Persons and Causes: The Metaphysics of Free Will
-
Upload
others
-
View
0
-
Download
0
Embed Size (px)
Citation preview
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 fol- lowing: 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 chil- dren, 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 pic- ture 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 philoso- phers of the past
managed to bequeath to us a perspicuous and immediately recog-
nizable articulation of that thought? Granted, whether any such
philosophical ac- count 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 de- gree—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 pre- cisely 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 under- stand 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—sur- prise!—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 un- avoidable 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 anteced- ent 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 rea- sons 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 indeter- minism 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 phi- losophers 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
disagree- ment 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 exer- cised. With event causation, the
capacity to generate a particular effect (in deter- ministic 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 contrast- ive, 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
lat- ter? 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 empiri- cally
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 famil- iar 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 circum- stances 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 interpreta- tion 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
indetermin- istic effects largely 'canceling out' at the
macroscopic level in ordinary circum- stances 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 hav- ing 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 deter- mined. 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 deter- minism. Again,
this hardly establishes the compatibilist's thesis. For, by the
same token, our nonphilosopher does not consider the possibility
that the person's deci- sion 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 philoso- phers (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, compati-
bilist responses to the argument, there is no good reason to doubt
the prima facie truth of incompatibilism. The late Alan Donagan
commented recently that philo- sophical 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 con- sidering 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 alter- natives, and (3) indeterminism at best doesn't
allow for anything not achievable under determinism, there must be
something wrong with any incompatibilist argu- ment. 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 de- fense 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.
6 PERSONS AND CAUSES
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.
Freedom and Determinism 7
(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 argu- ment, Peter van Inwagen uses a
generalized version of Transfer 1—one that gener- alizes 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 comple- tion 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.
8 PERSONS AND CAUSES
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 exam- ple: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.
Freedom and Determinism 9
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 char-
acter. 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
incompat- ibilist'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 neces- sity 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 determin- ism 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.
10 PERSONS AND CAUSES
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 indeter- ministic 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 dem- onstrate 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.
Freedom and Determinism 11
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,
12 PERSONS AND CAUSES
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
intro- duction. 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 un- avoidability such that
agglomerativity breaks down in some cases. A better ap- proach 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 discus- sion 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, result- ing
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.
Freedom and Determinism 13
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 indeter- ministic 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, intu- itively, 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
14 PERSONS AND CAUSES
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.
Freedom and Determinism 15
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 indeterminis- tic) 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
deter- minism 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
16 PERSONS AND CAUSES
how small.) Presumably, there would have been a difference in the
entire his- tory 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 abil- ity 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
unreason- able 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 incom- patibilist'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 logi- cal
entailment (and for this reason, of Transfer), even in application
to my artifi- cially strict notion of unavoidability.
Freedom and Determinism 17
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 un-
avoidability implies.)
I have already acknowledged that such a response as I have given to
the com- patibilists' 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 compati- bilists 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 indeter- minist
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 un-
avoidability, 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 neces- sity 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.
18 PERSONS AND CAUSES
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
deter- minism 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.
Freedom and Determinism 19
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 pre- serve 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
responsi- ble 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
refer- ences 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.
20 PERSONS AND CAUSES
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 (espe- cially in highly contrived scenarios such as we
have been considering) in determin- ing 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 free- dom 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 signifi-
cance 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 exam- ples of the sort to which Frankfurt appeals show that even
this general AP condi- tion 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.
Freedom and Determinism 21
brace "semicompatibilism"—accept that causal determinism is
incompatible with an agent's having alternative possibilities but
deny that causal determinism is incom- patible 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 deter-
minism and moral responsibility must be settled on independent
grounds. Indeed, if the incompatibilist argument set out earlier in
this chapter is sound, then deter- minism 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 re- sponsibility 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
pretheoret- ical commitment is false. Then the natural conclusion
to draw is that ordinary thought has misidentified the
freedom-relevant necessary condition on moral re- sponsibility 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 famil- iar 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 as- sumption
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, some- what
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.-
22 PERSONS AND CAUSES
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 im- portant
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 be- yond the artificial scenarios, with their hidden
counterfactual interveners. In par- ticular, it may be argued that
if determinism is true, every 'ordinary' choice cir- cumstance 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 responsi- bility 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