of the standard story of action, see Hornsby (2004). 17 I do not discuss here the cases in which omissions are the result of the agent’s deliberated decision, since they do not present any particular problem for any view of reflective endorsement. 18 See Shoemaker (1996, p. 28). 19 See Peacocke (1998, pp. 65-67). 20 Notice that there might be degrees of activities and passivity but, inasmuch as the supervisor is passive in at least some relevant conditions, this is not an objection to the following point. Cf. Bucelli (2014). 21 For the discussion of the various features of narrativity see Bucelli (2014). 22 Velleman (2008, p. 422). 23 I am not going further in exploring this aspect, since I will concentrate on the narrative model. However, surely some cases could present the agent deciding at some point, for example, to procrastinate, saying to himself “I will call my partner



Responsibility and Narrative Agency

193

 later to tell her I won’t be able to do the shopping”. I don’t think that all cases of negligence will present a similar stage in the agent’s thought and taking all cases of negligence as voluntary at some level would not match our experience of them. 24 See for example, Sher (2006). Here Sher gives a deep analysis of different ways to interpret Hume’s claim, connected to blame rather than to responsibility. For the present purposes, and exactly because my discussion concerns responsibility, it is sufficient for me to consider that an agent is responsible for manifesting, in some action, the corresponding vice. 25 Bucelli (2014). 26 However, it is important to notice that in the narrative model, unlike in Hume’s, it is possible to account for behaviour that the agents display only once. As long as the act makes sense to the agent in the light of whatever reason he can identify, there is no reason why the good deed done by the malicious person, or the greedy act done by the generous one, cannot make sense in the circumstances and cannot be coherent with these agent’s self-conceptions. As long as these acts count as actions there seem to be no problem in attributing responsibility. 27 I am not denying here that one could attempt to lead the discussion to establish these two concepts. I am suggesting that, at this stage, there is no warrant that this strategy would be more adequate than identifying as constitutive of responsibility some other feature that actions and activities have in common. 28 This seems to be Korsgaard’s strategy: see Korsgaard (1996). 29 For example, I can only acknowledge that working in this direction requires exploration of the connection between responsibility, obligation and the normative standards provided by this analysis of agency. It is important to stress that Velleman himself suggests that his views on responsibility would take this direction, and I believe this is so because of the broadly Kantian picture he provides. Cf. Velleman (2000, p. n127): “I have the obligation to be vigilant and keep my voice down, no matter what is causing it to rise. Responsibility arises from my having failed to prevent or control it, rather than from having initiated it”.



CHAPTER THIRTEEN EMPATHY AND CONCERN FOR OTHERS SARAH SONGHORIAN

1. Introduction There is a huge variety of ways in which the term “empathy” has been used in the literature (Goldie and Coplan 2011; Preston and de Waal 2002). The aim of this work will be to try to disentangle the confusion by providing a few internal distinctions and to understand whether or not there is actually a role for empathy in a moral theory. One first challenge, related to the first aim of this work, is that of considering some of the elements that have been provided in the literature as part of—or defining—“empathy”. Some of the questions that can arise when trying to determine the meaning of this concept, impose deep normative questions themselves. What is the object of empathy (Roughley 2013)? Is it someone else’s feelings, what that person will be feeling, what I would be feeling if I were in his shoes, or what he should be feeling given the situation he is facing? Is empathy affective or cognitive (Lamm, Batson and Decety 2007; Shamay-Tsoory, Aharon-Peretz and Perry 2009)? Is it solely an emotional response or is it also possible to have a cognitive component? If the latter is the most plausible answer, then what do we know about the relations between these two components? Do they work in parallel or in sequence? Do they interfere with each other? Does it entail specific behavior (Batson et al. 2003)? Is it sufficient to have an empathetic response to behave in a certain—moral— way or are there any other steps required? What is the relation between empathetic concern and personal distress (Slote 2010)? The difference between these two ideas is that, in the former, the focus is on the other person—the one with whom I am

Empathy and Concern for Others

195

empathizing—while, in the latter, the focus is on myself, on the way the other person’s situation makes me feel. For instance, if I see an ill person I can either be somehow moved to help her because of the understanding of how she might feel or I can keep a certain distance so as not to be infected—this example, of course, has a very clear behavioral pattern, which is not internal to empathy itself, as we will see, but it is useful to make the distinction clear. What does it mean to have an “appropriate” empathetic response (Roughley 2013)? In the literature we can find that we have an appropriate, a good, or a congruent empathetic response. This means that not only in our daily usage of the concept, but also in the literature, there is somehow the feeling of a relation between empathy and normativity— since all these are normative concepts. This last point leads to the second aim of this work, which is that of understanding whether such a feeling is somehow well-grounded or not: i.e. is there a relation between empathy and normativity? What kind of relation is it? Is empathy necessary for a normative theory? Is the relation unidirectional or bidirectional: that is, are they reciprocal in their influence or not? In this work I would like to provide a framework to try and answer these questions.

2. Different concepts of empathy As we have just noticed, there are several different aspects within the concept of “empathy” that have been analyzed and discussed in the literature. The aim of this section will be to provide an answer to the above mentioned questions so that a univocal definition of basic empathy will emerge. This definition will be a very narrow one, compared to some of those discussed in the literature (De Vignemont and Singer 2006; Gallese 2001, 2003), so that only the fundamental elements of empathy will remain, avoiding confusion with other related concepts. Describing empathy as a very basic capability will lead us not only to understand better its relations with more complex forms of fellow-feeling, but also to identify the passage from what we will consider as an empathic response to concern as a passage from the description of an embedded capacity to a normative understanding of how we should engage in our relations to others. To answer this and all the other questions we need to assess which level of empathic response are we interested in, since there has been a wide use of it in different contexts (Coplan and Goldie 2011, Introduction). The most basic form of empathy is what we consider as an

196

Chapter Thirteen

empathic response in toddlers (Eisenberg and Strayer 1987; Meltzoff and Moore 1977, 1983; Pfeifer et al. 2008; Simner 1971). The idea is that analyzing the elements of an empathic response in infants will tell us about the most basic and embedded features of empathy, that are not so distinguishable in adults. The basic form in toddlers is characterized by the emotional identification with the primary care-giver (Scheler 1923). From the infants’ point of view there is no distinction between themselves and their care-givers and the emotions are somehow shared. This is the most basic form of affective sharing. It is the extreme example of contagion. In an emotional contagion episode the person is not conscious about the source of the emotion she is feeling, but she is still conscious about the self/other distinction (Scheler 1923, pp. 12–15). In emotional identification this is not so. So, trying to focus on the minimum form of empathy that we can recognize in infants, emotional identification presents itself as such a minimum. It is with the aim of finding this minimum form that I will try to provide an answer to the questions presented above and to see how from such a basic form of affective sharing we can—if we do—move to a prescriptive account of how we should behave towards others.

2.1. The object of empathy The object’s individuation for this basic form of empathy cannot be guided by the cognitive ability of understanding counterfactuals or hypothetic situations, since toddlers cannot do that in their very first days of life. So the primary object of empathy is not what that person will be feeling, or what I would be feeling if I were in his shoes (see Batson, Early and Salvarini 1997) or what he should be feeling given the situation he is facing. It is simply someone else’s feeling at a certain time. This is not to say that the other objects I mentioned are never possible objects of shared emotions, but just that they cannot be considered as the primary ones. For any of those, we need to be conscious of the self/other distinction that is not explicit in infants. For instance, I cannot put myself in someone else’s shoes if I am not able to distinguish myself from others. In more complex and cognitive forms of empathy all of the abovementioned objects do exist, showing also the strong connection existing between empathy and normativity (i.e. what she should be feeling).

Empathy and Concern for Others

197

2.2. Emotional and cognitive components This last remark leads directly to the second question I mentioned above. As my interest is in the ability that infants have, I would say that the most basic and primeval form of empathy is an emotional response, which of course does not imply that a cognitive component is not present in subsequent and more complex forms of that ability. Adults’ empathy has both an emotional and a cognitive component, but that is not so for infants. As far as the relation between these two components is concerned, there is little knowledge and a huge debate in the literature to try and understand how and to what extent they interact or interfere with each other. Though it is a very interesting aspect, it is not my primary interest here for defining toddlers’ form of empathy.

2.3. Empathy and behavior Depending on how empathy is defined, there is a difference also in the relationship between this capacity and behavior. If empathy is construed as a complex ability with no internal differentiation between empathy, sympathy, contagion, concern, and so on, it can be said that empathy entails pro-social behaviors. That is not, though, the perspective I would like to take. I believe the distinction between all these concepts will prove helpful to understand how we can pass from simple emotional sharing to moral behavior. Empathy as the basic form of emotional identification between the toddler and her care-giver does not entail per se moral behavior. Further steps, first of all the acquisition of the self/other distinction and of the cognitive component of empathy, are needed in order to reach moral behavior. Furthermore, distinguishing an empathic response from a moral behavior makes it possible to understand perverted forms of empathy like sadism. A sadist is not someone who is not capable of feeling empathy, of sharing someone else’s feelings, he is just someone who has not reached the moral level. Instead of caring for others, he rejoices in their pains and feels sorry for their joys—we will see something more about sadism in what follows. To achieve a pro-social behavior much more than basic empathy is needed. There is certainly a connection between the two, but it is not a necessary, nor sufficient one. We can have empathy and never reach the moral level, like the sadist, and, even if we have it, we are not necessarily directly driven to moral behavior. We will see hereunder a possible

198

Chapter Thirteen

conceptual derivation from empathy to pro-social behavior: through sympathy, respect, concern for others, and care. This derivation is just one possibility and it is not necessarily the only way to reach moral behavior— it might be the case that other patterns to pro-social behavior are still possible.

2.4. Self-oriented and other-oriented empathy In the basic form of empathy I am considering here, the self/other distinction does not exist in the subject. At this level, it is not possible to take a stand in such a distinction. The infant does not direct his attention to someone else’s feelings nor to his own reaction, since he just feels what the other feels in an emotional identification. The distinction between self- and other-oriented empathy, that has been extremely relevant in the literature, is however useful to understand more complex forms of empathy and to distinguish between two styles of reaction. In real life situations we usually react somehow with both strategies, or rather with a combination of the two. But it is still possible to draw a conceptual distinction between them and see to which one every individual is more prone. Such a proneness characterizes a style of reaction in front of the emotional life of others that will prove useful to understand further steps to pro-social behavior. This distinction is not, strictly speaking, part of basic empathy. Basic empathy has to do solely with the fact of sharing someone’s feeling, while self- and other-orientedness deals with the particular and personal style by which every person reacts to such a sharing of emotions. First of all, the self/other distinction is needed to reach such a differential reaction. Second, I mean reaction in a minimal way: it is not a behavior, it is just an emotional reaction, the direction that each of us gives to that shared emotion. As I have already said, the self-oriented reaction towards someone else’s emotion focuses on the self, on how I will feel if I were in such a situation. The other-oriented reaction, on the contrary, focuses on how that particular person with her own specificities might feel and on trying to understand such a feeling. From this kind of differential reaction different behavioral patterns can of course be drawn. What is relevant here is just to see that in more complex forms of empathy, there can be a difference in the style through which people understand others, either focusing on one’s own emotions or on the other’s. As we have noticed, we usually use a combination of these two strategies going back and forth from ourselves to

Empathy and Concern for Others

199

others, but the conceptual distinction is still clear. Where one strategy is strongly prevalent it can have consequences in behavior: if someone is strongly more self-oriented in his emotional reactions, he will manifest more often an avoidance behavior and he will turn the worries or pains of others into his own. He will be concerned about himself, even though the source of that concern is not his own self. On the other hand, if someone is more prone to other-oriented reactions, they can lead to a supererogative altruism neglecting the self. As we have noticed, the self-/other-oriented reaction is possible only where empathy is more complex and cognitive and where the self/other distinction is present. It is not possible in toddlers. Furthermore, with this distinction we have seen a first glance of how empathy interacts with behavior. Such a link is not a direct one, but it is still interesting to notice how different styles of emotional reactions can, when not balanced, lead to specific styles of behavior.

2.5. Appropriate empathic responses As I have already mentioned, in the literature we can find that we have an appropriate, a good, or a congruent empathetic response. This means that not only in our daily usage of the concept, but also in the pertaining literature, there is somehow the feeling of a relation between empathy and normativity—since all these are normative concepts. Obviously, this has nothing to do with basic empathy: a toddler’s emotional identification cannot be good nor bad, it is just a shared emotion. When the idea of an empathic response being appropriate, good, or congruent is used, the focus is on more complex forms of empathy and, in particular, on the external expression of emotional states. It has to do, somehow, with a cognitive mediation of the strength of an emotion. For instance, it is not appropriate to express more pain than the subject of such pain. This kind of consideration, though, has more to do with the virtue of self-regulation rather than with empathy as an ability itself. In this respect, although there is a connection between empathy and normativity—though intended as correctness conditions and not as a moral ought—, it is not a direct link that connects the two and it cannot apply to basic empathy. By means of self-regulation, we can still learn what is appropriate and so modify accordingly our way of expressing empathy. The ability itself and its external expression are two different, thus related, things. Concerning the latter, normativity can certainly influence our expression of empathy, not modifying empathy itself, but modulating its expression. So that the

200

Chapter Thirteen

relation between empathy and normativity can be considered bidirectional in this respect: they have reciprocal influences.

2.6. Basic empathy As a first approximation, from what we have just said, we can say that there is basic empathy if: (i) one is in an affective state; (ii) this state is isomorphic to another person’s affective state; (iii) this state is elicited by the observation […] of another person’s affective state (De Vignemont and Singer 2006, p. 435).

And (iv) the subject is not conscious of the difference between subjects, between himself and others. This last condition (iv) is radically different from the one present in De Vignemont and Singer’s work, that is: (iv) one knows that the other person is the source of one’s own affective state (De Vignemont and Singer 2006, p. 435).

The difference between the definition provided by De Vignemont and Singer and my own is just that I have focused solely on basic empathy as the ability that we can identify in toddlers, while they conceive it as a more complex phenomenon as we see it in adults. Summing up what I consider basic empathy, we can say that it is the capacity we can identify in toddlers when sharing emotional states with their primary care-givers. The object of such a basic form of empathy, that is emotional identification, is simply the emotional state itself, with no conceptualization of how the person should feel, how I would feel, and so on. As my interest here is in such a primitive and minimal form of shared emotions, I would not consider the cognitive component, that is obviously extremely relevant, but not primitive. The link between basic empathy and behavior is not a direct one, as I have already mentioned and as we will see in hereunder. Furthermore, the self/other distinction which is pivotal for more cognitive and complex forms of sharing, is not present in the basic form of empathy. Finally, like the link between empathy and behavior, the link between empathy and normativity is also an indirect one. In both cases we can find several authors linking these two concepts together, by means of proneness to pro-social behavior in the former and by means of the

Empathy and Concern for Others

201

concept of an appropriate empathetic response in the latter, my idea is that these attempts show their deep relation, but that is not justified to believe those links are direct. Conceptually speaking, there are several things between empathy and behavior just as between empathy and normativity.

3. From empathy to sympathy One thing that can be considered between empathy and behavior is sympathy itself (Eisenberg 2000, Escalas and Stern 2003). First of all, we need to distinguish between the concepts of “empathy” and “sympathy”. x “em-pathy” (‫݋‬ȝʌȐșİȚĮ): to feel within oneself someone else’s feelings. x “sym-pathy” (ıȣȝʌȐșİȚĮ): to feel with someone else. To put oneself into someone else’s situation and to judge it. I will try to suggest a step forward, consisting in distancing sympathy from a purely emotional ability. Adam Smith, in his The Theory of Moral Sentiments (1759), pointed out the difference between “sympathy” and “compassion”, which are often considered synonyms, notwithstanding the evident difference that the etymological definition suggests. Pity and compassion are words appropriated to signify our fellow-feeling with the sorrow of others, sympathy […] may […] be made use of to denote our fellow-feeling with any passion whatever (Smith 1759, p. 49, italics mine).

Sympathy, like empathy, is completely indifferent to the content of the feeling, while compassion is not. What empathy and sympathy have as their object has nothing to do with the type or intensity of someone else’s feelings, but with the fact that it is a feeling. The ways in which they enable us to experience someone else’s emotions, however, are completely different. Empathy deals with an automatic and mostly unconscious ability to experience someone else’s emotional reactions—basic empathy, in particular, deals with infants ability to identify themselves with their caregivers’ emotional life; sympathy, on the other hand, is a more complex metaethical phenomenon.

