persons and causes: the metaphysics of free will

Persons and Causes
TIMOTHY O ' C O N N O R
OXFORD UNIVERSITY PRESS
OXFORD UNIVERSITY PRESS
Oxford New York Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
Sao Paulo Shanghai Singapore Taipei Tokyo Toronto
Copyright © 2000 by Timothy O'Connor
First published in 2000 by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016
www.oup-usa.org
First issued as an Oxford University Press paperback, 2002
Oxford is a registered trademark of Oxford University Press
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data O'Connor, Timothy, 1965-
Persons and causes : the metaphysics of free will / Timothy O'Connor.
p. cm. Includes bibliographical references and index.
ISBN 0-19-513308-0; 0-19-515374-X (pbk.) 1. Free will and determinism. I. Title.
BJ1461.027 1999 123'.5—dc21 99-20501
1 3 5 7 9 8 6 4 2
Printed in the United States of America on acid-free paper
This page intentionally left blank
Acknowledgments
Much of this book was composed in St. Andrews, Scotland, in 1996–1997 while 1 was supported by a Gifford Research Fellowship from the University of St. An- drews and by a fellowship from the Pew Scholars Program. I am grateful to both these institutions for their support. I thank the members of the Departments of Logic and Metaphysics and of Moral Philosophy in St. Andrews for their hospital- ity and fruitful philosophical discussion—in particular, Professors Crispin Wright and John Haldane.
I have used portions of previous articles of mine, more or less substantially revised, in the present work: "Indeterminism and Free Agency: Three Recent Views," Philosophy and Phenomenological Research 53 (1993); "Thomas Reid on Free Agency," Journal of the History of Philosophy 32 (1994); "Emergent Proper- ties," American Philosophical Quarterly 31 (1994); "Agent Causation," in Agents, Causes, and Events: Essays on Indeterminism and Free Will, ed. T. O'Connor (New York: Oxford University Press, 1995); and "Why Agent Causation?" Philosophical Topics 24 (1996; published by Arkansas University Press). I thank the publishers for permission to incorporate that material here.
I have read material ancestral to parts of this book to audiences at meetings of the American Philosophical Association and at the following universities: Arkansas, Brigham Young, Cornell, Edinburgh, Free (Netherlands), Glasgow, Indiana, Notre Dame, Purdue, Seattle Pacific, and St. Andrews. Other philosophers have helped me through private conversation or correspondence. I thank in particular the following: David Armstrong, Robert Audi, Roderick Chisholm, Mark Crimmins, Daniel Dennett, Fred Dretske, Laura Ekstrom, John Martin Fischer, Stewart Goetz, Anil Gupta, Bob Hale, William Hasker, Chris Hill, Al Howsepian, Jaegwon Kim, Keith Lehrer, Barry Loewer, David McCarty, Brian McLaughlin, Michael Murray, John O'Leary-Hawthorne, Alvin Plantinga, David Robb, Gregg Rosenberg, William Rowe, Tom Senor, Sydney Shoemaker, Michael Slote, Eleonore Stump, Michael Tye, Peter Unger, Peter van Inwagen, Rene van Woudenberg, Ted Warfield, David Widerker, and Timothy Williamson.
VIII ACKNOWLEDGMENTS
Now for some special appreciation. I thank my wife, Gail, and our three children, Brian, Laura, and Lindsay, for their love and joyful companionship.
I thank Jonathan Bennett and Samuel Gorovitz, whose article "Improving Aca- demic Writing," Teaching Philosophy (June 1997), struck me to the quick at just the right moment—shortly after I had completed a first draft of this book. The result was a couple of weeks of revising what I had written, line by line. I am not confident that they would be satisfied by my efforts, but what is in your hands is a far more readable text than that which preceded it. (An anonymous reader also deserves credit in this regard.)
For help in thinking about the problem of free will, I especially thank Ran- dolph Clarke and Robert Kane, with whom I have corresponded over several years, and above all Carl Ginet, who directed my original Ph.D. thesis on this topic at Cornell University and has been a friendly and trenchant critic ever since.
Finally, I acknowledge my great debt to another former teacher, Norman Kretzmann. He inspired in me a love of philosophy, taught me to see its problems through its history, attended meticulously to my work, and extended love and kindness to me in the manner of a father. Sadly, it is a debt I can no longer hope to repay, as Norman died shortly before this book went to press.
May 1999 T. O. Bloomington, Indiana
Contents
Introduction xi
1. Freedom and Determinism 3 1.1 An Ancient Argument 3 1.2 Some Modal Principles and the Argument for Incompatibilism 5 1.3 The Fixity of the Past and of Natural Laws 15 1.4 Freedom and Responsibility 18
2. Freedom and Indeterminism: Some Unsatisfactory Proposals 23 2.1 The Trouble with Incompatibilism 23 2.2 Simple Indeterminism: Carl Ginet on Choice and Control 24 2.3 Causal Indeterminism: The General Strategy and a Problem Posed 27
3. The Agent as Cause: Reid, Taylor, and Chisholm 43 3.1 The Agency Theory 43 3.2 Thomas Reid 43 3.3 Richard Taylor 49 3.4 Roderick Chisholm 55 3.5 Summary 60 Appendix: Chisholm's Later Writings on Agency 61
4. The Metaphysics of Free Will 67 4.1 Overview 67 4.2 Event Causation 68 4.3 Agent Causation 71 4.4 C. D. Broad's Objection to the Very Idea of Agent Causation 74 4.5 Remarks on a Contemporary Alternative Account of
Agent Causation 76 4.6 Ersatz Agent Causation? 79 4.7 Alternative Possibilities, Responsibility, and Agent Causation 81
X CONTENTS
5. Reasons and Causes 85 5.1 Reasons Explanation and the Agency Theory 85 5.2 Objections to the Account 86 5.3 Reasons and Contrastive Explanation 91 5.4 Reasons and Tendencies to Act: A Residual Problem and
a Speculative Proposal 95 5.5 When Is the Will Free? 101 5.6 Conclusion 107
6. Agency, Mind, and Reductionism 108 6.1 Introductory Polemics: 'The Emerging Scientific Picture of
the World' 108 6.2 Emergence 110 6.3 Emergence and Consciousness 115 6.4 The Emergence of Active Power 121 6.5 An Epistemological Objection to Agent Causation 123 6.6 Conclusion 125
References 127
Index 133
Introduction
The topic of this study is one of the oldest, most contentious, and most difficult topics in philosophy. That it should prove to be all of these things is itself very puzzling, at first glance. For the goal is merely to make explicit our everyday picture of ourselves as agents who adopt specific purposes in freely choosing how we shall act, choices that trigger and help sustain our actions. If virtually all of us think of ourselves as freely acting, purposive beings—when out living our lives, if not always when reflecting on the matter in our studies— why haven't philosophers of the past managed to bequeath to us a perspicuous and immediately recog- nizable articulation of that thought? Granted, whether any such philosophical account answers to the facts of the springs of ordinary human behavior is an open empirical question: the truly puzzling matter is that there should remain deep controversy over what empirical researchers should be looking for to answer the question.
Like other enduring philosophical conundrums, the problem of understanding the idea of free, purposive, responsible activity (free will) is difficult in part because it touches on other fundamental ideas: causation, explanation, and the nature of intentional states such as beliefs, desires, and purposes—just to start. Another source of difficulty is that we tend to assume that free agency admits of degrees. Philosophers not only disagree about the scope of free will in ordinary human beings (as opposed to God). They also dispute how genetics and environment influence a person's freedom of action. That is, once one fixes a basic concept of free will, it remains puzzling how free will so understood can be qualified.
The two features of the problem I've just mentioned pose difficulties for the individual philosopher in coming to his own view on the nature of free will (and so, a fortiori, for the community of philosophers to come to a shared view). A final obstacle directly to the goal of consensus stems from the fact that this project is rarely pursued in isolation. Philosophers want to do more than paint a com- monly held picture of ourselves. We also want to put forth a vision of human beings and their place in the wider scheme of things. And for some (no doubt
xi
Xii PERSONS AND CAUSES
most, to some degree), this will be a vision that competes with the conventional one. Clearly, our view of whether human beings act freely—and to what degree—is right at the center of our wider philosophical vision. Freedom of will is directly connected to the possibility or significance of moral responsibility, auton- omy, the uniqueness of persons (involving creativity, originality, and their life his- tories in general), dignity, love, and friendship. In short, it is connected to every- thing that fundamentally matters to us in our relationships with one another.1
There is at present a widespread trend in English-speaking philosophy toward 'deflationary' analyses of most traditional targets of philosophical inquiry. Al- though this project is all the rage, carried out under the banner of "Philosophical Naturalism," it is itself rarely made the subject of explicit articulation—what precisely qualifies and why it should be undertaken in the first place. The general, if vague, impetus is to analyze philosophical notions in a way that makes them hospi- table to a 'naturalistic' view of human beings that has apparently been handed down to us by 'Science.'
