Toward a plausible event-causal indeterminist account of free will

Abstract

For those who maintain that free will is incompatible with causal determinism, a persistent problem is to give a coherent characterization of action that is neither determined by prior events nor random, arbitrary, lucky or in some way insufficiently under the control of the agent to count as free action. One approach—that of Roderick Chisholm and others—is to say that a third alternative is for an action to be caused by an agent in a way that is not reducible to event causal terms. A different approach than the Chisholmian appeal to primitive substance causation is one that, instead, involves causal relations purely among events. This paper presents a particular event-causal indeterminist account of free action, describing both its attractions and recent objections to it, and then proposes a revised version, with the aim of supporting the plausibility of an event-causal indeterminist approach to free will.

Take the thesis of determinism to be this: There is at every moment exactly one physically possible future, where a physically possible future is one that is consistent with the actual past and the laws of nature.¹ Then the thesis of indeterminism, defined simply as a negation of the thesis of determinism, is as follows: It is not the case that there is at every moment exactly one physically possible future.² On the thesis of indeterminism, sometimes there is more than one way things can go on to be, given the way things have been and the way the natural laws are. In other words, if the thesis of indeterminism is true, then not all of the events of the future are fixed by the past and deterministic natural laws.

No one reasonably thinks that demonstration of the truth of the thesis of indeterminism so defined in itself would imply that anyone has free will, since the thesis says nothing about when or where there is sometimes more than one physically possible future, and it says nothing about human decision making or about overt human action. The thesis of indeterminism could be true while at the same time it is true that all human decision making proceeds deterministically and while at the same time it is true that all human actions are governed by deterministic natural laws and are the inevitable outcomes of what has come before, going back into the distant past. Indeed whatever indeterminism there might be in the world could be far removed from us or at a sub-atomic level having no effects on human action.

Still, some people think that the truth of the thesis of indeterminism might make space in the world for free will—the power to perform free actions—provided that the indeterminism were of the right sort and were located in the right places. It is these notions of “right sort” and “right places” that need specification. To cut to the chase, I think that the “right sort” is indeterministic causation among certain agent-involving events and that the “right places” are between the occurrence of reasons of the agent’s own and the agent’s subsequent decisions and other actions. If we are to give any sort of positive characterization of indeterminism that might be helpful to human agency, we should point our attention to the notion of indeterministic causation among events. It might not be clear what that is, and it might not be clear whether or not we should believe in such a thing. And once we clarified the notion of indeterministic causation, it would remain to be seen how such causation might play a role in human agency and what the implications of such involvement might be.

In what follows, I address these issues. First, I discuss indeterministic causation itself. Next, I present an account of free action that I have developed in previous work (Ekstrom 2000, 2003, 2011), which incorporates such causation, explaining the motivations for the view. In light of problems for that account, including objections raised by, for instance, Clarke (2003) and Pereboom (2014b), I give a revised version of an event-causal indeterminist account of free will and, finally, describe its virtues.

1 Indeterministic causation

Indeterministic causation is evidently widely accepted among theorists who work on the topic of causation and in the philosophy of science. But confusions about its nature and lurking doubts some may have about its existence seem to fuel worries about, and objections to, event-causal indeterminist accounts of free will. Spending a bit of time looking directly at the topic, then, proves useful for moving forward in the free will debate.

The issue about indeterministic causation that bears most directly on event-causal indeterministic accounts of free will concerns the nature of causal connection. The vast literature on causation contains numerous competing accounts of causal connection. These assorted accounts may be viewed as variations on two basic themes: one, an understanding of causal connection in terms of probability raising, and two, an understanding of causal connection in terms of processes of physical producing.³ On the first theme, to cause is to make more probable. Roughly, for one event to cause another is for the one to make the other more likely to occur. On the second broad approach—where accounts in terms of energy flow, physical processes, and property transference may be grouped—to cause is to produce. Causal connection is a matter of processes of production rather than of probabilities. Some find the problems for these views and the responses to them to motivate hybrid views (for instance, perhaps causes raise the probabilities of processes that produce effects).⁴

Let’s suppose that causal relata are events. This is somewhat controversial; some take them to be facts (or something else). But suppose, along with many theorists of causation, that what stand in causal relations to one another are events.