202

Chapter Thirteen

Sympathy is the basis of the moral sentiments for Smith because “to sympathize with” means “to align with the estimation of right or wrong based on fellow feeling” (Agosta 2011, p. 4, italics mine).

Sympathy, in my view, is a mechanism that balances different beliefs and emotions. The relevance of Agosta’s point on Smith is the fact that he introduces an evaluative element within the concept of “sympathy”. It is not merely a matter of basic ability to enter into a relation with cospecifics, but it is a way to judge behaviors and characters and decide which considerations we should take into account in order to act morally. Smith’s subtitle make this role of sympathy particularly clear: An Essay Towards an Analysis of the Principles by which Men Naturally Judge Concerning the Conduct and Character, first of their Neighbours, and Afterwards of Themselves.

Sympathy represents such a principle of moral sentiments. It can be seen as a metaethical criterion to judge and decide, that is, the point of view we have to adopt in order to pronounce any moral judgment. But how exactly does this principle work? […] though sympathy is very properly said to arise from an imaginary change of situations with the person principally concerned, yet this imaginary change is not supposed to happen to me in my own person and character, but in that of the person with whom I sympathize. When I condole with you for the loss of your only son, in order to enter into your grief, I do not consider what I, a person of such a character and profession, should suffer, if I had a son, and if that son was unfortunately to die; but I consider what I should suffer if I was really you; and I not only change circumstances, but I change persons and characters. My grief, therefore, is entirely upon your account, and not in the least upon my own. It is not, therefore, in the least selfish (Smith 1759, pp. 501–502, italics mine). The ideal spectator runs a cognitive simulation in which one may indeed begin with one’s own characteristics as input, but quarantines one’s own peculiarities in favor of those of the other (Agosta 2011, p. 5, italics mine).

Sympathy, as opposed to empathy, arises when the subject puts himself into a particular position in relation to the state of the other he is considering. Sympathy does not imply just the mirroring of someone else’s emotions, it is a much more complicated phenomenon implying both an emotional reaction, probably empathic and most of the time unconscious, and the fact of taking a certain position in order to be able to

Empathy and Concern for Others

203

express a judgment. The subject imaginatively changes his position with that of the other person, he does not simply imagine himself in that position, but he considers all the peculiarities of the person he is trying to understand as if they were his own. The second element I would like to consider here is the difference between a) the possibility, intrinsic in an empathic reaction to be selforiented; and b) the more complex and pondered ability to assume a thirdperson standpoint, i.e. the one of the so-called ideal spectator, who is capable of judging fairly both before and after the performance of the action, thus allowing moral judgment on the one hand and moral motivation on the other. While empathy can be, as we have seen, both selfand other-oriented, sympathy cannot be self-oriented, as Adam Smith says: sympathy cannot be “in the least selfish”. To feel someone’s emotion per se does not imply certain active behaviors in favor of him. The relation between sympathy and acting in favor of others is simply closer than the one linking beneficent behavior to empathy. In the phenomenon of sympathy I try to integrate both the emotional reaction and some information about the subject that is undergoing such a situation, since the aim of sympathizing is that of understanding how that particular person, with that particular character, should have behaved. Not only the point of view is different, but different are also the sources of the information that the two phenomena elaborate: empathy concerns only emotive reactions, while sympathy needs to integrate different kind of information, i.e. emotive responses, knowledge of the person and of her character, comprehension of the situation, its features and—as far as possible—its consequences (see also Batson et al. 2003). I believe sympathy, together with the concept of the ideal spectator, constitutes the principle Adam Smith had in mind for establishing a criterion to judge and to direct moral behavior (Smith 1759, p. 11). In this sense, they are not normative concepts in the strong meaning of normativity, but they are surely metaethical ones. If the definitions of “empathy”, “sympathy”, and the “ideal spectator” prove useful and consistent, there will not be an actual “is/ought passage” (Celano 1994), since “sympathy” will be defined as a metaethical mechanism and not as a normative concept itself. The fact of deriving metaethical implications from factual premises does not give rise to so many problems as the derivation of normative conclusions from the same premises, as Joyce points out: Even if there were an a priori prohibition on deriving evaluative conclusions from factual premises, this need not stand in the way of

204

Chapter Thirteen

metaethical implications being drawn from factual premises, for a metaethical claim is not an ethical “ought” claim; it is more likely to be a claim about how we use the word “ought” in ethical discourse—which is a perfectly empirical matter (Joyce 2008, p. 371).

So, even accepting such an a priori prohibition—that I do not accept— the shift from the description of a neural mechanism (Botvinick et al. 2005, Carr et al. 2003) to metaethical implications can be itself interesting and can be complicated by the fact that metaethics does not only concern our usage of the ethical language, but sometimes provides us with a criterion for assessing the correctness of our usage of such language. In this latter meaning, it has something to do with prescribing. In so far as some metaethicists offer prescriptions about how the word “ought” ought to be used, metaethics sometimes steps beyond the descriptive. Even in such cases, however, metaethicists are still not pushing ethical “ought” claims (Joyce 2008, p. 371).

As far as “metaethics sometimes steps beyond the descriptive” is concerned, I believe the relation between the two concepts analyzed here can provide some further insight and can represent a mediate way to reach a solution. It is not a matter of deriving a normative conclusion from a pure description of how our brain is, but to derive it by means of a metaethical criterion that takes into account both how we are actually wired and the way in which we use the evaluative language. Hume’s concern about the absence of any explanation or justification for the passage from sentences containing “is” and “is not” to propositions connected with “ought” or “ought not” (Hume 1969, pp. 469–470), has been widely misrepresented, for instance in Joshua Greene’s works (Greene 2003, 2009, Greene and Haidt 2002, Greene et al. 2001). Besides the evident relevance of his experiments concerning human beings’ actual capacity to produce moral judgments, he drew rash conclusions concerning the relevance of such an analysis for the “is/ought question” debate: If these empirical claims are true, they may have normative implications, casting doubt on deontology as a school of normative moral thought (Greene 2003, p. 36).

The misconception derives from the fact of disregarding the metaethical level. The denial of such a level of analysis counts as an unjustified passage from description to normativity. For this reason, I

Empathy and Concern for Others

205

believe the concepts of “empathy” and “sympathy” could provide a more detailed picture of the relations between “is” and “ought” since “sympathy” will not count as a normative concept, but as a metaethical one.

4. From sympathy to respect, concern, and care The last step I will take in this attempt to link basic empathy to prosocial behavior is that of briefly considering the normative level. Since sympathy is just a metaethical criterion, it does not prescribe anything to us. What guides our behavior is the fact of applying sympathy to our relation to others, thus leading to three kinds of behavioral attitudes: respect, concern, and care. These behavioral attitudes are what is desirable as a guide of our interaction with others—compared to the alternative possibility I will explore—and they are a consequence of applying a sympathetic criterion. These three attitudes, that cannot be deeply analyzed here, do not constitute the entire normative level, since other things are also relevant to it and do not concern the relation with others, but respect, concern, and care are the three normative attitudes that concern how we ought to interact with others. Respect deals with the fact of respecting other people’s dignity (for a deeper analysis of “respect” see Kant 1785, 1788, 1797, Darwall 1977, Wood 1999, Rawls 2000) and their free choices about their own lives understanding them by means of sympathy and the ideal spectator. Respect has often to do with not interfering with their value choices, as long as they seem to be legitimate after having analyzed them through sympathy. Concern is an attentional attitude towards others and the consequences of our own and other people’s behavior towards them. It can lead to care as far as care regards the interest for the well-being of others (Gilligan 1982, Held 1993, 1995, 2006, Kittay and Feder 2002). The link between sympathy and these three attitudes is characterized by the fact that, in order to have them, we need to undergo a sympathetic response, trying to put ourselves into someone else’s shoes. In this sense, there is a passage from a metaethical criterion for judging and acting to proper forms of considering others in our pro-social behavior.

206

Chapter Thirteen

5. Conclusion: an “is/ought question” There is obviously much more to say about the concepts of respect, concern, and care, than the observations made in the preceding paragraph, but that was not the aim of this work. My purpose here has been to show the complexity underlying the concepts of empathy, sympathy, and concern and how these concepts can help us understand the passage from descriptive premises to normative conclusions. I have tried to operate some distinctions within what is usually considered, broadly speaking, as empathy in the literature, so that a more narrow concept of our basic ability can emerge. Thus, from the very basic form of empathy that infants possess, I have tried to see how we can come to a more cognitive and complex ability: sympathy as a metaethical mechanism enabling us to judge a posteriori both other people’s behavior and our own and to decide a priori—by means of the imaginary projection (and/or simulation) of an ideal spectator—whether this conduct is appropriate or not (Gordon 1995). Is this an overall normative theory? Of course it is not, but the combination of the ability to put ourselves in someone else’s situation and in that of the ideal spectator—that directs towards ourselves the same kind of considerations we apply to others when judging them from a sympathetic point of view—can constitute the criterion to balance beliefs, emotions, principles and so on. It does not tell us how we should behave in a strict sense, but it shows how we judge behavior a posteriori and how we can educate ourselves to foresee that kind of judgment and make it relevant for our own decision-making. This issue has to do with a different level of description, compared to the one involved in the description of neural areas, but it is still description unless we attribute to it an evaluative element. If we do so, if we show that sympathy is somehow a good, or the best, way to judge and decide in morally-laden situations, then the step from “is” to “ought” is less radical. The idea is that the fact that we ordinarily have—save in cases of impairment (Baron-Cohen 2011, Dapretto et al. 2006) or of reduced activation—an “empathic circuit” in our brain (see also Dalgleish 2004) does not imply that we will necessarily endorse a sympathetic criterion of judgment and decision-making. To endorse such a criterion will be the best “choice”—it is obviously not a choice that a subject completely has in a conscious way since we are partly the result of the interaction between deterministic factors, but it is possible to educate ourselves to it. But, if it is a “choice”, there has to be an alternative. My idea is that different degrees of sadism are the alternative. At least conceptually, sadism is the

Empathy and Concern for Others

207

opposite of a sympathetic attitude: while the latter makes you feel a certain emotion as if you were in the other subject’s situation, you feel it with him, exactly as he feels it; the former, on the other hand, leads you to rejoice in other people’s pain, the other person’s pain causes your own joy. In this sense, if sympathy ought to be the criterion to judge and decide morally (as opposed to a sadistic attitude), the transition from “is” to “ought” is twofold: the first move is a sort of shift within the descriptive domain, from the neural description to the metaethical one; the second is the move from sympathy as the criterion we actually use to a normative theory prescribing pro-social attitudes—such as respect, concern, and care—as normative. Given the natural ability we have been embedded with, our education, culture, and environment in general can make us direct this capacity to one of the two possible attitudes: either sympathy or sadism. Obviously it is not an all-or-nothing alternative and it is not a choice than can be made once and for all: we are always merged in a continuum and we will face every situation with a certain degree of sympathy and sadism. The metaethical issue is, therefore, related to the concept of sympathy and shows the shift from a descriptive level to a different one that has an evaluative element, i.e. the fact that sympathy is desirable, it ought to be the direction that we given to our ability to empathize. This sense of “ought” is quite broad, and it is not the sense we want for a normative ethical theory. A normative theory requires a connection with the metaethical concept of sympathy and needs to derive from it. But this connection will not be enough, since the meaning of “ought” we desire for a moral theory is more binding than the one that can be ensured by “sympathy” alone. My idea is that such an aim could be obtained by means of the concepts of respect, concern, and care, I have briefly analyzed. So, on this account, empathy can provide the natural and embedded basis for our ability to understand those who are similar to us, but it will not be moral until we somehow decide that we are going to take the emotions of others as relevant for our moral judgment and decisionmaking and to give a precise direction to this consideration. As long as it is just an instinctive reaction, it is far from a moral theory telling us how we should behave towards others. The innate capacity is, thus, necessary for the sympathetic consideration of other people, but it is not a sufficient condition to make that relationship moral: having such a mechanism is totally compatible with making immoral decisions and judgments—it is, for example, totally compatible with some degree of sadism. If we lack such a automatic

208

Chapter Thirteen

mechanism for sharing other people’s emotions, it is much more complicated to understand them, though it is not impossible by means of a more complex and mentalizing mechanism that cannot provide an immediate and emotional sharing, but an intellectual comprehension through the attribution of mental states to others. So, even though the link between empathy as a capacity and sympathy is indeed necessary, it does not follow that the link between empathy and pro-social behavior is necessary too. There are other ways to perform pro-social behavior that do not entail empathy (Slote 2010, Prinz 2007). The neural circuits give us a possibility. Normativity—as far as the relations to others are concerned—is grounded in our capacity to empathize with others. But the fact of having such a circuit does not mean that it is necessary to develop it in a desirable way: the realization of oneself as a sympathetic and caring being is totally open to this concept of possibility.1

References Agosta, L. 2011, “Empathy and Sympathy in Ethics”, In Internet Encyclopedia of Philosophy, Accessed May 2012, http://www.iep.utm.edu/emp-symp/ Avenanti, A., D. Bueti, G. Galati and S.M. Aglioti 2005, “Transcranial Magnetic Stimulation Highlights the Sensorimotor Side of Empathy for Pain”, Nature Neuroscience 8, 955 Baron-Cohen, S. 2011, The Science of Evil: On Empathy and the Origins of Cruelty, New York: Basic Books Batson, C.D., S. Early and G. Salvarini 1997, “Perspective Taking: Imagining how Another Feels versus Imagining how You Would Feel”, Personality and Social Personality Bulletin 23, 751 Batson, C. D., D.A. Lishner, A. Carpenter, L. Dulin, S. Harjusola-Wevv, E.L. Stocks, S. Gale, O. Hassan, and B. Sampat 2003, “‘…As you Would Have Them Do Unto You’: Does Imagining Yourself in the Other's Place Stimulate Moral Action?”, Personality and Social Psychology Bulletin 29, 1190 Botvinick, M., A.P. Jha, L.M. Bylsma, S.A. Fabian, P.E. Solomon, and K.M. Prkachin 2005, “Viewing Facial Expressions of Pain Engages Cortical Areas Involved in the Direct Experience of Pain”, Neuroimage 25, 312 Carr, L., M. Iacoboni, M.C. Dubeau, J. C. Mazziotta, and G.L. Lenzi 2003, “Neural Mechanisms of Empathy in Humans: A Relay from Neural

Empathy and Concern for Others

209

Systems for Imitation to Limbic Areas”, Proceedings of the National Academy of Sciences of the United States of America 100 (9), 5497 Celano, B. 1994, Dialettica della Giustificazione Pratica. Saggio sulla Legge di Hume, Torino: Giappichelli Cremaschi, S. 2005, “Adam Smith Antiutilitarista”, La società degli individui 24/VIII/ 3, 1 Dalgleish, T. 2004, “The Emotional Brain”, Nature Reviews Neuroscience 5, 583 Dapretto, M., M.S. Davies, J.H. Pfeifer, A.A. Scott, M. Sigman, S. Y.Bookheimer, and M. Iacoboni 2006, “Understanding Emotions in Others: Mirror Neuron Dysfunction in Children with Autism Spectrum Disorders”, Nature Neuroscience 9, 28 Darwall, S.L. 1977, “Two Kinds of Respect”, Ethics 88 (1), 36 De Vignemont, F. and T. Singer 2006, “The Empathic Brain: How, When and Why?”, TRENDS in Cognitive Sciences 10 (10), 435 Eisenberg, N. 2000, “Empathy and Sympathy”, In Handbook of Emotions, edited by Lewis M., J.M. Haviland-Jones and L. Feldman Barrett, New York/London: Guilford Press, 677 Eisenberg, N. and J. Strayer 1987, “Critical Issues in the Study of Empathy”, in Empathy and Its Development, edited by Eisenberg, N. and J. Strayer 3–16, Cambridge: Cambridge University Press Eisenberg, N. and J. Strayer 1987, Empathy and Its Development, Cambridge: Cambridge University Press Escalas, J. E. and B.B. Stern 2003, “Sympathy and Empathy: Emotional Responses to Advertising Dramas”, Journal of Consumer Research 29 (4), 566 Gallese, V. 2001, “The ‘Shared Manifold’ Hypothesis: From Mirror Neurons to Empathy”, Journal of Consciousness Studies 8/5-7, 33 —. 2003, “The Roots of Empathy: The Shared Manifold Hypothesis and the Neural Basis of Intersubjectivity”, Psychopathology 36, 171 Gallese, V., L. Fadiga, L. Fogassi and G. Rizzolatti 1996, “Action Recognition in the Premotor Cortex”, Brain 119, 593 Gallese, V., C. Keysers, and G. Rizzolatti 2004, “A Unifying View of the Basis of Social Cognition”, TRENDS in Cognitive Sciences 8, 396 Gilligan, C. 1982, In a Different Voice: Psychological Theory and Women's Development, Cambridge: Harvard University Press Goldie, P. and A. Coplan (eds.) 2011, Empathy. Philosophical and Psychological Perspectives, Oxford: Oxford University Press Goldman, A. I. 1993, “Ethics and Cognitive Science”, Ethics 103 (2), 337 Gordon, R.M. 1995, “Sympathy, Simulation, and the Impartial Spectator”, Ethics 105 (4), 727