My own approach to the subject of free will in this work is quite counter to the fashion for apologies on behalf of 'Naturalism.' I am driven in the first place to clearly understand the prephilosophical view of human agency, let the chips fall where they may concerning its compatibility with 'Naturalism.' But I am no more satisfied with mere conceptual analysis than are Naturalists. I, too, want to understand human nature as it is—its frailties, as well as its glories—and recognize that empirical work in relevant branches of psychology and biology will, in the end, have quite a lot to say about that. But we can assess the significance of the verdict of the relevant science—when it comes—only if we have already reflected on our values and their requirements. It is 'wretched subterfuge' to settle on a scaled-back notion of what we value about ourselves and then declare victory when—surprise!—it is patently clear on a little reflection that no empirical work is ever likely to undermine that conception.
So how is it that we prereflectively think of ourselves? Chapter 1 constitutes a first step toward an answer to this question. In it, I consider the issue generating the most basic divide among theorists of free will: roughly, whether or not free will is compatible with the thesis that human choice and consequent behavior is a causally determined outcome of antecedent factors in and around the agent. In contemporary philosophy, this perennial debate has centered around the validity of a certain 'modal' style of argument for incompatibilism that turns on principles concerning the logic of 'unavoidability' (as in 'Brian's kicking his sister was unavoidable for him at the time, given his unfortunate affliction with sibling aggres- sion syndrome'). Although the incompatibilist argument is easily stated, a number of subtle issues connected to it have been insufficiently understood by previous discussants. My exploration of these matters suggest two conclusions: (1) some current formulations of the incompatibilist argument are clearly invalid, but (2) they are naturally repaired in a manner that restores intuitive assent. I then defend
1. The most comprehensive recent discussion by a philosopher of these conceptual interconnec- tions is chapter 6 of Robert Kane's excellent book The Significance of Free Will (New York: Oxford University Press, 1996).
Introduction xiii
the validity of my favored version of the argument against common compatibilist replies.
If, as I contend, free will requires that our choices not be causally determined, then a positive account of it must show how the causal 'loose fit' between antecedent factors (including the agent's having reasons for various alternatives) and choice does not reduce to an injection of mere randomness in the springs of action. One needs to show how certain causally indeterministic mechanisms would confer (or contribute to) a kind of power or control over one's own choices that would be lacking in any sort of deterministic scenario. I consider, in chapter 2, four such accounts that I believe to be wanting. I first consider the broad position of simple indeterminism, in which the agent's control over, or determination of, his own choices is held to be entirely noncausal in character and is instead a consequence of intrinsic, noncausal features of the choice itself. I argue against this position in the course of examining the account of its most prominent proponent, Carl Ginet. I then turn to the thesis of causal indeterminism, in which the agent's control over his own choices resides in the indeterministic ('chancy') causal efficacy of his reasons for so choosing. The idea here is that in every free choice, several options have a nonzero probability in the circumstance, and the actual outcome in every case will be caused by factors that prominently include the agent's reasons for so choosing. A prima facie problem for this position is to explain how the agent directly controls the outcome in a given case. There are objective probabilities corresponding to each of the possibilities, but within those fixed parameters, which choice occurs on a given occasion seems, as far as the agent's direct control goes, a matter of chance. I examine three versions of causal indeterminism that try to overcome this objection. Robert Nozick's strategy is to characterize choice as 'self- subsuming,' in that one's choice may itself conform to the very values reflected in the choice. Storrs McCall argues that the intentional explanation of choice is independent of any probabilistic causal explanation. Finally, Robert Kane focuses on the deliberative process that gives rise to (what he takes to be) paradigmatic cases of freedom of will—cases in which the agent struggles to prioritize conflicting values and desires. Kane suggests that the agent's control over the outcome consists in (1) the agent's close identification with each of the conflicting sets of motiva- tions and (2) the activity of the 'self-network,' a stable network of values, prefer- ences, and so on that constitutes the agent's character at the time of acting. I argue that all of these strategies fall short, despite the considerable ingenuity each displays (and the genuine insights on particular issues that one can glean from their efforts).
Indeed, by the end of this chapter, it becomes clear that both simple indeterminism and causal indeterminism founder because they try to establish a kind of control distinctive of free and responsible agents in the absence of correspondingly distinctive ontological resources. This gap is precisely what the traditional (and, nowadays, notorious) notion of 'personal' or 'agent' causation is intended to fill. Near the end of his instructive and thoughtful book, The Significance of Free Will, Robert Kane skeptically allows that "maybe theories of agent-causation can be resuscitated. But the burden of proof must be on anyone who would do so" (p. 195). I accept this assessment, and in the remainder of this book, I try to discharge that burden.
XIV PERSONS AND CAUSES
I begin, in chapter 3, by examining in some detail the three most influential accounts of free will that make recourse to the notion of agent causation: the eighteenth-century Scottish philosopher Thomas Reid and the contemporary philosophers Richard Taylor and Roderick Chisholm. In the discussion of each, I argue a central thesis (a point of agreement in the case of Reid, but one of disagreement with both Taylor and Chisholm). With Reid, I emphasize that an instance of agent causation ("exertion of active power," to use Reid's term) is not prior to or logically independent of the intention that is the agent's immediate effect. Against Taylor, I argue that an exertion of active power cannot itself be causally produced (although a wide range of factors are of course necessary for an agent to have such a causal capacity at the time of its exercise). Finally, contra Chisholm, I argue that recognizing that an agent's exertion of active power is intrinsically a direct exercise of control over one's own behavior suffices to dispel the worry that the agency theory must somehow tell a further story to explain how the agent controls this event.
These three contentions constitute the initial 'fixed points' from which my own articulation of the metaphysics of agent causation is developed in chapter 4. I there contend that we should think of causal notions in general in terms of the notion of 'causal power' or 'causal capacity.' Given this more general ontological framework, the difference between event (or broadly mechanistic) causation and agent (or personal) causation concerns the way in which causal capacities are exercised. With event causation, the capacity to generate a particular effect (in deterministic cases, that effect will be only one of a range of possible effects) is exercised 'as a matter of course': having in the right circumstances the cluster of properties that ground the capacity directly gives rise to one of the effects within its range. By contrast, having the properties that subserve an agent-causal capacity doesn't produce an effect; rather, it enables the agent to determine an effect (within a circumscribed range). Whether, when, and how such a capacity will be exercised is freely determined by the agent. After responding to some objections to the co- herence of this basic account, I critically examine Randolph Clarke's alternative account of agent causation, as well as some recent 'deflated' (ersatz) accounts.
In chapter 5, I defend an account of how agent-causally generated activity may be explained in terms of reasons. Central to the account is the assumption that what an agent directly causes is an action-triggering state of intention. The content of that intention (which endures throughout the action and guides its completion) is that an action of a specific sort be performed for certain reasons the agent had at the time. It is this twofold internal relation of direct reference and of similar content that grounds the explanatory link. I then respond to several objections stemming from Donald Davidson's influential critique of noncausal accounts of reasons explanations. Along the way, I defend the position that such explanations need not be contrastive in character—that is, whereas there may be a reasons-based explanation of the agent's acting as he did, this does not entail that there will also be an explanation of his so acting rather than performing any of the alternatives that had been contemplated. Near the end of the chapter, I suggest a refinement of the basic account in which the having of reasons generates or raises a carried tendency to act in particular ways, which tendency probabilistically structures the
Introduction xv
basic agent-causal capacity. It remains up to the agent, nonetheless, to determine which such tendency will be acted on. This refinement of the basic account allows both (1) a straightforward interpretation of the fact that we have relative tendencies to act, even when apparently acting freely, and (2) an account of strongly contrastive, as well as noncontrastive, explanations of actions.
Finally, in chapter 6, I discuss the assumption of many 'Naturalists' that this agent-causal account of free will is not consistent with 'the emerging scientific picture of the world.' I contend that there is little basis for this claim. We must sharply distinguish the plausible claim that macrophysical phenomena in general arise out of and are causally sustained by microphysical phenomena ('The Causal Unity of Nature Thesis') from the far from evident claim that all such higher level phenomena are constituted by more fundamental, lower level phenomena ('The Micro-Macro Constitution Thesis'). Granted the former thesis, why accept the latter? It does not follow from the former, as is shown by the possibility of some higher level features being emergent in a robust sense. And when the issues are properly sorted out, I argue, it is plain that the Constitution Thesis is not empirically established. To be sure, difficult issues concerning the underlying basis for "active power" would have to be sorted out before a decent conception of it as emergent is to be had. But as things now stand, such a construal isn't precluded by present knowledge. So although agent causation may be widely disdained by Naturalists, it is not at odds with naturalism.