Approaches to causation in terms of probability are a natural fit with an event-causal indeterministic account of free agency, since on those approaches, causes are probability raisers of their effects, and the probabilities may be less than 1.0, giving sense to non-deterministic causal relations that could hold between agent-involving events during decision making and acting. The basic form of a probabilistic theory of causation is as follows:

$$\begin{aligned} C\hbox { is a cause of }E\hbox { just in case P}(E\,|\,C)> \hbox {P}(E\,|\,{\sim }C). \end{aligned}$$

That is, an event C is a cause of event E just in case the probability that E occurs given than C occurs is higher than the probability that E occurs given that C does not occur. The idea is that C changes the probability of E: its occurrence makes E more probable than it would have been if C had not occurred. This basic characterization faces problems, including spurious correlations and handling the asymmetry of causal connections. Hans Reichenbach proposed to deal with such problems with the addition of a “screening off” condition (1956). Cartwright (1979) and Skyrms (1980) aim to rectify problems with his account by adding the requirement that causes raise the probability of their effects in various background contexts. Cartwright proposes that C causes E if and only if $\hbox {P}(E\,|\,C\, \& \,B)> \hbox {P}(E\,|\,{\sim }C\, \& \,B)$ for every background context B.⁵ Skyrms’s proposal involves a weaker condition: a cause must raise the probability of its effect in at least one background context, and lower it in no contexts. Specifying a background context requires making precise which factors will be conditioned upon or held fixed.

Other approaches to causation than those put centrally in terms of probability may be suitable for an event-causal indeterminist free will theory, including a view on which causation is a primitive bringing about, which cannot be further analyzed, so long as the bringing about of one event by another is not assumed to require necessitation.

The crucial matter concerning indeterministic causation, whichever theory best captures its nature, is that, if there is such a thing, then some events are causally related to their effects without necessitating them. Events that indeterministically cause other events make a difference for those effects, but the effects might not have occurred, in the same circumstances and holding fixed the natural laws. It can happen that the events in question occur but are causally inert: they occur, but the effects they might have brought about, they do not (this time) bring about. An indeterministic or nonnecessitating cause is one that can fail to produce its effect, even without the intervention of anything to frustrate it.

Not everyone finds it natural to accept the idea that events can cause without determining their effects, but there are cases that certainly look to be instances of causation but in which there is no determination. Elizabeth Anscombe’s example of an indeterministic or nonnecessitating cause is a collection of radioactive material that causes a Geiger counter to be activated, which—since the Gieger counter is connected to a bomb—causes the bomb to explode (1981). It was not determined that the radioactive material emitted particles in such a way as to activate the Geiger counter sufficiently to set off the bomb. Given prior events and natural laws, the material might not have emitted particles in the way it did; nonetheless prior factors brought about the particle emission, in turn causing the explosion.

Most causation theorists now take it to be a desideratum of a theory that it make room for causal connections without a presumption of determinism. David Lewis, for instance, writes:

I certainly do not think that causation requires determinism...plenty of people think that our world is chancy; and chancy enough so that most things that happen had some chance, immediately beforehand, of not happening. These people are seldom observed to deny commonplace causal statements...We had better provide for causation under indeterminism, causation of events for which prior conditions are not lawfully sufficient. (1986, p. 175)

Lewis’s counterfactual theory allows for causation in indeterministic contexts, understanding causation in terms of causal dependence among a sequence of events, where an event Ecausally depends upon the distinct event C just in case both occur and the probability that E would occur, at the time of C’s occurrence, was much higher than it would have been at the corresponding time if C had not occurred. This counterfactual is true, for Lewis, just in case, in the nearest possible worlds where C does not occur, the probability of E is much lower than it was in the actual world.⁶

Now the central point I want to make in the present section is this: Certain criticisms of event-causal libertarian theories, made most often by proponents of agent-causal libertarian theories, are off base, and they are off base because of inattention to a positive characterization of indeterminism—which I have suggested points toward the notion of indeterministic causal connections between events. For instance, Timothy O’Connor writes the following:

However, according to many critics (myself among them), indeterminist event-causal approaches falter just here, in the fact that the free control they posit is secured by an absence, a removal of a condition (causal determinism) suggested by the manifestly inadequate varieties of compatibilism. (2009, p. 192)

O’Connor calls varieties of compatibilism “manifestly inadequate” and contends that event-causal incompatibilist approaches to characterizing free action “falter” because they claim to secure freedom “by an absence, a removal of a condition (causal determinism).” But it is incorrect to say that event-causal libertarian theories appeal only to an absence—the absence of causal determination. It’s not a removal or an absence that is posited at the crucial juncture in the production of action on the event-causal indeterminist account of freedom. It is certainly not an absence of causation. It is, instead, causation of decisions and other actions by the occurrence of reasons of the agent’s own—in particular, indeterministic causation rather than deterministic or necessitating causation. Decisions and other actions are not absences. Indeterministic causal connections between an agent’s reasons and her actions are not absences. Furthermore, indeterministic causal connection between events, as we have seen, is something of which we can make good sense.