210

Chapter Thirteen

Greene, J.D. 2003, “From Neural “Is” to Moral “Ought”: What are the Moral Implications of Neuroscientific Moral Psychology?”, Nature Reviews Neuroscience 4, 847 —. 2009, “The Cognitive Neuroscience of Moral Judgment”, in Gazzaniga, M.S. (ed.) The Cognitive Neurosciences IV, Cambridge: MIT Press, 987 Greene, J.D. and J. Haidt 2002, “How (and Where) Does Moral Judgment Work?”, TRENDS in Cognitive Sciences 6 (12), 517 Greene, J.D., R.B. Sommerville, L.E. Nystrom, J.M. Darley and J.D. Cohen 2001, “An fMRI Investigation of Emotional Engagement in Moral Judgment”, Science 293, 2105 Held, V. 1993, Feminist Morality, Chicago: University of Chicago Press —. (ed.) 1995, Justice and Care: Essential Readings in Feminist Ethics, Boulder, CO: Westview Press —. 2006, The Ethics of Care: Personal, Political, and Global, New York: Oxford University Press Hume, D. 1739-1740, Treatise of Human Nature, London: John NoonThomas Longman, re-edited 1969, London: Penguin Iacoboni, M. 2009, “Imitation, Empathy, and Mirror Neurons”, Annual Review of Psychology 60, 653 Iacoboni, M., R.P. Woods, M. Brass, H. Bekkering, J.C. Mazziotta, and G. Rizzolatti 1999, “Cortical Mechanisms of Human Imitation”, Science 286, 2526 Jackson, P.L., A.N. Meltzoff and J. Decety 2005, “How Do We Perceive the Pain of Others? A Window into the Neural Processes Involved in Empathy”, Neuroimage 24, 771 Joyce, R. 2008, “What Neuroscience Can (and Cannot) Contribute to Metaethics”, in Moral Psychology Volume 3: The Neuroscience of Morality: Emotion, Brain Disorders, and Development, edited by Sinnott-Armstrong, W., Cambridge: MIT Press, 371 Kant, I. 1785, Grundlegung zur Metaphysik der Sitten. Translated by Gregor, M. 1996, “Groundwork of the Metaphysics of Morals”, in Immanuel Kant Practical Philosophy, New York: Cambridge University Press —. 1788, Kritik der practischen Vernuft. Translated by Gregor, M. 1996, “Critique of Practical Reason”, in Immanuel Kant Practical Philosophy, New York: Cambridge University Press —. 1797, Die Metaphysik der Sitten. Translated by Gregor, M. 1996, “The Metaphysics of Morals”, in Immanuel Kant Practical Philosophy, New York: Cambridge University Press

Empathy and Concern for Others

211

Keysers, C., B. Wicker, V. Gazzola, J.L. Anton, L. Fogassi and V. Gallese 2004, “A Touching Sight: SII/PV Activation during the Observation and Experience of Touch”, Neuron 42, 335 Kittay, E.F. and E.K. Feder (eds.) 2002, The Subject of Care: Feminist Perspectives on Dependency, Totowa NJ: Rowman and Littlefield Lamm, C., C.D. Batson and J. Decety 2007, “The Neural Substrate of Human Empathy: Effects of Perspective-taking and Cognitive Appraisal”, Journal of Cognitive Neuroscience 19 (1), 42 Meltzoff, A.N. and A.K. Moore 1977, "Imitation of Facial and Manual Gestures by Human Neonates", Science 198 (4312), 75 Meltzoff, A.N. and A.K. Moore 1983, "Newborn Infants Imitate Adult Facial Gestures", Child Development 54, 702 Meyer, M.L., C.L. Masten, Y. Ma, C. Wang, Z. Shi, N.I. Eisenberger and S. Han 2012, “Empathy for the Social Suffering of Friends and Strangers Recruits Distinct Patterns of Brain Activation”, Social Cognitive and Affective Neuroscience, 1 Morrison, I., D. Lloyd, G. di Pellegrino and N. Roberts 2004, “Vicarious Responses to Pain in Anterior Cingulate Cortex: Is Empathy a Multisensory Issue?”, Cognitive, Affective & Behavioral Neuroscience 4, 270 Pfeifer, J.H., M. Iacoboni, J.C. Mazziotta and M. Dapretto 2008, “Mirroring Others’ Emotions Relates to Empathy and Interpersonal Competence in Children”, NeuroImage 39, 2076 Preston, S.D. and F.B.M. de Waal 2002, “Empathy: Its Ultimate and Proximate Bases”, Behavioral and Brain Sciences 25, 1 Prinz, J. 2007, The Emotional Construction of Morals, Oxford: Oxford University Press Rawls, J. 2000, Lectures on the History of Moral Philosophy, edited by Herman, B., Cambridge, Mass: Harvard University Press Rizzolatti, G., R. Camarda, L. Fogassi, M. Gentilucci, G. Luppino and M. Matelli 1988, “Functional Organization of Inferior Area 6 in the Macaque Monkey. II. Area F5 and the Control of Distal Movements”, Experimental Brain Research 71, 491 Rizzolatti, G. and L. Craighero 2004, “The Mirror-Neuron System”, Annual Review Neuroscience 27, 169 Rizzolatti, G., L. Fadiga, V. Gallese and L. Fogassi 1996, “Premotor Cortex and the Recognition of Motor Actions”, Cognitive Brain Research 3, 131 Rizzolatti, G., L. Fadiga, V. Gallese and L. Fogassi 2001, “Neurophysiological Mechanisms Underlying the Understanding and Imitation of Action”, Nature Reviews Neuroscience 2, 661

212

Chapter Thirteen

Rizzolatti, G. and C. Sinigaglia 2006, So quello che fai. Il cervello che agisce e i neuroni specchio, Milano: Raffaello Cortina Roughley, N. 2013, Empathy, Approval and Disapproval On Michael Slote’s Moral Sentimentalism, http://www.uni-due.de/~se753pa/arbeitsgruppe_roughley/pdf/slote_ workshop_2013_paper_roughley.pdf Scheler, M.F. 1923, Wesen und Formen der Sympathie, in Gesammelte Werke. Vol. VII, edited by Frings, M., Bern und München: Francke Verlag. Translated by Heath, P. 2008, The Nature of Sympathy, New Brunswick, New Jersey: Transaction Publishers Shamay-Tsoory, S.G., J. Aharon-Peretz and D. Perry 2009, “Two Systems for Empathy: A Double Dissociation between Emotional and Cognitive Empathy in Inferior Frontal Gyrus versus Ventromedial Prefrontal Lesions”, Brain 132, 617 Simner, M.L. 1971, "Newborn's Response to the Cry of Another Infant”, Developmental Psychology 5, 136 Singer, T., B. Seymour, J. O’Doherty, H. Kaube, R.J. Dolan and C.D. Frith 2004, “Empathy for Pain Involves the Affective but not Sensory Components of Pain”, Science 303, 1157 Singer, T., B. Seymour, J. O’Doherty, K.E. Stephan, R.J. Dolan and C.D. Frith, 2006, “Empathic Neural Responses are Modulated by the Perceived Fairness of Others”, Nature 439, 466 Slote, M. 2010, Moral Sentimentalism, Oxford: Oxford University Press. Smith, A. 1853, The Theory of Moral Sentiments or An Essay Towards an Analysis of the Principles by which Men Naturally Judge Concerning the Conduct and Character, first of their Neighbours, and Afterwards of Themselves, London: Henry G. Bohn Stein, E. 1917, Zum Problem der Einfühlung, Halle. Translated by Stein, W. J. 1989, On the Problem of Empathy. The Collected Works of Edith Stein Sister Teresa Bendicta of the Cross Discalced Carmelite Volume Three, New York: Springer Wicker, B., C. Keysers, J. Plailly, J.P. Royet, V. Gallese and G. Rizzolatti, G. 2003, “Both of us Disgusted in My Insula: The Common Neural Basis of Seeing and Feeling Disgust”, Neuron 40, 655 Wood, A.W. 1999, Kant's Ethical Thought, Cambridge: Cambridge University Press Zahavi, D. 2001, "Beyond Empathy. Phenomenological Approaches to Intersubjectivity", Journal of Consciousness Studies 8/5-7, 151 —. 2007, "Expression and Empathy", in Folk Psychology Reassessed, edited by Hutto, D.D. and M. Ratcliffe, Dordrecht: Springer, 25

Empathy and Concern for Others

213

—. 2008, "Simulation, Projection and Empathy", Consciousness and Cognition 17 (2), 514 —. 2010, “Empathy, Embodiment and Interpersonal Understanding: From Lipps to Schutz”, Inquiry 53(3), 285

Notes 1

I am thankful to Greta Favara for helpful discussions and insightful comments on an earlier version of this paper.



CHAPTER FOURTEEN TRUTH AND MORAL RESPONSIBILITY P. ROGER TURNER

Most philosophers who study moral responsibility have done so in isolation of the concept of truth. Here, I show that thinking about the nature of truth has profound consequences for discussions of moral responsibility. In particular, by focusing on a very obvious fact about truth—that truth depends on the world and not the other way around—we can see that widely accepted alleged counterexamples to a key inference rule in an important incompatibilist argument can be shown not only to be unsuccessful, but also impossible. The Direct Argument goes, informally, as follows: If determinism is true, then our acts are the consequences of the laws of nature and events in the remote past. But we are not responsible for what went on before we were born, and neither are we responsible for what the laws of nature are. Therefore, we are not responsible for the consequences of these things (including our present acts). (van Inwagen 1983, p. 16)1

The Direct Argument rests on the following two rules of inference (where, “फ” stands for broadly logical necessity; “Š” stands for material implication; and “NRp” stands for “p and no-one is now or ever has been even partly directly morally responsible for p”): Rule A: From फp, we may infer NRp Rule B: From NRp and NR(p Š q), we may infer NRq In recent years, the Direct Argument has received a lot of critical attention, most of it paying special attention to the Argument’s inference rules, as well as some key metaphysical assumptions. In particular, Rule B has come under the most fire from compatibilists wishing to disarm the



Truth and Moral Responsibility

215

Direct Argument. This has happened in (at least) two ways. First, and most prominently, compatibilists have leveled alleged counterexamples to Rule B in an attempt to show that the rule is invalid. Second, compatibilists have leveled charges of dialectical impropriety with respect to Rule B.2 It is the first challenge against Rule B—i.e. alleged counterexamples to the Rule—that I wish to focus on in this paper. Specifically, I want to argue that if we reflect on the nature of truth, we will see that there’s good reason to think that counterexamples to Rule B are impossible. If I am right that the nature of truth shows us that counterexamples to Rule B are impossible, such conclusion is important. For, if Rule B is valid, then incompatibilism is true. Moreover, if Rule B is invalid, then there are possible cases that illustrate its falsity; that is, there are counterexamples to the Rule. So, it falls right out of my overall conclusion—viz., that an obvious fact about truth implies that there can’t be counterexamples to Rule B—that compatibilism commits its adherents to denying an obvious fact about truth. That compatibilists must deny an obvious fact about truth is very surprising (and important) indeed! To begin my defense of the claim that counterexamples to Rule B are impossible, I follow Trenton Merricks (2007, 2009, 2011a, 2011b) by noting that truth depends (in a very trivial way) on the world. It’s true, for example, that dogs bark because dogs bark; it’s true that Turner exists because I exist; that I write this paper at t is true because I write this paper at t, and so on. No one should disagree with this “truism about truth” (Merricks 2009, p. 31). That is, no one should disagree that truth depends on the world in this trivial way.3 Many philosophers disagree about how it is that truth “depends on” the world in this trivial way; that is, many philosophers disagree about how the truism about truth is true. But none should disagree that the truism about truth is true. And I do not think anyone would disagree with this truism about truth. But I think that reflection on this truism will reveal that any alleged counterexamples to Rule B must fail. They must fail because it’s impossible to give a bona fide counterexample to Rule B given this truism about truth. To begin to see why I say this, note that Merricks (2009, 2011a, 2011b) argues that the following is a corollary to the truism about truth, what I’ll call: Truth DependenceCHOICE [TDC]: For all agents, S, and all propositions, p, if S has a choice about what p’s truth depends on (in the sense of “depends on” in which truth depends on the world), then S has a choice about p’s truth.



216

Chapter Fourteen

To see that TDC is a corollary of the truism about truth, imagine that the proposition that Jones kills Smith is true. Given the truism about truth, that Jones kills Smith is true because Jones kills Smith; that is, that Jones kills Smith depends on what Jones does. And if Jones has a choice about whether or not he kills Smith, it seems easily to follow that he has a choice about whether or not it’s true that Jones kills Smith. For, suppose that we thought Jones doesn’t have a choice about the truth of that Jones kills Smith. We’d think this because we’d think Jones doesn’t have a choice about whether or not he kills Smith. So, not only do we see that there’s a close connection between “having a choice about” and the nature of truth, we can see that the truism about truth reveals a principle about “having a choice about” the truth of a proposition, namely, TDC. I think that something similar is true about the relationship between moral responsibility and the nature of truth. To see this, suppose it’s true that Jones kills Smith. It’s true that Jones kills Smith because Jones kills Smith; that is, the truth of that Jones kills Smith depends on Jones and what he does. So, given the truism about truth, it follows that that Jones kills Smith would not have been true had Jones not acted as he did. Now, if all of that is right, then I think the truism about truth has the following corollary in addition to TDC: Truth DependenceMORAL [TDM]: For all agents, S, and all propositions, p, if S is directly morally responsible for that which p’s truth depends on (in the sense of “depends on” in which truth depends on the world), then S is at least partly directly morally responsible for p’s truth. And if TDM is a corollary to the truism about truth, then, as I’ll go on to argue, since the truism about truth is necessarily true, counterexamples to Rule B are impossible. To see why I say that TDM is a corollary to the truism about truth, notice that given the truism about truth and Jones’s moral responsibility for killing Smith, it follows that that Jones kills Smith would not have been true had Jones not acted as he did. Now, suppose that we thought that Jones isn’t directly morally responsible for the fact that Jones kills Smith. I say we’d think this because we’d think that Jones isn’t directly morally responsible for the thing upon which the truth of that Jones kills Smith depends, viz., Jones’s killing of Smith. And this generalizes. Thus, for all S and all p, if S is directly morally responsible for the thing upon which the truth of p depends, then S is at least partly directly morally responsible for p’s truth. If TDM is true, so I’ll go on to show, then it is impossible to give a counterexample to Rule B. So, what I intend to do in the paper is the



Truth and Moral Responsibility

217

following. First, I will show how TDM’s (assumed) truth reveals a way in which two recent alleged counterexamples to Rule B fail. But these reasons will generalize; so, TDM’s (assumed truth) will reveal how all alleged counterexamples fail. Second, I will defend TDM from what I take to be the most pressing objection to TDM. I’ll conclude that this objection fails to undermine TDM; so if TDM is true, its truth undermines all attempts to provide a counterexample to Rule B. Such conclusion is important. For, I think that TDM is intuitively plausible. Moreover, I think that if the truism about truth is true, then TDM is true. That is, I think that the truism about truth implies TDM. Thus, to deny TDM (which any compatibilist must, assuming (i) that Rule B implies incompatibilism, and (ii) that there can be counterexamples to Rule B if it’s invalid) will come at great cost; for, the objector to TDM will have to deny the truism about truth.