This page intentionally left blank
Persons and Causes
1 Freedom and Determinism
1.1 An Ancient Argument
A moment before I began typing these words, I paused to consider how I should spend this afternoon. Shall I join my children on the floor, and await instructions about my role in their pretend play? Shall I return to my work instead? I am anxious to get started on this chapter, and I have planned an ambitious agenda of writing for the next few months. Or shall I go to the campus to retrieve my mail, thereby delaying a decision on how 1 shall spend the remainder of the day? I am dimly aware of some other possibilities, but these three are the options I'm taking seriously. After just a brief moment, I settle down to type at my computer. I would describe how making this decision seemed to me as follows: each of the options I considered (and perhaps some others) was open to me, such that I could have chosen it, just then. Put differently, it was entirely up to me to decide the matter, and I did so in a particular way, all the while being conscious that I might have chosen differently.
With characteristic eloquence, William James characterized causal determinism thus: "It professes that those parts of the universe already laid down absolutely appoint and decree what the other parts shall be. The future has no ambiguous possibilities hidden in its womb: the part we shall call the present is compatible with only one totality."1 Less lyrically, it is the thesis that there are comprehensive natural laws that entail that there is but one possible path for the world's evolution through time consistent with its total state (characterized by an appropriate set of variables) at any arbitrary time. The question I want to consider here is very familiar to philosophers: whether the truth of those beliefs about what was in my power are compatible with my decision's having been causally determined by circumstances that preceded it.
It might be thought that answering this question has been obviated by the strong confirmation of fundamental indeterminism by quantum theory, but this is
1. "The Dilemma of Determinism," in The Will to Believe (New York: Hafner,1987), p. 150.
3
4 PERSONS AND CAUSES
mistaken on at least two counts. First, there is an empirically adequate interpretation of the quantum formalism—that given by physicist David Bohm—which is deterministic. (This has had few adherents among physicists, although it is gaining in popularity among philosophical interpreters.) Second, and more important, quantum mechanics as standardly interpreted is compatible with such indeterministic effects largely 'canceling out' at the macroscopic level in ordinary circumstances involving the control of our own behavior. The result is a 'near-enough- as-makes-no-difference' determinism with respect to human action. Plausibly, if strict determinism is incompatible with human freedom, so is the looser variety, and this last is certainly an open possibility on current knowledge.
Returning to our central question, it is, of course, consistent with things having seemed to me to be as I described (consistent with my having those beliefs about what possibilities were open to me) that my choice had been causally determined. Yet it would be foolish to say that observation constitutes a sufficient basis for a compatibilist position with respect to freedom and determinism. To act freely is for the 'seeming' to be veridical—for it to be true that several options were open to me, while I deliberated. (Holding that the mere seeming is sufficient would be comparable to holding that there seeming to be a physical world with which I am interacting constitutes its being so.)
It is also true that when the nonphilosopher considers whether any of a variety of factors obtains that would suggest that a person was unable to have 'chosen' otherwise, she does not consider grounds for the thesis of universal causal determinism. Again, this hardly establishes the compatibilist's thesis. For, by the same token, our nonphilosopher does not consider the possibility that the person's decision was a direct result of remote-control manipulation by a clever Martian, unde- tectable to present human technology. You might object, "But she does not believe that such things happen." Yes, and that is precisely the point: she does not consider options that she is firmly convinced do not obtain and which she has no means to investigate. For all that, she implicitly believes that such manipulation would vitiate freedom of action. Likewise, it might be that the ordinary nonphilosopher disbe- lieves causal determinism, and she certainly has no means to evaluate its truth. I concede this is implausible, however, since the truly ordinary nonphilosopher has probably never entertained the thesis of causal determinism under any label. More realistically, then, what may be true, for all the absence of investigation into the truth of determinism in such contexts shows, is that the way people prereflectively think about 'the openness of the future to human decision' implies that if the future really is 'open,' causal determinism is false.
Does freedom of choice have this implication? It seems so to the typical under- graduate on first encountering the question. For, 'there ain't no changing the past,' and we, at least, have no choice in the matter of what laws of nature govern the way things operate. If these two factors together entail all of my future decisions, then those 'decisions' aren't really up to me. That they would occur was settled long ago. In Jonathan Edwards's pithy phrase, "Those things that are indissolubly connected with other things that are necessary, are themselves necessary."
However, at least until recently, a solid majority of English-speaking philosophers (or the most prominent among them) have held that there is no such impli-
Freedom and Determinism 5
cation. (For what it's worth, I have the impression that the climate of opinion on this topic has shifted somewhat. But I hope we can agree that this is worth very little, as counting noses is not, despite some recent tendencies, the method of true philosophy.) Many professed compatibilists have acknowledged that the informal reasoning above is seductive, as they might say, while insisting that careful exami- nation can reveal some hidden confusion or other.
Over the last several years, though, a number of philosophers have advanced formal versions of the incompatibility argument.2 In this chapter, I consider this sort of argument, with special attention to the central modal principle on which it turns. I contend that notwithstanding a few novel, as well as traditional, compatibilist responses to the argument, there is no good reason to doubt the prima facie truth of incompatibilism. The late Alan Donagan commented recently that philosophical discussion of this issue often resembles a "dialogue of the deaf." As he was surely right, explaining the purpose of the present exercise is in order. Al- though most of those professional philosophers who have thought about this issue and come to a firmly held compatibilism are unlikely to be moved by further variations on a familiar argument, my discussion is not directed primarily to them. I am much more hopeful of persuading those without firm convictions, who I have optimistically assumed will include some of my readers.3
I am also mindful that many compatibilists come to their position after considering what free choice would look like if indeterminism were true and deciding that it wouldn't help matters with respect to our freedom of choice. (According to some, it makes them far worse.) They basically reason that since (1) free choice must at least be possible, (2) determinism and indeterminism are exhaustive alternatives, and (3) indeterminism at best doesn't allow for anything not achievable under determinism, there must be something wrong with any incompatibilist argument. Different conclusions are then drawn about what that something is. The burden of most of the remainder of this book is to show that the third premise— that indeterminism cannot help matters—is mistaken. Thus, my subsequent defense of a particular indeterminist account of free will is, in a way, an extension of the argument of the present chapter.
1.2 Some Modal Principles and the Argument for Incompatibilism
Properly spelling out the notion of an agent's 'having it in her power to bring about some state of affairs' is not easy. We might simply state that an agent, S, can bring about a state of affairs provided that S can perform an action that, given the
2. See, for example, Carl Ginet's "The Conditional Analysis of Freedom," in P. van Inwagen, (ed), Time and Cause (Dordrecht: D. Reidel, 1980); and Peter van Inwagen, An Essay on Free Will (New York: Oxford University Press, 1983), pp. 93 ff.
3. Compare Peter van Inwagen's similar remarks in "Reply to Christopher Hill," Analysis 52 (April 1992), pp. 56-61.
circumstances and relevant causal laws, is logically sufficient for its obtaining.4 But this has the rather odd consequence that if Stewart Shapiro does not abandon his enthusiasm for second-order logic before midnight tonight, I can make it the case that he does not do so simply by, say, sitting in my chair. Something like the condition that S be able to perform an action that would play a central role in causing (or constituting) the state of affairs in question is needed. In contemporary discussions of the modal argument for incompatibilism, the notion of ability (and inability) is often left more or less at this vague level of analysis, perhaps with the dismissive comment that "no doubt more would need to be said in order to deal with philosophers' examples." (Much to my surprise, I recently found this compla- cent remark in an article written four years ago by Timothy O'Connor.5) For now, I will follow O'Connor and the rest in their lamentable laziness. I conjecture that it results from the following (implicit) thought: given the very general implication of the incompatibilist's thesis, one can focus one's attention on central cases. And since the focus of the discussion is a fragment of the logic of inability, we should be able to formulate some intuitive and sharp principles that govern this notion without worrying about the messiness around the edges of the notions of ability and inability. We consider the sustainability of this assumption in due course.