2 Self-direction and alternatives

I turn to addressing directly the nature of free will—the power to act freely—and, in particular, to explaining the motivations for the event-causal indeterminist account I have defended in previous works (Ekstrom 2000, 2003, 2011).

The various theories in the historical and contemporary literature on free will are plausibly construed as aiming to capture two central features, while varying in emphases and on the right interpretations of these: self-direction and alternative possibilities.⁷ With respect to the first, the thought is that free acts are directed by the agent herself—they derive from her self or from who she is. A natural starting place in developing the idea of self-direction is to say that free acts are the ones an agent does because she wants to and not because she is compelled, coerced or forced to do them. One problem for this initial characterization is that some animals act on desire without having what we would call free will. Another problem is that some acts by persons on desire do not reflect the self; they are ones from which the agent herself is alienated. Harry Frankfurt addresses this problem in a series of works by giving greater specificity to the notion of the self, observing that persons have the ability to rise above their desires to form desires concerning them, and proposing that a person’s self is delimited by her desires concerning which desires lead her to act, when or if she acts (that is, in Frankfurt’s terminology, her self is to be identified with her “second-order volitions,” which are second-order in virtue of having as their intentional objects desires to act, rather than having as their intentional objects actions themselves).⁸ Gary Watson objected to the idea that merely ascending orders of desire (from desires to act at the first order, to desires to desires to act at the second) somehow brings us closer to the real self, suggesting instead that we act freely in acting from our values, which are the judgments concerning the good that we can defend in a cool and non-self-deceptive moment.⁹ Watson later described this proposal as too rationalistic, but still maintained that in characterizing free action it is crucial to capture the idea of derivation of action from the self.¹⁰

My own account of the self in terms of preferences develops from these ideas. The term “preferences” I use not in the ordinary sense, but instead as a term of art, to refer to desires one forms or maintains as one aims to desire what is good.¹¹ This notion enables the account to avoid problems with the Frankfurtian model on which internal (second-level) desires can be formed and maintained on any basis whatsoever and even on no basis. “Acceptances” (elsewhere I call them “convictions”)¹² I understand as beliefs formed with the aim of assenting to what is true. These attitudes—preferences and acceptances—seem especially definitive of the self, or of one’s character or psychological identity. We might understand one’s central self or “core self” as comprised of those preferences and acceptances that cohere together in a network of mutually supporting attitudes. Action deriving from a cohering preference seems particularly self-governed or autonomous. (I defend a coherence account of autonomous action in Ekstrom 1993, 2005a, b.)

Preferences, as so defined, I envision as being decisively adopted or maintained and as involving the agent’s activity, rather than states of mind with which she passively finds herself. We do not always decide on our attitudes, of course. Our beliefs and desires are often not voluntarily adopted—instead we simply find ourselves wanting what we want and believing what we believe. But we can decide what attitudes to adopt: I can decide whether or not I prefer to vote for the Democratic candidate for President, for instance, and I can decide whether or not I assent to the proposition that God exists. On such occasions we decide what sort of person we want to be, settling on our character-defining attitudes. In the process we ask ourselves such questions as, “What should I prefer?” and “What should I accept to be true?” In Ekstrom (2000), I suggested: “Take an agent to consist of evaluated reasons (preferences and acceptances, as a group constituting the character), together with an evaluating and choosing faculty (in other words, a power to fashion and refashion character)” (p. 114). We might call the agent’s evaluating and choosing faculty or faculties her intellect and will, so that an agent—as a matter of her moral or psychological identity, her self—is understood as her intellect, will, and character-defining attitudes.¹³ A conception of the self can be used to explicate self-directed action: self-directed action is caused by reasons that are elements of the self, without the interference of (non-self-arranged) manipulation, coercion or constraint.¹⁴