1. Rule B and some alleged counterexamples Recall that the Direct Argument depends on the following rule of inference: Rule B: From NRp and NR(p Š q), we may infer NRq. This inference rule is the most controversial aspect of the Direct Argument.4 Where “NR” is an operator that means “nobody is now, or ever has been, even partly directly morally responsible for the fact that”, what follows is that, if Rule B is valid, then if you’re not (now or ever) even partly directly morally responsible for some fact, p, and you’re not (now or ever) even partly directly morally responsible for the fact that p implies some further fact, q, then you’re not (now or ever) even partly directly morally responsible for the fact that q. Van Inwagen attempts to motivate Rule B with various cases like the following: Plato: NR Plato died in antiquity NR (Plato died in antiquity Š Plato never met Hume) NR Plato never met Hume. (van Inwagen 1983, p. 187)

And surely the conclusion of Plato—that no one is now (or ever has been, or ever will be) even partly directly morally responsible for the fact that Plato never met Hume—follows given the truth of the first two steps.



Chapter Fourteen

218

And since Plato is just a substitution instance of Rule B, Rule B appears to be valid. But, as I say, Rule B is contentious, and lots of ink has been spilt in discussion of this controversial inference rule. So, now I wish to consider two recent alleged counterexamples to Rule B. My aim is to show that counterexamples to Rule B are impossible; so, I use these two recent purported counterexamples in order to illustrate this point. I begin by considering a well-known case called “Erosion”, introduced by John Martin Fischer and Mark Ravizza (1998). Many philosophers think that Erosion decisively tells against Rule B; so, if it fails as a counterexample to Rule B, this is important. Next, I’ll consider an alleged counterexample by Ishtiyaque Haji (2010).5 As we’ll see, each case is a different type of alleged counterexample. Fischer and Ravizza’s case is a case of overdetermination, while Haji’s case is a case of libertarianly free action where there is no overdeterminatin involved. Importantly, I’ll show that each of these cases fails to provide a counterexample for the same reasons. Thus, such reasoning will generalize and all alleged counterexamples to Rule B will fail. I’ll start with Fischer and Ravizza’s attempt to provide a counterexample. Consider: Erosion: Imagine that Betty [a soldier charged with destroying an enemy fortress] plants her explosives in the crevices of the glacier and detonates the charge at T1, causing an avalanche that crushes the enemy fortress at T3. Unbeknownst to Betty and her commanding officers, however, the glacier is gradually melting, shifting, and eroding. Had Betty not placed the dynamite in the crevices, some ice and rocks would have broken free at T2, starting a natural avalanche that would have crushed the enemy camp at T3. (Fischer and Ravizza 1998, p. 157)

Erosion is alleged to be a counterexample to Rule B because 1. The glacier is eroding and no one is, or ever has been, even partly responsible for the fact that it is eroding; and 2. If the glacier is eroding, then there is an avalanche that crushes the enemy base at T3, and no one is, or ever has been, even partly responsible for this fact;

But, given Betty’s responsibility, it is not true that 3. There is an avalanche that crushes the enemy base at T3, and no one is, or ever has been, even partly responsible for this fact. (Ibid.)



Truth and Moral Responsibility

219

So, it appears that Rule B is invalid. For, there are two paths that suffice for the enemy camp’s having been crushed by the glacier: one that in fact obtains; the other counterfactual since that path didn’t obtain (though it would have). But, since this counterfactual natural intervener— the erosion of the glacier—does not actually cause the avalanche, it does not remove Betty’s moral responsibility for the enemy camp’s having been crushed by the glacier. So, even though the enemy camp’s being crushed by the glacier is inevitable, it doesn’t follow that Betty isn’t morally responsible for its having been so crushed. Rule B is invalid. To put the point a bit more clearly, notice that Erosion contains two paths. The first path passes through Betty, a normally functioning agent. The second path, however, does not pass through Betty (or any other normally functioning agent). The second path is merely a counterfactual path that Fischer and Ravizza call the “Ensuring Path”. The Ensuring Path, obviously enough, ensures that the consequence—in this case, the enemy’s being crushed by the glacier—obtains. So, even though, 4. There is some Ensuring Path leading to a particular outcome and no one is, or ever has been, even partly responsible for this fact;

and 5. If there is this Ensuring Path, then the outcome is reached, and no one is, or ever has been, even partly responsible for this fact;

it does not follow that 6. The outcome is reached and no one is, or ever has been, even partly responsible for this fact.

6 doesn’t follow because, since the outcome (the camp’s being destroyed by the glacier) was not caused by the natural intervention of ice and rocks breaking free (but, rather, by Betty’s placing the dynamite), Betty is responsible for the enemy camp’s having been crushed by the glacier even though this would have happened even if she had never detonated her explosives at T1. Rule B is invalid. But I don’t think that Erosion successfully shows that Rule B is invalid; and reflection on the truism about truth (§I) will help us to see why. Recall that it’s true that Turner exists because I exist, and it’s true that I write this paper at t because I write this paper at t. The truism about truth asserts just the notion that truth depends on the world in this very



220

Chapter Fourteen

trivial way. And, as I argued above, the truism about truth has TDM as a corollary. There is an objection to TDM (or something like it) in the offing, and I’ll consider it in the next section. For now, however, let’s assume that TDM is true. Given TDM’s truth, does Erosion supply a successful counterexample to Rule B? I think that it does not. To see why I say this, consider: 1*. NR The glacier is eroding 2*. NR (The glacier is eroding Š there is an avalanche that crushes the enemy base at T3) 3*. NR There is an avalanche that crushes the enemy base at T3. 1* – 3* is just 1 – 3 written in the form of Rule B. And the idea is that, even though 1* and 2* are true, 3* is false because Betty is directly morally responsible for the fact that there is an avalanche that crushes the enemy base at T3. Let’s assume, with Fischer and Ravizza, that Betty is morally responsible for the fact that an avalanche crushes the enemy base when it does. Now, let’s write the equivalent of the conditional in 2* as follows: 2**. ~ (The glacier is eroding) v (There is an avalanche that crushes the enemy base at T3). Since the first disjunct of 2** is false, we can see that 2** is true just in virtue of the fact that the second disjunct is true.6 But notice that 2**’s truth depends on the fact that there is an avalanche that crushes the enemy base at T3. Since Betty is, ex hypothesi, morally responsible for the fact that there is an avalanche that crushes the enemy base at T3 because she’s morally responsible for the avalanche that crushes the enemy base when it does, we can conclude on the basis of TDM that 2* is false. That is, Betty is morally responsible for the fact expressed in 2** because 2**’s truth depends on (in the sense of “depends on” in which truth depends on the world) the fact that there is an avalanche that crushes the enemy base at T3, something for which Betty is ex hypothesi morally responsible. Therefore, Betty, contra Fischer and Ravizza’s claim, is morally responsible for the fact that lies within the scope of the “NR” operator in



Truth and Moral Responsibility

221

2; and so, 2 itself is false. Thus, Erosion fails as a counterexample to Rule B. Now, Fischer and Ravizza might respond as follows. You’ve failed to show that Betty is morally responsible for the conditional contained in 2 because you’ve failed to show that the relevant portion of 2’s truth depends on anything that Betty has done. By hypothesis, that there is an avalanche that crushes the enemy base at T3 would have been true no matter what Betty did; there was an “ensuring condition”—namely, the erosion of the glacier—that would have led to the crushing of the enemy base even if Betty had never planted the explosive device. Thus, you’ve failed to show that Betty’s actions are what the truth of 2 depends on; so, you’ve failed to show that the truth of 2 depends on Betty. Erosion hasn’t yet been undermined.

But in reply I ask the following question: Is Betty morally responsible for the fact that there is an avalanche that crushes the enemy base at T3, or not? If she is responsible for the fact that there is an avalanche that crushes the enemy base at T3, and the truth of the conditional that lies within the scope of the “NR” operator in 2 depends on this fact, then, given TDM, Betty is morally responsible for the truth of the conditional that lies within the scope of the “NR” operator in 2.7 Now, if Betty isn’t directly morally responsible for the fact that there is an avalanche that crushes the enemy base at T3, then, of course, there is no problem for Rule B. In such a case, Rule B is confirmed rather than refuted. In any case, equipped with TDM, we can see that Erosion fails as a counterexample to Rule B. Ishtiyaque Haji (2010) offers the following attempt at a counterexample to Rule B: Hal’s Creation: Hal-2 [an essentially omniscient, sempiternal, amoral— that is, lacks knowledge of moral right, wrong, or obligatory—being] has the ability to create (or actualize) any one of an infinite number of possible worlds…Suppose Hal-2 creates a world, W1, in which, after due reflection, Yasmin [in a libertarianly free way] donates a large sum of money (at some time, ts) to a credible charity, UNICEF. Yasmin really cares about the plight of the needy children; she donates because she wishes to help the kids and not, for instance, because she wants a big tax break. We safely suppose that she is morally praiseworthy for her bountiful donation. Under appropriate circumstances normal agents would be deserving of praise for such an act (Haji 2010, p. 125).

Hal’s Creation generates the following substitution instance of Rule B:



Chapter Fourteen

222 Hal’s Argument:

(4H): NR (Hal-2 creates W1). (5H): NR (Hal-2 creates W1 Š Yasmin donates (at ts in W1)). (6H): Therefore, NR (Yasmin donates (at ts in W1) (Ibid.).

The upshot of Hal’s Argument is that, since Hal is an amoral agent, 4H is true because neither he nor anyone else is morally responsible for the creation of W1; 6H is false because, ex hypothesi, Yasmin (libertarianly) freely donates the money at ts in W1 in the right sort of way for moral responsibility; but, 5H is true because neither Hal nor Yasmin are morally responsible for the fact that (Hal-2 creates W1 Š Yasmin donates (at ts in W1)). Of course, one might object that Yasmin is morally responsible for the truth of the conditional that lies within the scope of the “NR” operator in 5H. But, if so, then Yasmin would have to be morally responsible for YC: Hal-2 creates W1 Š at ts Yasmin donates to UNICEF (Haji 2010, p. 127).

But, thinks Haji, for Yasmin to be morally responsible for YC, she’d have to meet the following epistemic condition on moral responsibility, namely: E3: S is morally praiseworthy [blameworthy] for seeing to the occurrence of state of affairs, a, only if S morally ought to have known or believed that a is morally obligatory [morally wrong] or, as some might prefer, morally good [morally bad] (in some specified sense of “good” [“bad”] (Ibid.).

And how could Yasmin be expected to meet an epistemic condition like E3 with respect to YC? Indeed, “there are no compelling grounds to require that in Hal’s creation, Yasmin ought to have known or believed that YC is obligatory or morally good” (Ibid.). So, Yasmin is not morally responsible for YC even though she’s responsible for donating, at ts, to UNICEF. But, since Hal isn’t responsible for YC either, 5H is true. Thus, thinks Haji, we have a successful counterexample to Rule B. But I think that, given TDM, Hal’s Creation does not generate a counterexample to Rule B; for, I think that TDM’s truth reveals that 5H is false. To see that this is so, let’s—as we did with the Erosion case— rewrite 5H’s conditional this way:



Truth and Moral Responsibility

223

(5H*): ~ (Hal-2 creates W1) v (Yasmin donates (at ts in W1)). As with 2** from the Erosion case, we can see that since the first disjunct of 5H* is, ex hypothesi, false, 5H* is true just in virtue of the truth of the second disjunct. That is, 5H* is true because Yasmin donates to UNICEF when she, in fact, does. But Yasmin is responsible for this fact. Thus, given TDM, Yasmin is at least partly directly morally responsible for the truth of 5H*. And if Yasmin is at least partly directly morally responsible for the truth of 5H*, then it follows that she’s at least partly directly morally responsible for the truth of the conditional contained in 5H of Hal’s Argument. So, 5H itself is false, and Hal’s Creation is no counterexample to Rule B. I’ve shown that Erosion, and Hal’s Creation fail to provide successful counterexamples to Rule B if TDM is true. But it should be clear that I’ve shown more than that. For, since every alleged counterexample to Rule B will have to be of the same form, every alleged counterexample to Rule B suffers the same affliction. Namely, in each case, someone is (or was, or will be) ex hypothesi directly morally responsible for the thing upon which the truth of the consequent contained within the second premise of each substitution instance of Rule B depends. And this means that someone is (or was, or will be) at least partly directly morally responsible for the thing upon which the truth of the conditional contained within the second premise of each substitution instance of Rule B depends. Thus, if TDM is true, counterexamples to Rule B are impossible.

2. Is Truth DependenceMORAL true?: an objection to TDM I will now consider an objection to Truth DependenceMORAL. The objection I’ll consider begins with the following story.8 Zombie Case: Sara does cutting edge scientific research. She knows her craft well. She fulfills the requisite requirements to be morally responsible for her research, and reports her research to her boss, Ted. Unbeknownst to Sara, however, Ted is a mastermind controlling a large conglomerate of labs. Ted uses Sara’s work, along with the work of many other scientists (whose work we can safely assume Sara would not grasp), to create a virus that, when released, turns half of the world into flesh-eating zombies. Sara bears moral responsibility for a part of the way things are: namely, that her research took place. But Sara doesn’t bear part of the moral responsibility for the way things are. It would be inappropriate to blame Sara for the



224

Chapter Fourteen zombie outbreak. Plausibly, Sara is non-culpably ignorant, and so gets off the hook.9

The Zombie Case is, I think, supposed to show that someone could be morally responsible for some fact but lack the epistemic requirements for responsibility with respect to the “dependent” fact. So, for example, Sara bears moral responsibility for the fact that her research took place, but it’s alleged that she bears no responsibility for the fact that her research took place and there’s a zombie outbreak, and this is because she doesn’t have any good reason to believe that the truth of this second fact—this conjunctive fact—depends on her. But, I think that this objection can be met with the following story. Suppose that I am a mason, tasked with the making, and laying, of a single brick that’s part of the foundation of a beautiful mansion. And suppose that my creating and laying this brick is, other things equal, a morally praiseworthy action. Now, suppose, also, that all I know how to do, at least when it comes to building things, is make and lay bricks; I don’t know anything about engineering, blue-prints, or anything else relevant to the task of building a house. Moreover, I’m not so much as capable of knowing such things (for whatever reason). According to TDM, if I’m directly morally responsible for the fact that this particular brick is made and the truth of that this particular brick is made and the beautiful mansion is built depends on whether or not I make this particular brick, then I’m at least partly directly morally responsible for this conjunctive fact; I’m partly directly morally responsible for the fact that things are such that that this particular brick is made and the beautiful mansion is built. Now, suppose that the homeowner, the person who designed the house and commissioned me to build and lay the single brick, came to me and thanked me for the fact that her house is built. That is, suppose the homeowner thanked me in such a way as to be a type of moral praise given to me for the fact that her home is built. Is the homeowner out of line? Has she gotten her wires crossed with respect to whom she ought to praise for the building of her house? I don’t see any obvious reason to think the answer to either of these questions is “yes.” Indeed, it seems perfectly natural for the homeowner to thank me in this way since her home’s being built the way it was (we may assume) wasn’t so much as possible without my having constructed the relevant brick. And this seems true to me regardless of whether or not I understood (or could understand) the larger state of affairs of which my constructing the brick was a part. But suppose that the homeowner is out of line. Does this show that TDM is false? I think that it does not. For, TDM doesn’t imply that I am



Truth and Moral Responsibility

225

responsible for the fact that her house is built; rather, TDM implies that I am responsible for the fact that this particular brick is made and her house is built. So, it seems to me that meeting the epistemic conditions for responsibility for the making of the brick suffices for meeting the epistemic conditions for responsibility for the whole of which the brick plays a part. Moreover, I think that this objection confuses what’s at issue. TDM doesn’t imply that a person is responsible for all of the conjuncts in a conjunction, for example. Nor does it imply that a person is responsible for the antecedent in a conditional. All it implies is that a person is responsible for the truth of the conjunction, or the truth of the conditional (or, etc.) given that the truth of those things depends on what they do. For, suppose it’s true that Sara’s research took place. Now, suppose that Sara isn’t responsible for this fact. I think the only thing to conclude here is that she’s not responsible for this fact because she’s not responsible for the thing upon which the truth of the proposition depends, viz., doing her research. But this is just to say that if she’s responsible for the thing upon which the truth of the proposition depends, then she’s responsible for the truth of the proposition. And given the Zombie Case, the truth of the proposition that her research took place and there’s a zombie outbreak depends on what Sara does. Thus, though Sara is responsible for the fact that her research took place and there’s a zombie outbreak, it doesn’t follow that she’s at all responsible for the fact that there’s a zombie outbreak. 10 And it seems to me that this is what the Zombie Case is supposed to elicit. Thus, I conclude that TDM is safe from this objection. Moreover, I think that this type of objection is the most pressing objection against TDM. So, I think that TDM is safe from objection, full stop. And if TDM is safe, then counterexamples to Rule B are impossible.