There are various ways to express the thought that more than one alternative is available to a person, such as "It is open to Susan to do x or y," "She is able to do x or y," and "It is up to her whether she does x or y." It will be helpful in what follows to adopt a bit of canonical language. If a true proposition p describes a circumstance an agent S could act at time t to prevent, we say that "p is avoidable for S at t." Naturally enough, if S lacks such control over the circumstance, we say that "p is unavoidable for S at f." What is thus 'unavoidable' for one agent at a particular time need not be unavoidable for another. Suppose that human beings have more than one course of action open to them and consider the proposition
(R) Tim pauses from his typing at 2:32 P.M. to review what he has written.
At the time I typed these words (2:31 P.M.), R presumably was not unavoidable for me, but given your location it was unavoidable for you.
Let "NS2t (p)" abbreviate "p is true and is unavoidable for person S at time t." Several philosophers have suggested that the following inference rule underlies the incompatibilist's reasoning:
From NS,t (p) and Ns,t (if p, then q), it follows that Ns2t (q).
4. Van Inwagen takes something like this route in his much-discussed book, An Essay on Free
Will. 5. "On the Transfer of Necessity," Nous 27 (1993), pp. 204-218.
(Recall Jonathan Edward's dictum: "Those things that are indissolubly connected with other things that are necessary, are themselves necessary.") Let us call this rule Transfer 1, in virtue of the fact that it licenses a certain kind of transmission of unavoidability from a pair of propositions to a third, relative to a particular agent and time.6 (In his much discussed treatment of the incompatibilist's argument, Peter van Inwagen uses a generalized version of Transfer 1—one that generalizes over all agents and all times—which he calls Beta.)
The plausibility of this sort of inference is best judged through a consideration of examples. Suppose a certain endearing philosophy professor, Stewart, has been tied to a chair at 10:00 P.M.. one Friday evening and is about to be force-fed two pounds of haggis by his abandoned students. Suppose it follows from certain facts about Stewart's biological constitution that if he eats all that haggis, he will become sick. It clearly is a consequence of these suppositions that Stewart will soon become sick and that this imminent sickness is unavoidable for him at the time of his unfortunate feeding. Of course, the mere fact that there are particular premises and conclusion conforming to Transfer 1 that are all true does not establish its validity. Instead, reflection on the example reveals a natural inclination to judge that it gives a correct way of reasoning to the conclusion. Because certain facts about Stewart's present and immediately future situation are unavoidable for him at 10:00 P.M. and (what is also unavoidable for him) if those facts obtain he will become sick, it follows that it is likewise unavoidable for Stewart that he is about to become sick.
Here is another example. At 10:00 P.M. the following Friday ( t1) , our hapless hero Stewart inadvertently presses the computer keys that entail the command "erase file 'Book,'" his only copy of his magnum opus in the philosophy of mathe- matics. It follows from facts about the computer's design and its local environment that the file will be erased two seconds later (t3). We may further suppose that there was nothing Stewart could do in the two-second interval to stop the completion of the process once started. Transfer 1 seems to give a truth-preserving means of reasoning from these facts to the conclusion that it is unavoidable for Stewart at the intervening moment (t2) that the file will be erased at t3: this seems to follow from the facts that Nstewart,t2 (the keys were pressed at tl) and Nstewart,t2 (if the keys are pressed at tl, the file will be destroyed at t3).
Transfer 1 is a plausible principle. But it must be judged as a first approxima- tion only to our target reasoning, for some clever examples of David Widerker's reveal a clear defect in its formulation. The following example was presented in an article of his in Analysis:7
(Case A) Suppose that by destroying a bit of radium at tl, Sam prevents its indeterministically emitting a subatomic particle at (2. Suppose further that this is the only way by which Sam can make sure that it won't emit radiation at t2.
6. John Fischer uses this term in The Metaphysics of Free Witt: An Essay on Control (Oxford: Basil Blackwell, 1994); I dubbed the principle "TNP" (for "Transfer of Necessity Principle") in "On the Transfer of Necessity," pp. 204-218).
7. "On an Argument for Incompatibilism," Analysis 47 (1987), pp. 37-41.
If we let p = The bit of radium does not emit a subatomic particle at t2, and q = Sam destroys the radium at tl, then Transfer 1 licenses us to conclude that Sam was unable at tl to refrain from destroying the radium, for both of the needed premises are satisfied. Clearly, Sam did not have control over the truth of p—he couldn't ensure that a particle was emitted at t2, even though this might have occurred had he not destroyed the radium. So NSam,tl (p). Consider now the second premise, NSam,tl (ifp, then q). This also holds because the conditional (if p, then q) is true and its truth was not within Sam's control. To have control over the truth of the conditional, Sam must have been able to make it the case that not (if p, then q). This is equivalent to (p and not q). If Sam had made true the second conjunct (not-q)— that is, had he refrained from destroying the radium—then he would have had no means of ensuring that the first conjunct (p) is also true (though, again, this might have been the case nonetheless). But surely it is consistent with these facts about the example to suppose, contrary to the conclusion licensed by Transfer 1, that Sam was able to falsify q, that is, to not destroy the radium at tl. Therefore, as stated, the inference rule is invalid.
As a first step toward seeing what has gone wrong here, we should think of Transfer 1 as a kind of "modal slingshot," as John Fischer has put it, transmitting the feature of unavoidability by a link that is itself necessary. Notice that in Wider- ker's example, the principle is used to sling the necessity back in time. Recall its formulation:
Transfer 1 From Ns,t (p) and Ns,t (if p, then q), it follows that Ns,t (q).
In Widerker's example, the proposition corresponding to q is made true at a time prior to the time at which the proposition corresponding to p is made true. In applying the principle to propositions that are temporally ordered in this way, we are allowing that future events (relative to a time t) may be relevant to what is avaoidable for an agent at t. But one might justifiably think that this is improper, quite apart from there being convincing counterexamples that demonstrate the invalidity of such an unrestricted principle. Intuitively, Transfer 1 was intended as a forward-directed8 'slingshot.' Incorporating the intended restriction gives us the following:
Transfer 2 From Ns,t (p) and Ns,t (if p, then q), it follows that Ns,t (q), for all propositions p, q such that q is made true no earlier than p is made true.
But this is not the true source of the problem. Consider the following example:9
(Case B) Suppose that Helen is deliberating about whether or not to insult Stewart. She decides not to do so at t2, and her decision is pre-
8. Including a 'side-long' shot as a limiting case, in which there is a necessary connection between two propositions that are made true simultaneously. (The most obvious case of this is a relation of logical implication.)
9. David Widerker presented a similar example to me in correspondence. In O'Connor (1993c), I argued that Widerker's example failed (see n.16, pp. 217—218, of that article). Further reflection has convinced me that my argument for this claim is not cogent.
ceded by some appropriate sign Z, occurring at tO, that makes it probable that she will not insult Stewart (perhaps a relaxation of certain facial muscles). Crispin detects Z and, understanding its significance, does not change his opinion concerning Helen's character. However, he might have done so had he not seen it.
Let p = Crispin does not change his opinion about Helen at tl, and q - Helen does not insult Stewart at t2.
Transfer 2 wrongly implies that it is not up to Helen at the earlier time, tO, to determine whether she insults Stewart at f2 [i.e., it wrongly implies NHelen.t0 (q)] It clearly is true that NHeien,io (p)> since Crispin doesn't change his opinion about Helen at tl (= p) and Helen doesn't have control over whether he does so. (She can only do something that might prompt such a change.) Is it also true that NHe!en,to (if p, then <j)? For Helen to be able to falsify the embedded conditional, she must be able to make it the case that (p and not q). But if Helen were to insult Stewart (i.e., to make it the case that not q), then in the absence of the appropriate sign at tO, Crispin might change his opinion about her (making it the case that not p). P might have obtained had Helen taken those other steps, but it's not true that it would have. Helen can, at tO, play a causal role in determining whether (p and not q) is true, but the matter is only partly in her hands. However, if this is right, Transfer 2 is apparently invalid.
Now Widerker has his own suggestion about how best to render the incompatibilist's reasoning. Earlier I used the metaphor of thinking of Transfer as a modal slingshot. Widerker's analysis may be summarized thus: the thing keeps breaking down, so clearly what we need is a more durable slingshot. Instead of using our unavoidability operator, N, as our slingshot, we should move to the logical necessity operator (' '). He suggests, then:
W From Ns,t (p) and (if p, then q), it follows that Ns,t (q).
If we can plausibly argue, by familiar reasoning, that the first premise of our schema holds where p stands for the conjunction of a description of a past state of the world and a statement of the basic laws of nature, then the thesis of determinism will yield the truth of the second premise required by W. (Such a conjunction, given determinism, will logically entail p, for all true propositions p describing any future state of the universe.) And W is clearly even more intuitive than either of the previous principles, stating only that unavoidability is closed under logical entailment.
I do accept the validity of W. However, one might well complain here that until we have a clearer and, ideally, uniform understanding of why our first two candidate principles failed, although intuitive in their own right, we should wonder whether our unavoidability operator is at all well behaved, as any Transferlike principle must assume. I now try to lay out such an understanding. Once it is in place, we see that Widerker's proposed substitute, although valid, is probably insufficient to the task, despite what I've just said in a deliberately hand-waving fashion about its apparent applicability. But we also see the way toward a natural and valid recasting of the incompatibilist's argument.
Let us recall Helen's happily refraining from behaving rudely toward Stewart. In the example, Helen can bring it about that q—that Helen does not insult Stew- art at t2—is false. (That is, she can insult him.) According to both Transfer 1 and 2, it must then either be in her power to make p false or be in her power to make (if p, then q) false. What is going on in this example is that because of an indeterministic causal link between the events described by p and q, if Helen were to exercise her power to make q false, it is determinately true only that one or the other of p and (if p, then q) would be false. It is indeterminate which of these would be false as a result of her undertaking not q (insulting Stewart). Thus, Helen has the ability to falsify the conjunction of p and (if p, then q) while lacking the ability to falsify either of the conjuncts taken separately.10
This suggests that the problem with our first two Transfer principles is that they falsely presuppose the validity of closure under conjunction introduction (sometimes called 'agglomerativity') for unavoidability. And indeed, we can demonstrate that this is so." Under a trivial assumption, the original Transfer entails the validity of, and its validity is entailed by, the conjunction of two more basic principles, closure under conjunction introduction and closure under logical en-
10. Another feature of Widerker's third example—not bearing on the discussion in the text—is that it involves an agent's ability at a time t to make true some proposition at a significantly later time. While it is certainly meaningful to talk of such abilities, the issue of which such abilities we have is considerably more muddied than that pertaining to abilities we have at t to bring something about 'straightaway,' in the immediate future of t. In the former case, this will partly depend on the world's cooperation in a whole host of ways in the intervening period (my not getting killed by a meteor and so on, of course, but also the actions of other people with whom I am in constant interaction). If the obtaining or nonobtaining of those necessary conditions on my effectively exercising my power is a deterministic consequence of the present state of the world, the issue of what I will be able to bring about at a later time is tolerably clear (in principle, even if I am not in a position to determine this with any confidence).
On the other hand, if such factors involve significantly indeterministic processes, there may not be a fact of the matter about whether they would obtain, were I to carry out a particular sequence of actions necessary to put me in a position to perform the final action required. Yet consider the following sort of case: a friend from out of town is visiting my area (unbeknownst to me) and will dine at the Snow Lion tomorrow evening. We may also suppose that I will deliberate tomorrow about whether to eat at the Snow Lion or Siam House and that I will choose Siam House. If neither of our respective decisions is now causally determined to occur, then it seems that I now have it in my power to meet and dine with my friend tomorrow night, although the fact that this is in my power depends on the obtaining of a causally undetermined fact about the future, viz., my friend's decision to eat at the Snow Lion restaurant.
In general, where the possibility of my making a proposition p true at t2 is dependent on certain actually obtaining factors X that are (a) beyond my control but (b) essentially causally "insulated" from the path I would have to take beginning at t1 in order to make p true at t2, we should hold those factors fixed in order to determine whether it is in my power at tl to make p true at (2. But we should not do this for cases such as Widerker's, where the obtaining of X would be causally influenced by my activity.
11. I owe the first half of this simple proof (applied to van Inwagen's principle B) to Thomas McKay and David Johnson's "A Reconsideration of an Argument Against Incompatibilism," in Philo- sophical Topics 24 (fall 1996), pp. 113-122. I commented on an early version of this paper at the 1995 meetings of the Central Division of the American Philosophical Association in Chicago.
tailment. The further assumption needed to see this is what van Inwagen calls Alpha:
Alpha From p it follows that Ns,t p, for all s and t.
Here are proofs of entailment relations between the principles (" " represents the material conditional, and "&" represents logical conjunction):
Assume Transfer 1. I Proof of closure under conjunction introduction: Ns,t p, NS,t q Ns,t (p &
q) 1. NS,t p premise 2. NS,t q premise 3. {p [q —> (p & q)]} necessity of a logical truth 4. Ns,t{p [q (p & q ) ] } 3, alpha 5. Ns,t [q ( p & q ) ] 1,4 Transfer 1 6. Ns,t (p &q) 2,5 Transfer 1
II Proof of closure under logical entailment: Ns,t (p), (p q) Ns,t (q) 1. Ns,t premise 2. (p q) premise 3. Ns,t (p —> q) 2, alpha 4. NS,t q 1, 3 Transfer 1
Assume the closure principles. III Proof of Transfer 1: Ns,t (p), Ns,t (p —> q) Ns,t (q)
1. Ns,t (p) premise 2. Ns,t (p —> q) premise 3. Ns,t [p & (p —> q)] 1, 2 closure under conjunction intro-
duction 4. {[p & (p —> q)] —> q} necessity of a logical truth 5. NS,t q 3, 4 closure under logical entailment
Once we see that our Transfer principle effectively comprises these two further principles, and also that closure under conjunction introduction is not valid for unavoidability, it is tempting to simply fall back on the other closure principle (Widerker's W). One might even argue that this is a reasonable, rather than ad hoc, recasting of ordinary reasoning. For just as I suggested in motivating the temporal restriction on Transfer 1, one might claim here that the Transfer principle shown to be invalid is unintendedly strong. Were one to be presented, apart from a counterexample of the sort given, with the distinction between being able to ensure only that at least one of p and (if p, then q) was false and being able to ensure that a particular one of them was false, one would unhesitatingly judge that only the former ability is required to block the inference to Ns,t (q).
But there is a lacuna in this strategy that Widerker has failed to notice: in order to use W (closure under logical entailment) in arguing for incompatibilism, we will have to establish at some point that the following holds: NS,t (p & q), where p is a proposition that captures the entire state of the world at some point in time,
and q is a conjunction of the laws of nature. Yet we cannot validly infer this from Ns,t p and NS,t q, as we've seen. Widerker might rightly point out that one can judge that Ns,t (p and q) holds in the context of a particular instantiation of this schema without a commitment to the general validity of closure under conjunction introduction. In particular, it would be very odd indeed to concede that I have no control over a true proposition concerning a past state of the world (that its truth is now unavoidable for me) and that I have no control over truths encapsulating the basic laws of nature, while supposing that nonetheless I do have control over the holding of the conjunction of these propositions. If one ever can reasonably judge that Ns,t (p and q) holds in the context of a particular instantiation of this schema without a general commitment to agglomerativity, this would seem to be such a case.12
Even if this is a reasonable thing to say, though, we should expect to be able to do better. Apart from the appeal to an ill-defined restricted applicability of agglomerativity, we are still puzzled by the nature of our ordinary concept of unavoidability such that agglomerativity breaks down in some cases. A better approach to the logic of the concept begins by considering the particular notion of ability on which it is parasitic and recognizing that the latter is a member of a family of closely related notions. (Here I follow in certain respects a recent discussion by Thomas McKay and David Johnson.)13
The statement 'I am able here and now to make it the case that p' admits of different interpretations, depending on the degree and kind of control that is in view. Obviously, all of my present abilities depend in part on the cooperation of my environment. To note only an obvious such dependency, if the oxygen is re- moved 'in a flash' from my surroundings, my abilities will likewise rapidly contract. The extent to which a given 'ability' to make some proposition true also depends on the friendly outcomes of nonstable features of my environment will differentiate corresponding degrees of control. The limiting case is where the instability is not merely epistemological (i.e., unpredictable by me, given my limited knowledge)
12. In earlier (unpublished) work, I made similar comments in defense of a different principle that I had formulated on recognizing the invalidity of agglomerativity for (our ordinary notion of) inability. That principle simply deleted the agglomerativity-entailing 'component' of Transfer, resulting in
Transfer 3 From Ns,t [p & (if p, then q)], it follows that Ns,t q.
In correspondence, Leonard Schulte pointed out to me a rather substantial dialectical disadvantage resulting from the use of such a principle in a defense of incompatibilism. Given the strict equivalence of [p & (if p, then q)] and [p & q], and given that 'q' in this context will stand for some arbitrary future action, it would be an unusually inept compatibilist that would grant such a premise as Ns,t [p & (if p, then q].