3. Conclusions This is a truism about truth: that dogs bark is true because dogs bark; that I write this paper at t is true because I write this paper at t, and so on. I follow Trenton Merricks (2007, 2009, 2011a, 2011b) when I say that that truth depends on the world and not the other way around denotes just this truism, the truism about truth. But this truism, I claimed, has the following corollary: Truth DependenceMORAL [TDM]: For all agents, S, and all propositions, p, if S is directly morally responsible for that which a p’s truth depends



226

Chapter Fourteen

on (in the sense of “depends on” in which truth depends on the world), then S is directly morally responsible for that p’s truth. If I’m right about this, then accepting the truism about truth requires accepting TDM; and denying TDM requires rejecting the truism about truth. But no one should reject the truism about truth; so, no one should reject TDM. But, if TDM is right, then counterexamples to Rule B—the Direct Argument’s most contentious transfer principle—are impossible. I argued for this claim by, first, showing that two recent attempts to provide a counterexample (Erosion, and Hal’s Creation) fail to so provide given the truth of TDM. Moreover, I argued that TDM’s truth shows more than the failure of Erosion and Hal’s Creation. It shows that any alleged counterexample to Rule B will fail. For, any purported counterexample to Rule B will have the same form. The form of any alleged counterexample to Rule B will go as follows: by hypothesis, no one is (or was, or ever will be) even partly directly morally responsible for the propositions governed by the “NR” operator in the first two premises of the substitution instance of Rule B. But someone is, by hypothesis, at least partly directly morally responsible for the proposition governed by the “NR” operator in the conclusion of the substitution instance of Rule B. But, if TDM is true, then counterexamples of this form (which is all of them!) will fail. For, the truth of the conditional in the second premise depends on what the agent in question does. Thus, in all attempts to give a counterexample to Rule B, someone is at least partly directly morally responsible for the truth of the conditional in the second premise. Thus, if TDM is true, then it’s impossible to give a substitution instance of Rule B where the conclusion is false but the premises are true. Which is just to say that if TDM is true, then it’s impossible to give a counterexample to Rule B.11

References Campbell, J.K. 2007, “Free Will and the Necessity of the Past”, Analysis 67, 105 —. 2008, “Reply to Brueckner”, Analysis 68, 264 —. 2010, “Incompatibilism and Fatalism: Reply to Loss”, Analysis 70, 71 Fischer, J.M. and M. Ravizza 1998, Responsibility and Control: A Theory of Moral Responsibility, Cambridge: Cambridge University Press Fischer, J.M. 2006, My Way: Essays on Moral Responsibility, New York: Oxford University Press



Truth and Moral Responsibility

227

Fischer, J.M. and E. Stump 2000, “Transfer Principles and Moral Responsibility”, Philosophical Perspectives 14, 47 Haji, I. 2009, Incompatibilism’s Allure: Principal Arguments for Incompatibilism, Toronto: Broadview Press —. 2008, “Reflections on the Incompatibilist’s Direct Argument”, Erkenntnis 68, 1 —. 2010, “On the Direct Argument for the Incompatibility of Determinism and Moral Responsibility”, in Grazer Philosophische Studien 80, 111 Kearns, S. 2011, “Responsibility for Necessities”, Philosophical Studies 155, 307 McKenna, M. 2008, “Saying Good-bye to the Direct Argument the Right Way”, Philosophical Review 117, 349 Merricks, T. 2007, Truth and Ontology, New York: Oxford University Press —. 2009, “Truth and Freedom”, Philosophical Review 118, 29 —. 2011a, “Replies to Hasker and Zimmerman”, in Perszyk, K. (ed.) Molinism: The Contemporary Debate, New York: Oxford University Press, 90 —. 2011b, “Truth and Molinism,” in Persyzk, K. (ed.) Molinism: The Contemporary Debate, New York: Oxford University Press, 50 Shabo, S. 2010a, “Against Logical Versions of the Direct Argument: A New Counterexample”, American Philosophical Quarterly 47, 239 —. 2010b, “The Fate of the Direct Argument and the Case for Incompatibilism”, Philosophical Studies 150, 405 Schnall, I.M. and D. Widerker (forthcoming), “The Direct Argument for Incompatibilism”, in Palmer, D. (ed.) Libertarian Free Will, New York: Oxford University Press van Inwagen, P. 1983, An Essay on Free Will, Oxford: Oxford University Press Warfield, T.A. 1996, “Determinism and Moral Responsibility are Incompatible”, Philosophical Topics 24, 215 Widerker, D. 2002, “Farewell to the Direct Argument”, The Journal of Philosophy 99, 316

Notes 

1

What I’ve written, here, is van Inwagen’s “Consequence Argument” cashed out in terms having to do with “responsibility”. Where the Consequence Argument speaks in terms of the past and the laws of nature and whether or not they’re “up to” anyone, the Direct Argument speaks in terms of the past and the laws of nature and whether or not anyone is “responsible” for them.



228

Chapter Fourteen



2 See, e.g., John Martin Fischer (2006, chapter 8 in particular), and Michael McKenna (2008). 3 This isn’t to say that there’s nothing more substantive to say about the way truth depends on the world. One might wonder, for example, how true counterfactuals can “depend on the world”. But I leave aside discussions of such analyses of truth—e.g. Truthmaker theory, Correspondence theory, Truth Supervenes on Being (TSB) theory, etc.—because a discussion of such analyses not only takes us too far afield, it’s also irrelevant to the present discussion. What’s crucial to my project is that all agree that what’s true depends on how things are; how things are doesn’t depend on what’s true. So, e.g., if it’s true that if Roger is in C, then he would A, this is true (minimally) because I would A if I was in C. Perhaps there’s more that can be said about how this (assumed) truth “depends on” the world, but it’s not important for the present discussion. Even so, for a thorough treatment of the competing theories about truth, see Merricks (2007). 4 Rule B is the most controversial aspect of the Direct Argument, but there are other objections to the argument in the literature. Stephen Kearns (2011) questions the first inference rule upon which the Direct Argument rests; namely Rule A. Rule A says that no one is now (or ever was, or will be) even partly directly morally responsible for a necessary truth. Kearns attempts to call this Rule into question. Joseph Keim Campbell (2007, 2008, 2010) calls into question the assumption that the “determinism” thesis necessarily includes the idea of a remote, or distant, past. The upshot of Campbell’s objection is that, if he’s right, though the Direct Argument still goes through, it fails to establish incompatibilism, since incompatibilism, if true, is necessarily true. But, if there could be deterministic worlds that don’t include a remote, or distant, past, then all the Direct Argument establishes (if it establishes anything) is that moral responsibility and causal determinism are incompatible in worlds that include a remote, or distant, past. But this is consistent with the possibility that there are deterministic worlds that don’t include a remote, or distant, past but include morally responsible agents. In my dissertation, I argue that Kearns’s alleged counterexamples to Rule A fail, and that, while Campbell’s objection goes through, it can be met by a revised version of the Direct Argument that continues to rely on both Rule A and Rule B. So, Rule B’s validity continues to be important to a properly revised version of the Direct Argument for incompatibilism about moral responsibility and causal determinism. 5 There are, of course, many other attempts to provide a counterexample to Rule B. See, e.g., David Widerker (2002), Michael McKenna (2008), Seth Shabo (2010a; 2010b), Ishtiyaque Haji (2008; 2009). 6 Ira Schnall and David Widerker (forthcoming) use this exact move against Haji’s (2009) attempt to provide a counterexample to Rule B. 7 Suppose that TDM isn’t true. There is still reason to doubt that Erosion provides a counterexample to Rule B. For, there is a problem with 3*. The problem with 3* is that it misses the point of Rule B, that is, the point of direct moral responsibility. Following David Widerker (2002, pp. 118-119), we can think of direct moral responsibility in this way: “S is directly responsible for p just in case S is responsible for p, but not in virtue of being responsible for some other fact”. In



Truth and Moral Responsibility

229

 other words, direct moral responsibility—the relevant sort of responsibility to which Rule B applies—is the sort of responsibility one bears for the truth of some morally significant fact, but not because one bears responsibility for it in virtue of some other fact. Thus, Betty is directly morally responsible for the fact that there is an avalanche that crushes the enemy base at T3 only if she’s not responsible for this fact because of her responsibility for the truth of some other fact. And since Betty is, ex hypothesi, responsible for the fact that there is an avalanche that crushes the enemy base at T3 only in virtue of the fact that she caused the avalanche, she is not directly responsible for the fact that there is an avalanche that crushes the enemy base at T3. Thus, even if TDM isn’t true, Erosion is defeated. 8 There is, also, a potential worry that TDM entails the following principle: (W1): If no one is even partly morally responsible for a conjunction, then no one is even partly morally responsible for either conjunct of the conjunction. (Warfield, 1996, p.218) The worry with respect to W1 is from Fischer and Stump (2000). They rightly claim that W1 entails: (W3): Given a true antecedent of a conditional, a person is partly responsible for the conditional’s being true if he is partly responsible for the truth of the consequent of the conditional. (Fischer and Stump 2000, p. 52) And this is a problem because, among other things, W3, if true, provides reason to think that Rule A—the transfer principle that says no one is even partly directly morally responsible for a necessary truth—is false. For, notice that the following is necessarily true: if Roger murders Jones, then Roger murders Jones. Given that I’m at least partly directly morally responsible for the fact that Roger murders Jones, by W3, it follows that I’m also responsible for the fact that if Roger murders Jones, then Roger murders Jones. But this is a necessary truth; so, if W3 is true, then Rule A is false. And, of course, there’s no point in defending Rule B if we’re just going to sacrifice Rule A in the process. I don’t have room to deal with this potential worry as fully as it perhaps needs. But, here’s a gesture at what I’d say if I had the room to do it. I do not think that this objection, even if successful against W3, at all undermines TDM. This is because TDM does not entail W3. Now, I agree that TDM entails a principle similar to W3, but I argue that any such principle must have an appeal to the relevant notion of dependence. For, suppose that it’s true that if hobbits don’t exist, then I write this paper. Given that it’s true that hobbits don’t exist, the foregoing conditional is true if and only if I write this paper. Thus, I’m responsible for the conditional’s being true if and only if I write this paper. Now, suppose that I’m not responsible for it’s being true that if hobbits don’t exist, then I write this paper. If I’m not responsible for the truth of this conditional, then this is because I’m not responsible for writing this paper; that is, I’m not responsible for the thing upon



230

Chapter Fourteen

 which the truth of the conditional proposition depends. Thus, I conclude that any W3-like principle that TDM entails must include the relevant notion of dependence. But W3, itself, does not include the relevant notion of dependence; so, TDM does not entail W3. 9 I owe this case to Kyle Fritz. 10 Objection: Doesn’t the truth of that there is a zombie outbreak depend on Sara? If so, then, by TDM, isn’t Sara at least partly directly morally responsible for the fact that there is a zombie outbreak? If your answer is yes to either of these, then the Zombie Case is, contrary to what you say, a counterexample to TDM. Reply: No, the truth of that there is a zombie outbreak does not depend on what Sara does. At least, it does not depend on what Sara does in the relevant way. Recall that the notion of “dependence” at issue in TDM is the very trivial sort of dependence in which truth depends on the way the world is. So: is it true that there is a zombie outbreak if and only if Sara does her research? According to the details of the case, the answer must be no (e.g., Sara’s boss’s nefarious plan plays a part, etc.). There are, presumably, very many ways in which it could be true that there is a zombie outbreak Thus, I conclude that the truth of that there is a zombie outbreak does not depend on Sara—not, anyway, in the sense of “dependence” at issue in TDM. Thus, the Zombie Case is no counterexample to TDM. 11 I’d like to thank Joe Campbell, E.J. Coffman, John Martin Fischer, Kyle Fritz, Trenton Merricks, David Palmer, Derk Pereboom, Daniel Speaks, the participants at the Florida State University Graduate Conference on Free Will and Moral Responsibility, and the participants at the 2013 meeting of the Pacific Division of the American Philosophical Association for helpful and instructive comments on earlier versions of this (or parts of this) paper.



CHAPTER FIFTEEN “I WAS NOT AWARE I WAS DOING THAT” FRANZISKA MÜLLER

1. Introduction In her seminal book Intention, Elizabeth Anscombe argued that intentional actions are ones to which a certain sense of the question “Why did you do this?” has application (Anscombe 1957/2000, §5). In order to elucidate this particular sense of the question, Anscombe analysed different cases in which the question can not be asked in this sense. One reply that according to her successfully blocks application of the question is the answer “I was not aware I was doing that” (Anscombe 1957/2000, §6). In what follows, I will refer to this answer as “the unawarenessclaim”. Anscombe argues that if someone replies in this way to the question “why?” this means that the question was not applicable in the first place and “that” was not an intentional action of this person (excluding the possibility of a lie). This short passage of Anscombe’s book received a lot of attention, mainly for its close link to her controversial claim that agents have direct knowledge of their own actions. I do not want to enter into the topic of knowledge of actions here; even though on the cited paragraph, Anscombe uses “I was not aware” and “I did not know” interchangeably, I will concentrate on awareness. Further, in what follows, “S could truly answer the unawareness-claim” will be used as synonymous with “S was not aware she was doing that”. This may be somewhat untrue to Anscombe’s account. However, in this paper I am not concerned with analysing Anscombe’s actual stance on the matter, but with the issue her passage raises: the relation between intentional action and awareness of action. We can then formulate P that is derived from Anscombe’s passage as follows: P: If S is not aware she is doing X, doing X is not an intentional action of hers.

232

Chapter Fifteen

As P (or some variation of it) seems widely accepted, I will not specifically argue for its truth but focus on analysing its basic notions. The first part of the paper is dedicated to an analysis of “doing X” as it appears in P; the second part deals with the notion of awareness at issue.

2. What the agent is unaware of What kind of things can we do in the absence of awareness of our doing so? I start with some general remarks on P that shape the further investigation. First, we should note that P is ambiguous about what follows from the absence of awareness. All that is claimed is that doing X in the absence of awareness cannot be an intentional action. This leaves open whether it can be an unintentional action or whether absence of awareness implies absence of action altogether. Which of these stances we take depends on our theory of action. I will not take a stand on that here: “Doing X” is used as an umbrella term for all kinds of doings and actions an agent is active in, and should not be understood as to exclude nor imply action. Second, I want to insist on the phrasing of P in which doing X is presented as factive. If P is true there must be something the agent is doing in the absence of awareness of it. Accepting P then requires accepting that a subject can do something without being aware of that. Situations in which the agent is not actually doing anything are not at issue here. A third remark concerns what can be filled in for X in P. A close reading of the little conversation cited above suggests that X should be an event in the world: there is some event, e.g. a bodily movement that is observed by the inquirer and she suspects that the person questioned brought it about. I propose that we should understand “doing X” as “the agent is bringing it about that the event X obtains”. This is supposed to remain neutral on whether the alleged action is identical with the event in the world or with her bringing that event about (or with a combination of both). This clarified, I propose three different situations in which an agent could claim that she is not aware of doing X. The situations vary depending on what about doing X she is unaware of. i. In the first situation, the agent has no awareness of the supposed action altogether. If she utters the unawareness-claim the agent means that she is and/or was not aware that the event X took place at all, let alone that she was involved in it.