Does Widerker's use of W in this context face the same objection? No. Transfer 3 is effectively a combination of (the trivial) closure under logical equivalence and closure under conjunction elimina- tion, the use of which requires no interesting formal relationship of p and q. By contrast, W imposes a more stringent condition on the relation of p and q, and draws its inference from that relation. The compatibilist can cheerfully grant the requisite premises, and is apt to balk only at the significance of the relation (as captured by W).
13. See note 11 above.
but metaphysical, as would be true of indeterministic causal processes such as radium decay. Finally, there are also features internal to the agent that are relevant here. In some contexts, we imply that it is enough for an agent to be able to perform an action that is part of a causally sufficient condition for the truth of the proposition, whereas in others we require that he has knowledge or true beliefs about what is required on his part to make the proposition true and also that he has a reliable mastery of the behavior needed. (On the more minimal notion, which does not require these further conditions, a thief can turn the dial on the safe to the numbers that will result in unlocking the safe, even though he hasn't a clue to the correct combination. Similarly, the novice at the carnival game can successfully toss the ring onto the peg and thereby win the prize, even though she hasn't acquired the skill of knowing how one ought to do so to make success likely.)
Each of these additional conditions, when in play, infects the logic of the corresponding notion of inability, introducing an element of 'slippage.' In Wider - ker's examples, it's the fact that any action I might carry out, together with sur- rounding circumstances, would not ensure a unique individual outcome, although some such actions would ensure the falsity of the relevant conjunction. There are other examples, though, that are entirely independent of action-sensitive indeterministic processes: McKay and Johnson give a simple coin toss example. Suppose Stewart is trying to decide in which academic department he wants to make his permanent home. Torn by conflicting reasons and feeling the terrible weight of his dreadful freedom, he almost resorts to tossing a coin that is in his pocket. In the end he does not do so, though he well might have. The conjunction the coin does not land heads and the coin does not land tails, although true (because it is not tossed), is certainly not such as to be beyond Stewart's control. Any random toss will do the trick of making it false. However, as he cannot reliably falsify either of the conjuncts individually, the truth of each is, in at least one ordinary sense, beyond his control.
Thus, aspects of the ordinary notion or notions of ability, having to do with their incorporation of conditions such as justified belief, behavior mastery, and direct causal influence over outcomes—conditions that admit of degrees—render them insusceptible to a sharp logic of ability or even inability robust enough to support Transferlike inferences. Now that we can see what features of the concept pose the obstacle toward giving such a completely general logic, it is possible to recast the argument in terms of an artificially sharpened notion that does, intuitively, permit such inferences. Let us simply define a minimal notion of ability, such that one is able to make it the case that either p or not-p in this sense just in case it is open to one so to act (reliably or not) that it might be the case that p, and open to one so to act that it might be the case that not-p. The weakness of such a notion of ability mirrors a relative strength in the corresponding notion of inability, or unavoidability. We say that Ns,t p holds, for our strengthened operator, just in case p is true and the agent s cannot act (at or subsequent to t) in such a way that it so much as might be the case that not-p.
The prima facie plausibility of our original Transfer principle is due to the tendency, when running 'test' cases, to restrict ourselves to paradigm cases where
the looseness of the concept is not readily apparent. By setting such features aside, our tightened notion of inability, I submit, permits us to reason with Transfer in a generally valid fashion. And because this notion is stronger than ordinary notions of inability, any conclusion we are able to draw in terms of it will directly follow for the vaguer ordinary notions. (If I am unable to so act that it might even be the case that not-p, then I am also unable to knowingly and reliably ensure, relative to stable features of my environment, that not-p be the case.)14
Our earlier analysis shows that if we formulate the incompatibilist argument in terms of our artificially strengthened operator, we might just as well set aside Transfer (and Alpha) in favor of the principles of closure under logical entailment and closure under conjunction introduction. These are adequate to the task and together are slightly weaker assumptions, logically speaking, than the combination of Transfer and Alpha (i.e., the closure principles are entailed by but do not entail Transfer and Alpha). More important, using these more basic principles allows us to isolate the precise point of disagreement in the debate over the incompatibilist's argument.
The incompatibilist applies these forms of inference in the following way. Let P stand for a true proposition describing the state of the world at some time in the past—a mere five minutes ago, if you like. Let L stand for a conjunction of the fundamental laws of nature. And let A stand for a true proposition describing an action I am about to perform (say, typing the word "surely.")- Surely it is true at the present time (t) that NTim], (P). Whatever may have been true a short while ago, nothing I can do now can in any way influence the truth of P, as it concerns
14. Our strict notion of inability—of an agent's being unable to act in such a way that it so much as might be the case that p—is not vulnerable to an example of Kadri Vihvelin's that purports to show the invalidity of Transfer when the operator is interpreted in terms of a fairly strong notion of inability. Her example runs as follows:
The government runs a lottery with the following rules. The draw is by an indeterministic process from the list of social security numbers; every person with a number is automati- cally in the lottery. Although anyone's number may be picked, you win only if you've paid the lottery fee. If the number drawn belongs to someone who hasn't paid the fee, there is no winner and the government keeps the money. Betty did not pay the fee, her number wasn't drawn, and she didn't win. She could have paid the fee, and her number could have been drawn. (It was a fair lottery.) ("The Modal Argument for Incompatibilism," Philosoph- ical Studies 53 [1988], p. 239)
Now let tl be a certain time during the day before the drawing of the number, when it was still open to Betty to pay the fee. Then, Vihvelin argues, where p is 'Betty's number is not drawn at (2,' and q is 'Betty doesn't win the lottery,' using Transfer in accordance with a fairly strict notion of inability wrongly leads us to infer that NBetty,t1 (q)—that Betty cannot so act at tl that it might be the case that not-p. Why does Vihvelin suppose that this conclusion is false? Her claim turns on the consideration that "although Betty doesn't win the lottery, she can, in [a] weak, luck-dependent sense, win. Among the closest world's at which she pays the lottery fee is a world at which the indeterministic draw turns out differently and hers is the lucky number" (pp. 239—240).
Grant for the sake of argument Vihvelin's assumption that if Betty had paid the fee, her number might have been drawn (despite the fact that her paying the fee presumably would have been causally isolated from the draw), and so grant that Betty was able to act so that she might have won. But if one does grant this, then of course one must grant that Betty was able to act so that her number was drawn, thus falsifying the first premise of the argument needed for Vihvelin's counterexample.
the past, which is, as we say, "over and done with." Surely it is also true that NTim,t
(L), as the content of natural laws (whether be they deterministic or indeterministic) is not something I have any say about, now or at any other time. By closure under conjunction introduction, it follows from these two facts that NTim,t (P&L).
Now suppose that causal determinism is true. In that case, P&L directly entails that I will perform action A. But if I lack control (in even our artificially weak sense) over a true proposition concerning a past state of the world and the laws governing its evolution (alternatively put, if its truth is strictly unavoidable for me), then it is scarcely credible that I should nonetheless have some measure of control over the obtaining of any of its logical consequences. So, by an application of closure under logical entailment, we arrive at the conclusion that NTim,t (A)— that it is not open to me at t to act so that it might be the case that not-A And this, in turn, implies that at t, A is unavoidable for me in any of its ordinary senses.
This argument can be generalized clearly enough to show that causal determinism entails for any human person S, time t, and true proposition p, Ns,t (p). That is, if determinism is true, then no one ever has it in his power to make it the case (or even contribute to making it the case) that anything happens other than what in fact occurs.
1.3 The Fixity of the Past and of Natural Laws
I now consider the two 'standard' compatibilist replies to this argument. They both begin in the same fashion but diverge in filling in a certain detail. Let us note this fact by representing them as speaking with one voice at the outset:
(Compatibilists United) Our position will be seen to be less than incredible once we make a simple distinction. It would be incredible to assert that you have an ability to act so that your action would either constitute or causally result in a violation of natural law. Likewise, it would be incredible to assert that you have an ability to act so that your action would have made the past course of events to have been different from what they in fact were. We wouldn't dream of making either of these claims. The actual exercise of any power you have would always match your previously determined beliefs and desires in an appropriate (which is to say, lawlike) way. So, you often have the ability to act differently from the way you are causally determined to act, but were you so to act . . .
[At this point, our compatibilists separate into two camps.]
(Camp A—'backtrackers') . . . the past would have been different from what it actually was. Most relevant, something about the nature of your beliefs and desires—which ones you have or the relative importance to you of those you actually have—would have been different. But, of course, for those features of the situation to have been different, there would have had to have been a difference among the factors leading to your actual beliefs and desires. (We don't go in for miracles, even in counterfactual situations and no matter
how small.) Presumably, there would have been a difference in the entire history of the world. But there is nothing unreasonable in all this since, as we've already said, we are certainly not claiming that any of us has the ability to bring about a series of events in the past.
(Camp B—'scofflaws') . . . there would have been a tiny law-breaking event just before you did so that would then have caused your action in a perfectly lawlike fashion. (Yes, a miracle, but only one that is just big enough to have the action follow in the usual way.) This strikes us as less extravagant than supposing (with our confreres speaking above) that the entire past history of the world would have been different. In any case, there is nothing unreasonable in this since, as we've already said, we are certainly not claiming that any of us has the ability to do anything miraculous.
I regard the difference between the backtrackers and the scofflaws as an in- house squabble of little importance to the general issue at hand. They unite in acknowledging that asserting either of the following would be incredible:
1. I can perform an action, the performance of which would constitute or bring about a miracle.
2. I can perform an action, the performance of which would have brought about a series of events in the past.
But according to their preference, they suggest that asserting one or the other of these others is perfectly reasonable:
3. I can perform an action, the performance of which requires a prior miracle.
4. I can perform an action, the performance of which requires the entire past history of the world to have been different.
In making either of these claims, they are not denying a premise of the incompatibilist's argument. For whereas the backtracker compatibilists do claim that I can perform an action such that if I were to do so, the past (including the state of the world five minutes ago) would have been different, they do not claim that my performing this action would have brought about (or would have in any way causally contributed to) that previous state of affairs. The scofflaw compatibilist has a parallel position concerning the violation of an actual law of nature that would have preceded my performing any actually unrealized action that is within my power. Furthermore, neither camp maintains that I have the ability to falsify the conjunction of true propositions concerning the past and laws of nature. (In general, neither is committed to rejecting closure under conjunction introduction.) Instead, they are united in holding that my inability to bring about the falsity of any truths about the past or the laws of nature does not render me unable to bring about the falsity of certain truths concerning the future that those other truths logically entail. In so averring, they are rejecting the validity of closure under logical entailment (and for this reason, of Transfer), even in application to my artificially strict notion of unavoidability.
Statements (1) and (2) certainly are incredible. Yet—and here I hold out little hope of persuading any battle-hardened compatibilists among my readers—are (1) and (2) really any more incredible than (3) or (4)? Granted that they are distinct pairs of claims (as some are fond of reminding us), what is the relevance of this distinction to the issue of my ability to act differently from the way I will act? When I wonder what it is now in my power to do, I am wondering what is open to me, given the way things are and have been and the laws that constrain how things might be. And I am not, of course, merely wondering what general abilities I have. Rather, I want to know which of those abilities I am able to exercise in the present circumstances. Asserting (3) or (4) in response to this question is, on the face of it, no less absurd than either (1) or (2). An 'ability' to act here and now, the actual exercise of which strictly requires a prior condition that is lacking and which I cannot in any way contribute to bringing about, is, in the sense at issue, no ability at all. (This is essentially what closure under logical entailment for unavoidability implies.)
I have already acknowledged that such a response as I have given to the compatibilists' position will not move the firmly persuaded, though it seems to me no less correct and decisive for all that. And as best I can tell, it seems that way to the vast majority of people on first encountering the issue. (Consider the difficulty one has in trying to make compatibilism seem so much as plausible to undergradu- ates.15) My purpose here is mainly to explore carefully the reasoning that leads people to see an incompatibility between determinism and the freedom to act in any of a variety of ways, as well as to bring out the unintuitiveness of certain attempts to rebut such reasoning. It is worth remarking again that many compatibilists will acknowledge that, at first, it seemed that way to them, too. Those of my readers who reject the argument primarily because they fail to see how an indeterminist metaphysic could confer greater freedom of choice than a determinist one are invited to read on and consider the merits of the positive account developed in the rest of this book.
Before leaving this matter, however, I want to briefly call attention to a strategy used by Michael Slote for showing the invalidity of Transfer.16 (His argument would tell against the validity of Transfer for my artificially strong notion of unavoidability, if it has any force at all, and so we may construe it as so directed.) Slote's strategy is an indirect one: to show that the principle fails when it is applied to other forms of necessity in order to motivate its rejection when applied to unavoidability. Slote begins by assuming (what we earlier demonstrated in relation to unavoidability) that acceptance of a Transferlike principle for any form of necessity requires that this operator be closed for both conjunction introduction and logical entailment. Slote then tries to show that these closure properties are lacking for certain other kinds of necessity, thereby paving the way for arguing that certain compatibilist intuitions may be construed as the assertion that in our ordinary
15. At least when it is put forth in a fair-minded fashion, rather than as the opposition to 'hyste- ria' or 'mystery' mongers. Daniel Dennett's Elbow Room is an unfortunate example of the latter tactic.
16. "Selective Necessity and the Free-Will Problem," Journal of Philosophy 79 (January 1982), pp. 5-24.
understanding of the notion of "unavoidability," this concept, too, lacks these properties.
I will not discuss Slote's argument here, as I believe I have shown elsewhere that it fails.17 Although Slote fails to make good on his claim, his general strategy is perhaps the most promising one for a compatibilist-leaning philosopher to take. The inclination to see strict unavoidability as governed by Transfer and its consti- tutive principles of closure for conjunction introduction and logical entailment is natural and fairly robust. Accordingly, the most plausible way of undermining that inclination is perhaps not the traditional, direct-assault method of attributing it all to the confusions of metaphysicians (such as myself!). A better tack is to provide a kind of 'critical distance' by seeing the notion of unavoidability in relation to different relative-necessity notions, making the case that these operators fail to conform to the relevant inference forms and showing that the reason they fail to do so plausibly generalizes to the target case of unavoidability. I doubt that this can be done, but it seems to me the right sort of project for a serious-minded compatibilist.
1.4 Freedom and Responsibility
Philosophers who have maintained that determinism is incompatible with moral responsibility typically have done so on the basis of the following two premises:
1. A person is morally responsible for what he has done only if he could have done otherwise.
2. A person could have done other than what he in fact did only if determinism is false.
The second of these claims is a consequence of the argument defended earlier. Harry Frankfurt has dubbed the first claim The Principle of Alternate Possibilities (PAP). Although this principle is deeply intuitive and has been widely accepted, Frankfurt18 has brought to light a range of cases that in the judgment of many, provide grounds for rejecting it.
One of his well-known examples concerns a man named Black, who wants Jones to perform a certain action:
[Black] waits until Jones is about to make up his mind what to do, and he does nothing unless it is clear to him (Black is an excellent judge of such things) that Jones is going to decide to do something other than what he wants him to do. If it does become clear that Jones is going to decide to do something else, Black takes effective steps to ensure that Jones decides to do, and that he does do, what he wants him to do.
17. See my "On the Transfer of Necessity" (1993). Effective criticisms are also made by John Martin Fischer in his "Power Necessity," Philosophical Topics 14 (fall 1986), pp. 77-91, criticisms that he develops further in his The Metaphysics of Free Will (1994).
18. "Alternative Possibilities and Moral Responsibility," Journal of Philosophy 66 (December 1969), pp. 829-839.
As it turns out, Jones decides for his own reasons to perform the desired action, and Black does not intervene in any way with the deliberative process that leads to the decision or the carrying out of the action. Jones, we want to say, was responsible for his action and its immediate consequences. (We assume that the scenario is "normal" in other respects, so that there are not other considerations affecting our evaluation of Jones's responsibility.) And yet it seems that he could not have done otherwise. (The power and intentions of Black ensure that this is so.) So, Frankfurt concludes, PAP is seen to be false.
Peter van Inwagen19 has argued that the incompatibilist can concede Frank- furt's verdict on PAP by endorsing the following variant of that principle:20
(PPP2) A person is morally responsible for a certain state of affairs only if (that state of affairs obtains and) he could have prevented it from obtaining.
The basic strategy of van Inwagen's argument exploits the fact that abstract states of affairs can be more or less fine-grained. For example, let us suppose that the desired action in Frankfurt's scenario is that Jones shoots Stewart. Correspond- ing to the particular event of Stewart's death are these states of affairs: Stewart's dying at t, Stewart's being killed by someone at t, Stewart's being intentionally shot by someone at t. Now, from the facts that an agent is responsible for a state of affairs S and that S entails S*, it does not follow that the agent is responsible for S*. (Contrast this with lack of responsibility for P, analogous to unavoidability.) Stewart's being killed by someone at t, for example, entails The universe's existing at t. Before one considers Frankfurt-type scenarios, it is quite natural to say that the point of "cutoff in terms of responsibility in a sequence of increasingly less specific states of affairs (where each entails the one subsequent to it) is precisely the point at which a state of affairs is such that the agent could not have prevented it. If we can show that we needn't absolve the agent of moral responsibility in Frankfurt cases to preserve this intuition, we would seem to have sufficient reason to preserve it.