“I was not aware I was doing that”

233

Admitting P in the sense specified by this scenario is to say (a) the agent can do things she is not aware of at all, and (b) what she is doing in this situation cannot be an intentional action of her. Someone admitting P in this sense has thus to allow that an agent can do things that are completely unintentional. A plausible candidate for this may be what has been called sub-intentional action (see Steward, 2009). To deny P in this reading means either to claim that in doing something an agent necessarily is aware of it somehow (thus denying a). This has been defended by Brian O’Shaughnessy in the last edition of The Will (2008, pp. 349–51). Or else denying P here is to claim that we can not only do things in the absence of awareness but even perform intentional actions without being aware of it (thus denying b). In this sense, Alfred Mele (2009, pp. 37–39) claimed that in driving absent-mindedly one may be completely unaware of one’s signalling that nonetheless is an intentional action. ii. In the second scenario, the agent is aware of X but denies awareness of self-authorship; she is not aware of it as something she is doing. An example: Princess drops a handkerchief in order to catch Hero’s attention. Yet she is so ashamed that she denies her action to herself and experiences the dropping as a mishap, such that she is completely unaware that she is active in the dropping.1 This is not a case in which she lacks awareness that an event is a consequence of what she is doing, but she is not aware of doing anything at all. Someone accepting claim P in the sense specified by ii. has to say (a) one can be active in something one experiences as something that happens to oneself, and (b) such a doing cannot be an intentional action. Someone denying P as specified here has to defend that doing implies awareness of it, such that if princess was really not aware of her dropping, it was not her doing, after all. iii. In a third situation, the agent is aware that she is performing an intentional action, but not aware of what it is she is doing. The agent is not aware of the fact that what she is doing is X. I may for example be aware that I am sawing a plank and do so intentionally but not be aware that I am sawing an oak plank. If P applies here, as I am not aware that I am sawing an oak plank, sawing an oak plank is not an intentional action of mine. Yet sawing a plank is, and it seems that sawing an oak plank is nothing further that I’m doing, nothing over and above my intentional action of sawing a plank. What I am not aware of is that this identity obtains: that what I am intentionally doing (sawing a plank) is sawing an oak plank. I am aware of the action, but as something else. As an answer to the question “why?”, the unawareness-

234

Chapter Fifteen

claim in this situation denies awareness of the fact that this particular description applies to an—otherwise intentional and experienced—action. Whereas accepting P when interpreted in the sense specified by i. or ii. is rather demanding as it requires that I can do something without being aware that I am active in what is going on, it seems harder to deny P in iii. As such, even someone denying the possibility of unintentional doings can accept claim P when interpreted in the sense specified by iii.—as long as she accepts a view of action individuation allowing that a doing can be intentional under one description and not intentional under another.2 Denying P as specified by situation iii. means to hold that an action can be intentional under a description even though the agent is not aware that this description applies to her action. Example iii. is the example Anscombe uses. Her topic then is not some kind of zombie action one performs in complete absence of awareness but the scope of intention as far as an intentional action is concerned. The claim is that the action is not intentional per se but intentional in some sense or under a certain description. And it is intentional under a particular description only if the agent is aware of the action under this description.3 It is important to note the difference between these scenarios. They are, however, not completely unrelated. Accepting P in the sense specified by iii. does have consequences for P’s plausibility in i. and ii. as well, as becomes clearer if we describe all three cases in terms of actiondescriptions: in i. the agent is aware of the action under no description at all, in ii. under no description that depicts it as something she is doing. Acceptance of P in i. and ii. depends on further stands on a theory of action. For example, for someone denying the possibility of nonintentional actions or doings, accepting P in i. yields that the subject is not doing anything at all, and P does not apply. We can thus treat scenario iii. as basic for analyzing P because it has implications for the other scenarios but also because it can do with less theoretical presuppositions. For this reason, in the following analysis of the notion of awareness in P, I will concentrate on cases of the kind of scenario iii.

3. The notion of awareness at issue According to P, awareness is a necessary condition for intentional action. But awareness in what sense? What does it mean that I am aware of my doing? Does it require conscious thought (A)? Is there some special kind of phenomenal awareness that we have of our own actions (B)? Or is it simply a belief about my doing that is required (C)?

“I was not aware I was doing that”

235

A. Typically, authors arguing that we can perform actions or even intentional actions in the absence of awareness deal with a narrow notion of awareness that requires the agent to be “fully” aware of her action, for example in the sense of paying attention to it or having a conscious thought about it.4 As it seems obvious that we can perform many intentional actions without being fully aware of it in this sense, restricting the notion of awareness in this manner allows these authors to claim somewhat convincingly that the agent lacks “awareness” in a number of cases that are clearly intentional actions. It then seems that disagreement here is less about the nature of the case in question but about the notion of awareness. However, there is wide agreement that awareness in this narrow sense is not necessary for intentional action and thus cannot be the relevant sense of awareness in P. B. In recent years, several authors argued that intentional action does not require awareness of the action as conscious thought yet that the agent is phenomenally aware of her intentional actions in a particular kind of way—that has been called the phenomenology of action or the sense of agency (see Pacherie 2007, Horgan et al. 2003 or Bayne 2008). There is no general agreement on what is understood as the phenomenology of action. Yet in a nutshell, the general claim is that if I act intentionally, there necessarily is something it is like for me to perform this action at this moment. This does not require that I pay any attention to my action, think of it or even really notice that I am performing it, yet nonetheless my action shapes my general phenomenal state. It seems plausible that if an agent is phenomenally aware of her action in this sense, this involves some cognitive moment, some minimal kind of belief that she is performing the action in question, even though she may not in any relevant sense consciously self-ascribe it. For an illustration of the idea of phenomenal awareness of our actions, compare the following example: A man who is tapping his foot in time to music is only marginally or peripherally conscious of what he is doing. But he is still marginally or peripherally conscious of it. His tapping is not at all like the secretion of enzymes in his stomach, of which he is totally unconscious. (Searle 1991, p. 298. See also O’Shaughnessy 2008, p. 361)

C. A further possibility to understand awareness is as a simple belief. Understanding awareness in P in this sense is to say that one acts intentionally only if one believes that one is performing that action, whereas existence of this belief does not depend on the subject’s being

236

Chapter Fifteen

consciously or phenomenally aware of it (or of its content); no reflection or occurrent thought is required. It is generally held that one can believe e.g. that Bukarest is the capital of Rumania for years without thinking of it—it is enough one could if required. This kind of non-occurrent belief is also called dispositional belief. At first glance, this understanding of awareness as belief seems to outrival the competing positions: there is broader agreement about the nature of belief than about the nature of agentive phenomenology, and it seems to fit well with Anscombe’s question “why?”. However, I will argue that if we want P to be true awareness in P cannot be simple dispositional belief, but awareness as a necessary condition for intentional action should be understood in a stronger sense. I will not directly argue for such a stronger notion; what I will do in the rest of the paper is to show that a dispositional belief in the sense presented does not fully constitute awareness in P. Awareness as non-occurrent belief One author who argued in favour of a belief-account of awareness in P is Kieran Setiya. In a discussion of Anscombe’s Intentionality he claimed: It is important to stress that Anscombe’s doctrine does not require conscious belief or awareness. >…@ Once this is clear, it is tempting to say that one cannot do something intentionally in ignorance: without the belief that one is doing it. (Setiya 2007, p. 25)

In talking of Anscombe’s doctrine Setiya refers to her passage on the unawareness-claim. His statement that according to her intentional action does not require conscious belief or awareness is thus misleading, as Anscombe clearly claims that absence of awareness does exclude intentionality.5 I take it what he means to say is that we should not understand her message in a too demanding sense; it is not awareness in a narrow sense that is a necessary condition for intentional action, but all that is required under the title of awareness is a cognitive attitude—that I believe that I am performing the action in question. However, this leaves aside the third possibility above: phenomenal awareness. In what follows, I will argue that his argument does not show that it is belief, and only belief, that is necessary as far as awareness of our action is concerned, but that a stronger notion of awareness is required. To keep things clear, I will use “awareness” for the necessary condition for intentional action to be analysed, and “awarenessph” for phenomenal awareness. To defend a belief-account of awareness in P against a phenomenal one, Setiya would have to show that (i) absence of belief does exclude

“I was not aware I was doing that”

237

intentionality (i.e. P is true if we understand awareness as belief) and (ii) absence of awarenessph does not do so (i.e. P is not true if we understand awareness as awarenessph). Setiya (2007, p. 25) presents three examples in support of his position: 1) S has no idea that in humming Beethoven’s Ninth he is driving his wife crazy. 2) S is ignorant of the impatient tapping of his foot as he is poring over a draft. 3) S is not consciously reflecting on the fact that in turning on the light he is closing an electrical circuit. In all three examples, S is not aware in the narrow sense of something he is doing (that he is driving his wife crazy, tapping with his foot and closing a circuit). As this narrow sense is not necessary for intentional action, this by itself should not affect the intentionality of the action. The first two examples are supposed to be cases in which he further has no belief about doing the thing in question and they are not intentional; they support (i). In the third case however, Setiya claims S does believe that he is closing a circuit and is doing so intentionally even though he is not “consciously reflecting” on it. So this example is supposed to be a case in which a lack of awareness in a narrower sense does not affect the intentionality of the action as awareness in the sense of belief is present. It could be argued that Setiya only denies awareness in the narrow sense of conscious thought. Yet as I want to analyse whether the argument supports a belief-account further against a phenomenal account, I read this as denying awarenessph as well. Thus this is supposed to be a case for awareness in the absence of awarenessph and support (ii). Unfortunately, Setiya is very brief in describing his examples, and it is not easy to identify the basis for his claims that S is aware of the closing and the closing is intentional while the humming and the tapping in 1) and 2) are not. Why should we accept that S is aware of his closing the circuit? On Setiya’s account, S’s awareness of his closing the circuit is constituted by a belief. It seems plausible that S has a dispositional belief concerning the circuit in the following sense: if you were to ask S the moment he touches the switch whether he is closing an electrical circuit he would probably say yes without hesitation or need for reflection. On most standard accounts this is enough to grant him a belief that he is closing a circuit. However, the question directs his attention to the closing such that

238

Chapter Fifteen

the moment he gives the answer he certainly is aware of his closing of the circuit even in a narrower sense. The question is: Is the non-occurrent dispositional belief he has before the question is asked enough to grant him awareness of the action under the description in question? We should thus have a closer look on how 3) differs from the other two cases in order to better understand this dispositional belief about his action. A. One main difference between the three cases as presented by Setiya concerns S’s background information about the action. He not only has a disposition to occurrently believe that he is doing the thing in question when asked—which he has not in 1) and 2)—but he furthermore has stored beliefs about such cases in general. Whereas his wife never told him that she dislikes his humming, and he never realised that he has the habit of foot-tapping while reading, he once learned at school that to turn on a light means to close an electrical circuit. Thus he has a stored belief about circuits at his disposal that he has independently of the fact that he reflects on it or brings it to awarenessph. Should we grant S awareness of his closing of the circuit for this? I think we should not. The stored belief is not about his specific action. It is general in form, it is something like: “All turnings-on of lights are closings of electrical circuits.” But what is required here in order to claim that this closing of the circuit is intentional is awareness of the fact that this is a closing of a circuit, not a belief about circuits in general. What is required is that he realises that his stored belief applies to the case in question. The stored belief alone does not give us that. B. A further difference between the cases is the object S is unawareph of. Whereas in 1) and 2) he is unawareph that the event X is taking place at all (that his wife is getting crazy, that he is tapping with his foot), in 3) he is awareph of the event and his bringing it about, but under another description (turning on the light rather than closing a circuit). Thus 3) is an example of situation iii. above, whereas 2) is a situation-i.-case and in 1) the agent is unaware of a consequence of what he is doing. Should we accept awareness of the closing because it is one and the same doing as the turning on of the light? No. If awareness is constituted by a belief, awareness as belief shouldn’t be referentially transparent: from the fact that I have a belief that p and that p=q it does not follow that I have a belief that q. Thus, the identity alone cannot be enough to grant awareness, at least not unless S is aware of the identity as well.

“I was not aware I was doing that”

239

C. A combination of A and B may provide the best basis for accepting awareness about the circuit in 3): from B, we have awareness that he is turning on the light, and from A the stored belief that turnings on of lights are closings of electrical circuits. From this he could non-occurrently infer that what he is doing is closing an electrical circuit. In this sense he believes this even in the absence of awarenessph and in this sense it could be claimed he’s aware that he is doing so. Whether or not we find this case of an unconscious inference very persuasive on theoretical grounds, comparison with another example will show that it is mistaken and that we should not accept that there is awareness of the closing of the circuit. But before I come to this example we should analyse on what grounds we are supposed to accept that S not only believes that he is closing a circuit, but also does it intentionally. Why should we accept that the closing of the circuit is intentional? P posits the relation between awareness and intentionality as a conditional: if an action is intentional, the agent is aware of it. But it does not follow from P that if the agent is aware of something she is doing she also does this intentionally. Even if we accept awareness in 3), we still need a further argument to show that the action is also intentional under this description. There are two possibilities: either one can argue for the closing to be intentional on separate grounds, or one can argue that the relation between awareness and intentionality is a bi-conditional such that awareness implies intentionality. Setiya just stipulates that the closing of the circuit is intentional without giving any argument for it. All he explicitly claims about the relation between intentionality and awareness in this passage is that the former implies the latter. However, his claims about the case suggest that he is rather on the side of the supporters of a bi-conditional relation. The dispute between authors defending that awareness implies intentionality and others denying it is mainly about whether an action can be intentional under a description even though the agent does not intend to perform the action under this description. Bratman (1984, p. 400) for instance defends a bi-conditional relation between intentionality and awareness. He argues that if I am aware that my running a marathon wears down my sneakers and I run the marathon, wearing down my sneakers is intentional even though I do not intend to wear them down. He denies the intention because one would not engage in means-end reasoning on how to wear down the sneakers. But he claims the action is nevertheless intentional under this description because the agent is aware that this is what she is doing and does it nonetheless—no further desire or willing is

240

Chapter Fifteen

required. Searle (1983, p. 103), on the other side, claims that the fact that I am aware that some description applies to my action does not imply that I do it intentionally under this description. Someone like Searle, who claims that awareness does not imply intentionality, is thus someone defending that in order for an action to be intentional under a description the agent needs to intend it under this description (what Bratman dubs “the simple view”). Note that they both understand the claim that an agent intends to do something as meaning that the agent aims at performing the action under this description, what is shown by her means-end reasoning or her caring in case of a failure. Setiya claims that S is aware of the closing of the circuit and closes it intentionally. Thus, in order to know whether he defends a bi-conditional relation between awareness and intentionality, the relevant question is whether in turning on the light S intends to close the circuit. It seems that in the sense Bratman and Searle agree upon, we do not normally intend to close electrical circuits when we turn on lights. We would hardly reason about how to close the circuit if it weren’t for the light, and if there was a new technique invented for turning on lights without closing circuits, I doubt many people would think they failed to close one if they learn that they turned on a light without doing so. Yet if S does not intend to close the circuit and Setiya still argues that he intentionally closes it, he seems to be on the side of those defending that awareness implies intentionality. Note that this is not to say that intention is fully constituted by awareness but should rather be understood in the sense that awareness is a marker for intentionality such that whenever I am aware of my doing this doing is an intentional action. This is consistent with a view claiming that there are further necessary conditions for intentional action (namely some willing or desire-like motivational element) that happen to always cooccur with awareness in cases of intentional actions (this is in fact Setiya’s view; see e.g. Setiya 2007, p. 48). However this may be, Setiya’s claim that closing the circuit is intentional seems fully to depend on his claim that he is aware of the closing. Why we should not accept that S closes the circuit intentionally The example of the closing of the circuit is short and unfamiliar in this context, so that we may easily overlook certain features of it. To make those features more apparent I propose another example analogous to 3) but of a more familiar sort: 4) The electrical system in my house is old and weak. That is, it does not support many electrical devices at the same time. If I don’t turn

“I was not aware I was doing that”