In the preceding scenario Jones can't prevent any of the states of affairs. So, according to PPP2, he is not responsible for the fact that they obtain, as they are inevitable from the standpoint of his 'sphere of influence.' But there is at least one other, closely related state of affairs for which we may plausibly hold him responsible without abandoning PPP2: Stewart's being killed by Jones acting on his own.21
This indicates a general formula applicable to any Frankfurt-type situation for characterizing a state of affairs for which the agent may be held responsible. For in all such cases, the agent is in no way caused to act or decide as he does, but rather acts or decides "on his own" or freely. In 'ordinary' situations, there will be a variety of other, more broadly delineated states of affairs for which the agent is equally responsible.
19. "Ability and Responsibility," Philosophical Review 87 (April 1978), pp. 201-224; and An Essay on Free Will (Oxford: Oxford University Press, 1983), chap. 5. Unless indicated otherwise, page references in the text will be to the 1983 work.
20. He also defends two other, related principles that I won't bother to discuss here, "ppp" is an acronym for the "Principle of Possible Prevention."
21. That is, as a result of a free decision by Jones.
It bears emphasis that by ascribing responsibility to Jones for Stewart's being killed by Jones acting on his own but not for Stewart's being killed by someone at t, we needn't in any way diminish the extent to which his conduct is reprehensible and blameworthy. We are simply recognizing that care needs to be exercised (especially in highly contrived scenarios such as we have been considering) in determining precisely which of a number of closely related states of affairs the agent actually brought about by his action—and for which he is accordingly responsible—relying on the intuitive notion that an agent cannot be responsible for a state of affairs that he couldn't have kept from obtaining.
Van Inwagen's response to Frankfurt has attracted several critical replies.22
In particular, John Fischer has recently argued that such a "flicker of freedom" strategy—that of finding a sufficiently fine-grained state of affairs that the agent could have prevented in any Frankfurt-type scenario—is unsatisfactory because, he urges, the availability of such narrow alternatives does not ground the agent's responsibility nor does its presence guide our judgments of responsibility.23 Fischer suggests that this becomes plausible once one recognizes that the 'alternative' left available to the agent in such scenarios is always one in which the agent acts unfreely.
I am unconvinced by Fischer's argument. After elaborating an account of freedom of will later in this book, I make some points, stemming from that account, which call his argument into question. But my principal reply, which I will now develop, is that one's view on this matter doesn't (or shouldn't) have deep significance for the broader dispute over the compatibility of causal determinism and moral responsibility. Philosophers who hold that determinism is incompatible with moral responsibility typically draw on some alternative-possibilities condition on responsibility. Its specific content varies, but we may generalize over them with the following:
Alternative Possibilities [AP] Condition A person is morally responsible for his action or its consequences only if there were alternative possibilities open to him.
What is largely common ground in contemporary discussions is that if examples of the sort to which Frankfurt appeals show that even this general AP condition must be rejected, then the joint truth of determinism and its incompatibility with alternative possibilities is irrelevant to whether we are responsible for our actions. One who accepts both these claims may, without further argument, em-
22. Robert Heinaman, "Incompatibilism without the Principle of Alternative Possibilities," Aus- tralasian Journal of Philosophy 64 (September 1986), pp. 266-276; William Rowe, "Causing and Being Responsible for What Is Inevitable," American Philosophical Quarterly 26 (April 1989), pp. 153-159; John M. Fischer and Mark Ravizza, "Responsibility for Consequences," in J. Coleman and A. Buchanan, eds., Festschrift for Joel Feinberg (Cambridge: Cambridge University Press, 1992), and "Responsibility and Inevitability," Ethics 101 (January 1991), pp. 258-278.
I defended (PPP2) against these criticisms in my "Alternative Possibilities and Responsibility," Southern Journal of Philosophy 31 (1993), pp. 345–372.
23. Metaphysics of Free Will, op. cit., pp. 140-147; see also his introduction (with Mark Ravizza) to Perspectives on Moral Responsibility (Ithaca: Cornell University Press, 1993), p. 38.
brace "semicompatibilism"—accept that causal determinism is incompatible with an agent's having alternative possibilities but deny that causal determinism is incompatible with moral responsibility. (This is just the position advocated by Fischer.)
I suggest, to the contrary, that even if one concludes from Frankfurt-type cases that alternative possibilities of any sort are not necessary to moral responsibility, one may not plausibly draw the further conclusion of semicompatibilism. What- ever the proper verdict on Frankfurt's examples may be, the compatibility of determinism and moral responsibility must be settled on independent grounds. Indeed, if the incompatibilist argument set out earlier in this chapter is sound, then determinism is incompatible with responsibility, too. On this issue, Frankfurtian cases are inconclusive.
To see this, we should firmly remember that the AP condition on moral responsibility for one's action and its consequences is deeply entrenched in ordinary, pretheoretical thinking about responsibility. Suppose for the sake of argument that Frankfurt and others have established possible cases that show that this pretheoretical commitment is false. Then the natural conclusion to draw is that ordinary thought has misidentified the freedom-relevant necessary condition on moral responsibility by conflating the AP condition with some distinct condition it closely tracks. For we shouldn't overlook the obvious, that is, that Frankfurt cases are extremely contrived and (unless we are badly mistaken about the world) never instanced. Ordinary thinking about responsibility proceeds by reflecting on familiar cases. And the common conclusion is that for an agent to bear responsibility in such familiar cases, the condition of one or more significant alternatives must obtain.
So, even if we gave up the strict or conceptual necessity of the AP condition on moral responsibility, the fact that we rely on its presence or absence in actual cases strongly suggests that it must be tightly connected to what is a truly necessary condition. That is, the two conditions are coextensive in ordinary contexts, even if they can in principle come apart. As philosophers, we would want to characterize the truly necessary condition. But a constraint on any proposal is that it entails the presence of alternative possibilities, relative to (conditional on) a broad assumption about actual deliberative environments, that is, that it lacks a purely 'counterfactual intervener'—one who does nothing that influences the actual flow of events but merely would do so if circumstances were different in some respect.
Consider, by way of contrast, the strongly revisionary conclusion that some philosophers draw from Gettier counterexamples to the justified true belief (JTB) analysis of knowledge. Whereas some conclude merely that the examples show that JTB is insufficient for knowledge, others argue that reflection on the examples shows that being epistemically justified is wholly irrelevant to knowledge. Whether this stronger thesis is correct or not, it is an option that one may reasonably consider when reflecting further on a range of examples, for the JTB analysis is a theoretical analysis that philosophers have devised in applying the ordinary, somewhat inchoate notion to various cases. But it is implausible to make a similar move in response to Frankfurt examples, for the AP condition on responsibility is present in ordinary thought (and is a deep conviction at that). It isn't a claim that has to be teased out of our thinking.-
Semicompatibilism thus requires a defense beyond refuting the AP condition. What is needed is a further argument from the Frankfurt examples to the irrele- vance, even in ordinary circumstances, of alternative possibilities. As such, it has to be an argument that ordinary moral reflection is deeply confused in some important respect. I note two ways this argument might go.
First, one might argue that Frankfurt's examples bring out a conflict in our moral intuitions concerning the nature of responsibility. Conflict of this sort would call for revisionary analysis that develops the deeper strand in our moral thinking. It is not at all apparent that conflicting intuition is what underlies the surprising conclusion drawn by Frankfurt and his allies. It's not as if Frankfurt pointed out a perfectly ordinary case in which we don't apply the AP condition, causing us to see that we were implicitly thinking about everyday cases in a manner different from our ostensibly general principle. Nonetheless, it is open to the Frankfurtian to try to make plausible this sort of analysis.
A second strategy is to argue that Frankfurt-type cases generalize widely beyond the artificial scenarios, with their hidden counterfactual interveners. In particular, it may be argued that if determinism is true, every 'ordinary' choice circumstance is relevantly like the artificial Frankfurt-type cases. They are like those cases in that they are inconsistent with alternative possibilities (AP) of any sort and yet this fact is irrelevant to the question of whether agents ever bear responsibility for the consequences of their actions. In this sort of approach, we simply fail to grasp the significance of our notion of AP. This, too, is not a highly promising strategy on the face of it. It would entail that we have a clear no

persons and causes: the metaphysics of free will

Documents