241

out the freezer before I turn on the vacuum cleaner, the electrical system reliably breaks down. As this happened several times, I have a stored belief that to turn on the vacuum without turning out the freezer is to overload the electrical system. Yet it regularly happens to me that I turn on the vacuum without turning out the freezer. This example is analogue to Setiya’s third example with the difference that I added a bad consequence. As in 3) I am awareph of the action under one description (turning on the vacuum cleaner) but not awareph of it under another (breaking down the system). As in 3) I do have a stored belief about such cases in general and thus I could deduce that what I am doing is overloading the system. Further, I have a dispositional belief about that: if you ask me the moment I touch the button whether I am breaking down the system, I say yes without hesitation or reflection. Thus, analogue to example 3), in turning on the vacuum cleaner I should be aware that I am breaking down the system. And for all we said above, this should be enough to grant that I do it intentionally. However, I hope that this example makes it clear that we should not accept this reasoning: if on a particular account it turns out that I intentionally break down the electrical system in this situation, there must be something wrong with this account. One way to avoid the conclusion is to argue that 3) and 4) are not in fact analogue cases, as adding the bad consequence has it that in my example the dispositional belief is a sufficient reason not to perform the action (or: refers to such a reason). This is a relevant difference to Setiya’s case, where the belief does not interfere with S’s action in this way. For this, it could be claimed, it cannot really be true that I believe that I break down the system in turning on the vacuum as this would amount to irrationality. Yet even if we grant this it still remains open why the two cases should be treated differently—why is it that S’s belief is relevant and mine is not? Another way out is arguing that in 4) what is missing is not awareness but another necessary condition for intentionality—some kind of minimal desire-like motivational state towards the action that does not amount to intention in the sense introduced above but just to something like acceptance. However, this leads to an undesired multiplication of intentional actions: such an account would have to say e.g. that every time we speak we intentionally move around molecules of air, as we certainly have stored beliefs about this and accept it as well. I propose the two cases are in fact analogous and the best way to avoid the conclusion that I intentionally break down the electrical system in

242

Chapter Fifteen

example 4) is to deny awareness of the action under the relevant description in 4) as well as in 3). It seems plausible that I am not aware of my overloading the system in 4): to grant awareness in this case amounts to hold that in performing an action, I could be aware of my action fulfilling a description that I consider a sufficient reason not to perform the action—what seems to conflict with the idea of being a rational agent. But why were we tempted to grant S awareness of the closing of the circuit in 3)? Example 3) is very mundane. Normally, discussions about whether an action is intentional or not are related to questions of responsibility or guilt: cases in which something depends upon admitting or denying intentionality —killings, causings of harm or damage. In Setiya’s example, nothing depends on whether S’s closing of the circuit is intentional or not. For this reason, we may overlook distinctions we take to be important in other cases. One distinction that might be at issue is the distinction between responsibility and intentional action. As we hold S responsible for the closing of the circuit, we may be tempted to accept it as intentional. Yet this is mistaken, as there are cases of responsibility in the absence of intentional action— as e.g. cases of negligence. In a standard case of negligence, an agent brings about a condemnable result without being aware of it. Other than in a case of an accident, where she also brings about a wrong without being aware of doing so, in the case of negligence we hold her responsible for the bad result because on the basis of her beliefs and available information she could easily have realised that she was doing so. What we reproach to her is exactly that she did not realize it and thus was not aware that she was doing the thing in question and act accordingly, even though she could have known better. Cases without moral or criminal relevance—such as examples 3) and 4)—are usually not considered as cases of negligence. However, it is hard to deny that they share important features with negligence: as in negligence, the agents in these examples dispose of all the relevant information, yet for some reason or other fail to draw the required conclusion. It seems even though they have a dispositional belief about it in the sense that they have the disposition to occurrently believe it when prompted, they are not in the relevant sense aware of the fact that they are doing the thing in question. Example 4) makes it more apparent that a stored general belief alone cannot be enough to grant that I am aware of the action under the relevant description, as in this case being aware of it under that description would count as a sufficient reason not to perform the action.

“I was not aware I was doing that”

243

It then follows that all four examples are cases in which the agent lacks awareness that she is performing the action under the description in question and they are all not intentional under this description. Thus, it has not been shown that it is belief, and only belief, that is necessary as far as awareness of our action is concerned. Example 4) indicates that a stronger notion of awareness is necessary for intentional action. The most plausible candidate, I take it, is phenomenal awareness. However other options may be open here.

4. Conclusion This paper aimed at making some clarifications on the relation between awareness of one’s doing and intentional action. I started by assuming P, according to which absence of awareness of an action implies absence of intentionality. The first part of the paper was dedicated to a better understanding of the cases at the centre of interest in this context; cases in which an agent is doing something she is in some sense not aware of. I argued that the most uncontroversial examples are actions we are aware of under one description but not under another. That S was not aware of doing X, then, is to say that S was not aware of the fact that what she was doing was X. The second part of the paper dealt with the notion of awareness at issue in P. I proposed three general notions: conscious thought, phenomenal awareness and belief. Whereas there seems to be general agreement that the first cannot be required for intentional action, several authors argued for the third option. I discussed some examples by Kieran Setiya, who argues that belief is all that is required for intentional action under the heading of awareness, and I presented a further example to show that dispositional belief cannot be enough to grant awareness of the action under a description. In this case of quasi-negligence, the agent has a dispositional belief that she is performing the action in question in the sense that she has the disposition to occurrently believe it when prompted. But nonetheless she does not seem to be aware of the fact that this is what she is doing in the sense of awareness on which Anscombe’s dictum is true. It seems on the basis of these reflections that we should understand the notion of awareness at issue here neither as strong as conscious reflection nor as weak as simple dispositional belief.6

244

Chapter Fifteen

References Anscombe, E. 1957, Intention, Oxford: Blackwell, new edition 2000, Cambridge (MA): Harvard University Press Bayne, T. 2008, “The Phenomenology of Agency”, Philosophy Compass 3 (1), 182 Bratman, M. 1984, “Two Faces of Intention”, The Philosophical Review 93 (3), 375 Carruthers, P. 2007, “The Illusion of Conscious Will”, Synthese 159, 197 Goldman, A. 1971, “The Individuation of Events”, The Journal of Philosophy 68, 21, 761 Horgan, T., Tienson, J. and G. Graham 2003, “The Phenomenology of First-Person Agency“, in Walter, S. and H.D. Heckmann (eds.), Physicalism and Mental Causation: The Metaphysics of Mind and Action, Thoverton: Imprint Academic Mele, A. 2009, Effective Intentions: The Power of Conscious Will, Oxford: Oxford University Press O’Brien, L. 2003, “On Knowing One’s Own Actions”, in Eilan N. and J. Roessler (eds.), Agency and Self-Awareness, Oxford: Clarendon Press O’Shaughnessy, B. 2008, The Will, A Dual Aspect Theory, II, Cambridge: Cambridge University Press Pacherie, E. 2007, “Sense of Control and Sense of Agency“, Psyche 13, 1, 1 Searle, J.R. 1983, Intentionality: An Essay in the Philosophy of Mind, Cambridge: Cambridge University Press —. 1991, “Response: The Background of Intentionality and Action”, in Lepore E. and Van Gulick R. (eds.), John Searle and His Critics, Oxford: Blackwell Setiya, K. 2007, Reasons without Rationalism, Princeton, NJ: Princeton University Press Steward, H. 2009, “Sub-Intentional Actions and the Over-Mentalization of Agency”, in Sandis C. (ed.), New Essays on the Explanation of Action, Basingstoke: Palgrave Macmillan

Notes 1

I borrow this from a similar example by O’Brien (2003, p. 361). Contrast Goldman (1971) who claims that these are two different actions. 3 This does not imply that the content of the intention is given to the agent linguistically, but that she is aware of her doing in a sense that could be described in this way. 4 For example Carruthers (2007, pp. 208-9). 2

“I was not aware I was doing that”

5

245

This phrase is short for “absence of awareness that a description applies to one’s action excludes that the action is intentional under this description“. 6 I would like to thank Martine Nida-Rümelin, Jacob Naïto and the participants of the Monday Research Colloquium at the University of Fribourg for helpful comments and criticism of earlier drafts of this paper.

ABSTRACTS

Fernanda Samaniego, Causation and the Interventionist Vector of Explanatory Depth In the Interventionist Theory of Causal Explanation (Woodward 2003) three different criteria of explanatory depth can be identified. Intuitively, these criteria should work together: a shallow causal explanation should fail to meet the three criteria, while a deep causal explanation should successfully fulfill all of them. However, as it is shown in this paper, sometimes one criterion is clearly fulfilled while the other two are not—or they are fulfilled in a weak way. This will be illustrated with a medical example. The solution hereby proposed is to understand each criterion as one component of a “3-dimensional vector of explanatory depth”. It will be argued that this vector-notion of explanatory depth can actually make sense in the medical example analyzed here, and can be presumably generalized to multi-causal scientific explanations.

Alexandre Marcellesi, Explanatory Depth, Inference to the Best Explanation and Proportionality Not all causal explanations are created equal and some are better than others. But what are the factors the quality of a causal explanation depends on? It is to answer this question that James Woodward and Christopher Hitchcock have developed an account of what they call ‘explanatory depth’. David Harker has suggested that depth, as Woodward and Hitchcock define it, is a “peculiarly explanatory virtue” advocates of Inference to the Best Explanation can appeal to in order to flesh out their account. In this paper I argue that the notion Woodward and Hitchcock provide an account of is just predictive power by another name and is not properly seen as an explanatory notion. As a result, it is not a notion advocates of Inference to the Best Explanation should appeal to. I also argue that Woodward and Hitchcock’s account conflicts with the view that causal explanations are better when they cite causes that are ‘proportional’, in Stephen Yablo’s sense, to their effects, a view many—including Woodward—take to be plausible. If the arguments I develop are sound,

New Advances in Causation, Agency and Moral Responsibility

247

then there are good reasons to think that the account of explanatory depth developed by Woodward and Hitchcock is inadequate.

Ludovica Lorusso, Complex Causation in the Race Debate The paper investigates the issue of the use of self-identified racial categories in biomedical research. In particular the paper seeks to understand what are the causal factors that are differently distributed among racial groups in the U.S. and are likely to explain the differences in the occurrence of complex diseases among them. May the property of being of a specific self-identified race be considered as a projectible property for certain genetic properties causing a race-specific genetic susceptibility to complex diseases, or is this property more likely to be projectible for certain non-genetic properties causing a race-specific nongenetic susceptibility to complex diseases? By providing an answer to this question the paper shows whether and how the use of self-identified racial categories can be justified in biomedical research.

Lorenzo Azzano, Causal Sufficiency and Interferences I introduce and discuss the notions of necessity in causation, causal sufficiency, and interference—and prove that Mumford & Anjum’s argument against causal sufficiency is not to be easily overcome by the postulation of unbreakable processes at some fundamental level of causation. In order to postulate such unbreakable processes, one needs to make extremely robust claims on the nature of causation (such as the endorsement of some model of causation, or a choice in the structural properties of the relation such as transitivity) on the mathematical structure of space-time and even on the macro/micro properties debate. The conclusion is that causes hardly necessitate their effects; that however does exclude, as many anti-Humeans claim, an interesting modal value in causal processes taken individually that is not mere contingency.

Alessandra Melas, Absolute Coincidences and Salmon’s Interactive Fork Model It is a common opinion that chance events cannot be described in causal terms. Conversely, according to a causal conception of chance, chance events are the result of intersecting causal lines. More precisely, as Jacques Monod says in Le Hasard et la Nécessité, the intersections between independent causal chains are the origin of accidental events,

248

Abstracts

called absolute coincidences. Despite its importance, such a notion of chance tends to be neglected in contemporary literature and it seems to eschew a precise definition. The present study takes into proper consideration this causal conception of chance and tries to shed new light on it, providing a new account devoted to showing the strong relation between absolute coincidences and Wesley Salmon’s common cause model. The main attempt of this survey is to endorse the idea that not only is there a connection between chance intended as absolute coincidences and the Principle of Causality, but also that this causal notion of chance can be entirely described in terms of a causal criterion.

Elvira Di Bona, The Method of Contrast and the Perception of Causality in Audition The method of contrast is used within philosophy of perception in order to demonstrate that a specific property could be part of our perception. The method is based on two passages. I argue that the method succeeds in its task only if the intuition of the difference, which constitutes the core of the first passage, has two specific traits. The second passage of the method consists in the evaluation of the available explanations of this difference. Among the three outlined options, I demonstrate that only in the third option—as we shall see, the case of the scenario that remains the same but is perceived in two different ways by the same perceiver—the intuition purports a difference that possesses the necessary characteristics, namely being immediately evident and extremely complex and multifaceted, which determine its tensive nature. The application within auditory perception of this third option generates two cases, a diachronic one and a synchronic one, which clearly show that we can auditorily perceive causality as a link between two sonorous episodes. The causal explanation is the only possible explanation among those evaluated within the second passage of the method of contrast.

Carolina Sartorio, Actual-Sequence Freedom “Actual-sequence” views of freedom are views according to which freedom (the kind of freedom that matters to moral responsibility) is exclusively a function of actual sequences. These views are motivated by “Frankfurt-style” examples: examples that seem to show that freedom is not a matter of having alternative possibilities of action but, instead, is exclusively a function of the actual sequence of events. In this paper I first offer an interpretation of the actual-sequence claim in terms of a

New Advances in Causation, Agency and Moral Responsibility

249

supervenience claim about actual causal histories. I then go on to discuss some apparent counterexamples to that claim. I argue that the counterexamples can be disarmed by focusing on the role that the concept of causation plays in actual-sequence views and, in particular, by appeal to certain important connections that exist between causation (or the relevant concept of causation) and moral responsibility. I argue that, far from undermining the actual-sequence view of freedom, the examples in fact support a view of that kind.

Daniele Santoro and Marcello Di Paola, Causes and Blame: A Pragmatic Analysis Social sciences have made several attempts to dispense intentionality from the explanations of social phenomena. In this paper, we contend this view from a pragmatist perspective, partly inspired by Ludwig Wittgenstein and Robert Brandom. We argue that there is no distinction in principle between causal and intentional vocabularies, but that they rather presuppose a common discursive practice in which the individuation of a cause is possible only within the context of blaming responsibility for actions. In the second part of the paper we offer an account of how this social practice works and propose a pragmatic analsysis of the relationship between causes and blame within a broadly inferentialist framework, arguing that intentional and causal statements differ only in their inferential role in the social practice of scapegoating causes.

Federico L. G. Faroldi, Responsibility Regardless of Causation This paper deals with the relationships between causation, responsibility and normativity, both conceptually and in some contemporary philosophical literature. I consider and reject the claims according to which responsibility (i) is ascribed on a causal basis, and (ii) is therefore descriptive. I maintain that responsibility does not necessarily require causality to be ascribed, nor it is only descriptive. I offer two arguments to sustain my thesis: (i) there already exist several non-causal forms of responsibility; (ii) causation is not completely descriptive, but subsumes an evaluative part.

250

Abstracts

Sofia Bonicalzi, Skepticism and Control The appropriateness of responsibility judgments including a reference to basic desert is one of the main issues at stake in the debate between compatibilists and incompatibilists. If determinism obtains, moral judgments of praise and blame might be considered unfair, but indeterminism does not seem to be a better ally in order to defend the consistency of moral responsibility attributions. The first section of the paper—through the discussion of Kane’s Ultimate Responsibility and Fischer’s Semicompatibilism—analyzes the minimal requirements for consistent praise and blame attributions in terms of sufficient control and structure of the sequence of action, presenting Pereboom’s skeptical outcome as a coherent provisional standpoint given the credibility of our best scientific theories but in absence of a real disproof of agential views. The second part deals with the consequences of the skeptical solution on the system of morality and defends the consistency of a forward-looking and revisionist conception of morality as an answer to Haji’s objection, according to which determinism (absence of freedom to do otherwise) subverts deontic notions of rightness and wrongness.

Michael Brent, Understanding Strength of Will Richard Holton has presented an important criticism of two prominent accounts of action, a criticism that employs a notion of strength of will. Holton claims that these well-known accounts of action cannot explain cases in which an agent adheres to the dictates of a previous resolution in spite of a persistent desire to the contrary. In this paper, I present an explanation and defense of Holton’s criticism of these accounts of action, and then I argue that while Holton highlights a crucial deficiency in both, his own explanation of strength of will is problematic.

Irene Bucelli, Responsibility and Narrative Agency Reflective endorsement theories of agency distinguish full-blown actions, resulting from the agent’s self-governance, from other kinds of activities that even some non-human animals can perform, despite lacking our cognitive reflective capacities. I take it that reflective endorsement theories seem particularly promising in having an upshot for the debate on responsibility, and, within this framework, I believe that David Velleman’s theory has significant advantages. I explore some ways in which the enterprise of establishing the conditions for agency in these

New Advances in Causation, Agency and Moral Responsibility

251

terms can bear on responsibility. My very general question will be revolving around what consequences do reflective endorsement theories of agency have for responsibility. For the ambitious the project might sound, my aims will actually be quite modest. I limit my analysis to Velleman’s narrative model of agency. Moreover, I will only discuss a simple (simplistic) argument for a direct connection between agency and responsibility. By dismissing it I hope to point at the kind of questions that a reflective endorsement theory has to answer in order to have relevant consequences for the work done in responsibility.

Sarah Songhorian, Empathy and Concern for Others The purpose of this work is that of showing the complexity lying behind the concepts of empathy, sympathy, and concern and of showing how these concepts can help us understand the passage from descriptive premises to normative conclusions. I will operate some distinctions within what is usually considered, broadly speaking, as empathy in the literature, so that a more narrow concept of our basic ability can emerge. Thus, from the very basic form of empathy that infants possess, I will show how we can come to a more cognitive and complex ability: sympathy as a metaethical mechanism enabling us to judge a posteriori both other people’s behavior and our owns and to decide a priori—by means of the ideal spectator—whether this conduct is appropriate or not. Disentangling different meanings of “empathy” and “sympathy”, I will try to show that the former is an embedded capacity, while the latter is such a metaethical tool, so that the step from “empathy” to “sympathy”, and consequently to respect, concern, and care, will prove useful to understand the “is/ought question”.

P. Roger Turner, Truth and Moral Responsibility Most philosophers who study moral responsibility have done so in isolation of the concept of truth. Here, I show that thinking about the nature of truth has profound consequences for discussions of moral responsibility. In particular, by focusing on a very obvious fact about truth—that truth depends on the world and not the other way around—we can see that widely accepted counterexamples to a key inference rule in an important incompatibilist argument can be shown not only to be unsuccessful, but also impossible.

252

Abstracts

Franziska Müller, “I was not aware I was doing that” Elizabeth Anscombe famously claimed that if a person denies awareness of something she is doing, what she is doing cannot be an intentional action of hers. In this paper, I want to analyse how we should understand this claim. The first part deals with the object the person is unaware of— what are things an agent can do in the absence of awareness? The second part deals with the notion of awareness at issue in this context; the sense of awareness on which Anscombe’s dictum is true. I will argue that we should understand the notion of awareness at issue here neither as strong as conscious reflection nor as weak as simple dispositional belief.

CONTRIBUTORS

Lorenzo Azzano got his M.A. at the Vita-San Raffaele University (Milan) and he later studied as an external visiting student at the Corpus Christi College, University of Oxford. His main interests are in metaphysics, and more precisely in the cluster of issues regarding dispositions, causality, modality and laws of nature. Sofia Bonicalzi graduated in Philosophical Sciences from the University of Milan in 2011. She works on the topics of free will and moral responsibility. Currently she is a PhD candidate in Philosophy at the University of Pavia and a Visiting Ph.d. Student at Cornell University. Michael Brent completed his Ph.D. at Columbia University in 2012 and is an Assistant Professor of Philosophy at the University of Denver. Irene Bucelli is completing her PhD at King’s College London, where she is also Graduate Teaching Assistant in Modern Philosophy and 19th Century Continental Philosophy. Her main interests are in the philosophy of mind and her research focuses on action theory. Elvira Di Bona has recently got her PhD in Philosophy and Cognitive Sciences at the Institut Jean Nicod (Ecole des Hautes Etudes en Sciences Sociales), Paris, and at the Vita-Salute San Raffaele University, Milan. Her main research area is philosophy of perception, and auditory perception in particular. Marcello Di Paola is Research Fellow at LUISS University, Rome. He works on moral and political philosophy, applied ethics, and Wittgenstein. He has authored two books on environmental ethics. His articles have appeared in Environmental Values, Global Policy, and the Journal for the Study of Religion, Nature, and Culture. Federico L. G. Faroldi (University of Pisa, University of Florence) graduated in Philosophy at the University of Pavia (Italy), and studied at NAU (USA) and Trinity College, Dublin. He is the author of The Normative Structure of Responsibility. Law, Language, Ethics (College Publications, London).

254

Contributors

Ludovica Lorusso is currently a Post-Doctoral Fellow at the University of Sassari (Italy). She earned her PhD in Philosophy at the University of Padua and her M.S. in Biology at the University of Milan. Some of her notable publications are The Justification of Race in Biological Explanation, Clustering Humans: on Biological Boundaries, and Visual Judgments of Kinship: An Alternative Perspective. Alexandre Marcellesi is a PhD student at the University of California-San Diego. He is writing a dissertation on topics in the philosophy of causation and causal inference under the supervision of Nancy Cartwright. Alessandra Melas is a post-doctoral fellow in Philosophy at the University of Sassari. After receiving her PhD in Philosophy, defending a dissertation on the foundations of Quantum Mechanics, she spent one year as a visiting scholar at the Philosophy Department at Rutgers University, in New Jersey. Her main research interests and publications are in the areas of causation, chance and coincidences. Franziska Müller is a PhD student in Philosophy at the University of Fribourg, Switzerland, where she worked in a research project on the philosophy and phenomenology of action. She is currently teaching assistant in law at the University of Lucerne, Switzerland. After Fernanda Samaniego graduated in Physics and completed two master degrees in Philosophy of Science (at the National Autonomous University of Mexico UNAM, and at the London School of Economics), she obtained her PhD with European Mention at the Complutense University of Madrid, defending a Thesis on Interventionist Explanations of Entropy Rising. She is currently postdoctoral fellow at UNAM. Daniele Santoro teaches Political Philosophy and Philosophy of Social Sciences at Luiss University of Rome. His work focuses on responsibility and causation, distributive justice, and the epistemic aspects of rights. He has published on these topics for Philosophia, Philosophical Topics, an in edited volumes for Routledge and Continuum. Carolina Sartorio is Associate Professor of Philosophy at the University of Arizona. She received her PhD from MIT in 2003, and taught previously at the University of Wisconsin, Madison. She has published articles on causation, moral responsibility, agency, and free will, and is currently working on a book manuscript entitled Causation and Free Will.

New Advances in Causation, Agency and Moral Responsibility

255

With a background in Philosophy, Sarah Songhorian is currently a PhD student in Cognitive Neuroscience and Philosophy of Mind. Her research focuses on the possibility to find a relation between the concept of “empathy”, as a basic and embedded capacity, and that of “sympathy”, as a metaethical tool. P. Roger Turner is a PhD candidate at the University of Tennessee, Knoxville where he specializes in metaphysics, philosophy of action, and philosophy of religion. His dissertation is a defense of the so-called Direct Argument for incompatibilism about moral responsibility and causal determinism.

INDEX OF NAMES

Absher D.M., 37 Adelman L., 43 Agosta L., 202 Aharon-Peretz J., 194 Aldrich M.C., 32, 35 Anjum R., 50, 59, 61 Anscombe E., 120, 231, 234-236, 243 Arendt J., 141 Aristotle, 77, 170, 183, 192 Armstrong D.M., 80 Azzoni G.M., 141 Bach K., 120 Bacon F., 121 Bamshad M.J., 38 Barbujani G., 36 Baron-Cohen S., 206 Batson C.D., 194, 196, 203 Baumeister R., 178 Bayne T., 235 Beebee H., 143 Belle E., 36 Benjamin M., 141 Björnsson G., 159 Bobzien S., 141 Bok H., 147 Bolnick D.A., 37 Bookheimer S.Y., 206 Botvinick M., 204 Braham M., 141 Brandom R., 113-114, 120 Brandt R., 177 Brass M., 158 Bratman M., 177, 191, 239-240 Bratslavsky E., 178 Bucelli I., 192-193 Burchard E., 32, 34, Butterfill S., 79

Bylsma L.M., 204 Campbell J.K., 228, 230 Cappelen H.W., 120 Carpenter A., 194, 203 Carr L., 204 Carruthers P., 244 Caruso G., 141 Castellanos E., 43 Celano B., 203 Chakravartty A.,61 Clark A.G., 38 Clarke R., 147, 149-150 Cohen J.D., 204 Colangelo L.A, 35 Coleman J., 127 Collins F.S., 43-44 Colonna V., 36 Coplan A., 194-195 Copp D., 158 Corander J., 37 Corlett J.A., 141 Cournot A.A., 77 Cox N.J, 43-44 Coyle N., 32, 34 Craver C., 27 Dalgleish T., 206 Daly M.J., 48 Dapretto M., 196, 206 Darley J.M., 204 Darwall S.L., 205 Davidson D., 120, 177, 191 Davies M.S., 206 De Caro M., 158 De Vignemont F., 195, 200 de Waal F.B.M., 194 Decety J., 194 Dennett D., 147, 150-151, 192

New Advances in Causation, Agency and Moral Responsibility Dubbink W., 142 Dubeau M.C., 204 Duff A., 123 Dulin L., 194, 203 Duncker K., 79-80 Duster T., 37 Early S., 196 Ebers G.C., 44 Eisenberg N., 196, 201 Elster J., 120 Epstein R.A., 142 Escalas J.E., 201 Fabian S.A., 204 Faroldi F.L.G., 141, 143 Feder E.K., 205 Feinberg J., 127, 141 Finch A., 148 Fischer J., 103-105, 141, 144-145, 147-148, 152-153, 156, 158, 192, 218-221, 228-229 Frankfurt H. G., 94-96, 104, 143, 147, 150-152, 157, 191-192 French P.A., 142 Fullerton S.M., 38 Fullwiley D., 37 Gale S., 194, 203 Gallagher S., 192 Gallese V., 195 Garrett J.E., 142 Gilbert M., 141 Gilligan C., 205 Ginet C., 104 Goldie P., 194-195 Goldman A., 244 Goldstein D.B., 38 González E., 142 Gordon R.M., 206 Graham G., 235 Graham K., 141-142 Greene J.D., 204 Grice H.P., 80 Haavelmo T., 3

257

Hacking I., 77 Haidt J., 204 Haji I., 148, 156-157, 218, 221-222, 228 Hall N., 98, 105 Handel A.E., 44 Harjusola-Wevv S., 194, 203 Harker D., 18, 22-24, 29 Harman G., 22 Hart H.L.A., 121, 130-132, 141 Hassan O., 194, 203 Haynes J.D., 158 Hedstrom P., 120 Heinze H.J., 158 Held V., 205 Hempel C., 31 Hennessey K., 107, 120 Hirschhorn J.N., 48 Hitchcock C., 18-26, 30-31, 98 Holton R., 165-178 Honoré T., 121-122, 130-132 Horgan T., 235 Hornsby J., 191-192 Huemer M., 61 Hume D., 143, 148, 165-168, 170, 174-175, 178, 188, 193, 204, 209, 217 Hunt L.M., 34 Husserl H., 91 Hutto D., 192 Iacoboni M., 196, 204, 206 Isaacs T.L., 141 Jackson F., 86, 93 Jeffries M., 43 Jha A.P., 204 Joyce R., 203-204 Julienne Junkersfeld M., 77 Kane R.,147, 149-150 Kant I., 113-115, 120, 130, 189, 192-193, 205 Kearns S., 228 Kelsen H., 121, 130 Kennett J., 177

258

Index of Names

Kittay E.F., 205 Koksvik O., 82 Korsgaard C., 191-193 Kovitz B., 61 Kriebel D., 39, 43 Kriegel U., 82-83, 86-87, 93 Kumar R., 32, 35

Nahmias E.A., 141 Nelkin D.K., 141, 148 Nichols S., 145 Nida-Rümelin M., 93 Novembre J., 38 Nozick R., 144 Nystrom L.E., 204

Lamm C., 194 Lee S.S., 37 Lenzi G.L., 204 Lepore E., 120 Lewis D., 80 Lewontin R.C., 36 Li J.Z., 37 Libet B., 158 Lishner D.A., 194, 203 List C., 26, 31 Long J.C., 37-38

O’Brien L., 244 O’Connor T., 149 O’Shaughnessy B., 233, 235 Oppenheim P., 31 Ossorio P., 37 Owens D., 78

Mackonis A., 18, 22 Mäkelä P., 141 Manolio T.A., 43-44 Martin C.B., 53 Martin M., 43 Marttinen P., 37 Mazziotta J.C., 196, 204 McGinn C., 158 McKenna M., 144, 228 Megyesi M.S., 34 Mele A., 158, 177, 233 Meltzoff A.N., 196 Menzies P., 26, 31 Merricks T., 215, 225, 228, 230 Michotte A., 80-81 Miller S., 141 Monod J., 63-67, 73-74, 77-78 Moore A.K., 196 Moore G., 141 Moore M.S., 121, 123, 128-130, 132-136, 142-143 Morgan D.K., 44 Morris Ch.W., 120 Mumford S., 50, 59, 61 Muraven M., 178

Pacherie E., 235 Peacocke C., 192 Penner A.M., 33 Pereboom D., 153, 155, 158-159 Perry D., 194 Pettit P., 177 Pfeifer J.H., 196, 206 Plato, 217-218 Posner R.A., 142 Preston S.D., 194 Prinz J., 208 Pritchard J.K., 36-37 Prkachin K.M., 204 Ramagopalan S.V., 44 Ravizza M., 103-105, 145, 147, 156, 192, 218-221 Rawls J., 205 Reichenbach H., 68-71, 78 Ricoeur P., 192 Risch N., 32 Risser D.T., 141-142 Romejin J.W., 13 Rosenberg A., 120 Rosenberg N.A., 36-37 Roskies A.L., 158 Rothman K.J., 39 Roughley N., 194-195 Royal C.D., 38 Russell B., 50, 61, 110 Russell P., 143, 148

New Advances in Causation, Agency and Moral Responsibility

259

Swedberg R., 120 Salmon W.C., 63-78 Salvarini G., 196 Sampat B., 194, 203 Saperstein A., 33 Sartorio C., 98, 105, 129, 135, 147 Scanlon T., 154 Scheler M.F., 196 Schnall I.L., 228 Scott A.A., 206 Searle J.R., 235, 240 Seibold M.A., 32 Setiya K., 236-243 Shabo S., 228 Shamay-Tsoory S.G., 194 Sheehy P., 141 Sher G., 192-193 Shoemaker D., 141 Shoemaker S., 192 Siegel S., 79-85, 89-91, 93 Sigman M., 206 Silver D., 141-142 Simner M.L., 196 Singer T., 195, 200 Slatkin M., 44 Slote M., 194, 208 Smart J.J.C., 147 Smedley B., 43 Smilansky S., 158 Smiley M., 141 Smith A., 201-203 Smith J., 141 Smith M., 177 Soares C., 141-142 Solomon P.E., 204 Sommerville R.B., 204 Soon C.S., 158 Stapleton J., 143 Stern B.B., 201 Steward H., 233 Stocks E.L., 194, 203 Strawson G., 158 Strawson P., 120, 147, 155-156 Strayer J., 196 Stump E., 229 Sullivan S., 44

TallBear K., 37 Tang H., 32, 37 Templeton A.R., 34, 36 Thakur N., 43 Thomas B., 107, 120 Tice D.M., 178 Tienson J., 235 Tognazzini N.A., 144 Tollefsen D., 141 van Hees M., 141 Van Inwagen P., 96-97, 104, 147148, 214, 217, 227 van Loo H.M., 13 Varzi A., 120 Velasquez M., 141-142 Velleman J.D., 177, 180-182, 185187, 189, 191-193 Vineis P., 39, 43 Von Wright G.H., 120 Vranas P.B.M.., 158 Waldmann P., 37 Wallace R.J., 154 Warfield T.A., 229 Watson G., 147 Weber J.L., 36-37 Weiss K.M., 37 Welch J.R., 142 Weslake B., 22 Whitelaw E., 44 Widerker D., 228 Williams G., 141 Williamson T., 142 Wilmot S., 142 Wittgenstein L., 106, 109-112, 115116, 124, 126, 141-142 Wood A.W., 205 Woodward J., 1-6, 10, 13-15, 18-26, 30-31, 120 Wright R.W., 142 Yablo S., 26, 28-29, 105

260 Zhu J., 44 Zimmerman M.J., 144

Index of Names Ziv E., 32, 34