But plausible accounts of free will involve self-direction and alternative possibilities, both. Philosophers from a range of historical periods highlight the alternative possibilities condition of freedom. For Aristotle, the voluntary involves a dual ability: “where it is in our power to act it is also in our power not to act” (Nicomachean Ethics 1113b6). For Hume, liberty is “a power of acting or not acting according to the determinations of the will; that is, if we choose to remain at rest, we may; if we choose to move, we also may” (An enquiry concerning human understanding, Sect. 8). For Thomas Reid, the “power to produce any effect implies the power not to produce it” (1983, p. 523). And for Kant, for an act to be truly free, “the act as well as its opposite must be within the power of the subject at the moment of its taking place” (1960, p. 45). Bernard Williams describes the claim that free will exists as consisting in “something to the effect that agents sometimes act voluntarily, and that when they do so they have a real choice between more than one course of action; or more than one course is open to them; or it is up to them which of several actions they perform” (1995, p. 5).¹⁵

Where should we locate the alternative possibilities in a model of free action? In reflecting on this question I came to think—along with many others who have thought hard about free will—that what was important is that we could have had different selves or characters than we did. What counted was that it was not rigged from the start who we turned out to be. Our actions should flow from our characters in order for those actions to be self-determined, and we should have a chance to be other than the sorts of people we are. The past and laws had to leave some openness in this matter. Indeterministic formation of preferences in the specialized sense allows for a self to have been other than it is in fact. One could have been a different sort of person if one’s preferences are not the inevitable outcome of natural laws and prior events.

And so I ended up with a theory of free action on which one acts freely in acting from a preference, and not just any preference but one which could have been otherwise—more specifically, one which was non-coercively formed or maintained and which had an indeterministic causal history, with its causes being various considerations that occur to a person as she is contemplating the question of what she prefers to do (Ekstrom 2000, 2003).

Free Action as Action on Undefeated Authorized Preference: An agent’s act is free just in case the act results by a normal (non-deviant) causal process from the agent’s undefeated authorized preference for the act. A preference has undefeated authorization just in case the agent’s evaluative faculty was neither coerced nor causally determined by anything to form that preference, but rather the preference was indeterminately caused by the considerations that were brought to bear in the agent’s deliberation. (2000, pp. 108–109)

On this model, considerations of the agent’s are themselves indeterministic causes of the agent’s formation of a preference concerning what to do. The indeterminism, then, is between events that are not actions—the occurrence of particular considerations in the agent’s mind—and the agent’s mental act of forming a preference. The free agent acts on a decisively-formed preference, an attitude that in part constitutes her psychological identity and that is caused indeterministically by her reasons; its claim to be authentic to the agent is not undermined by the objection that she formed it because she had to form it, since it is not causally necessitated and not coercively imposed. Notice that the noncoercion condition and the requirement of indeterministic causation of the preference by the agent’s prior considerations are separate conditions.

3 New proposed event-causal indeterminist account of free action

There are reasons that speak in favor of revising the account proposed above. In particular, it seems advisable to loosen the requirement on the sorts of reasons that can motivate a free act—expanding these beyond preferences in a specially defined sense.

What about, for instance, a case in which one acts intentionally without first making a conscious decision and without engaging in any protracted process of critical evaluative deliberation concerning what is true and good (such as getting up to get a glass of water when thirsty): might not such an act count as free, if other conditions on freedom are met? One might reply that in such cases one acts from a standing preference, which may have been formed deliberatively, non-coercively, and indeterminately for reasons, and which is brought into play, allowing the act to count as freely done (Ekstrom 2000, pp. 116–117). But then what about cases in which one acts on a desire that has not, and would not, survive an agent’s critical reflection with respect to what she takes to be good, cases, for instance, of action on whim or caprice or passion? One might insist that such an action insufficiently reflects the self to count as freely done (Ekstrom 2000, pp. 117–118). If what we are interested in when we construct an account of free action is the sort of control over action needed for us to be morally responsible for it in the desert sense, this may seem a problematic response, as to many it seems that we are candidates for being morally responsible when we act from desires that do not count as preferences, and perhaps from other sorts of reasons, as well, if those desires are not compulsive and do not result from external coercion or manipulation (and any other conditions on morally responsible action are met). Should I really be let off the hook in terms of what blame I deserve if I harm someone in a fit of rage or out of a desire for revenge that I repudiate, when I prefer not to seek revenge? Perhaps so—perhaps I should be pitied—but perhaps not: the answer seems to depend on the nature of the “fit”, whether it, or the desire for revenge, truly overtakes me or ought to count as “mine” in a sense preserving the agency behind my act. Perhaps these kinds of cases are inconclusive; we might simply note that on the basis of them it may be too stringent to require that an act derive from a preference, as it has been defined, in order for that act to be free in the sense required for moral responsibility of the desert sort.¹⁶

There is a stronger reason to revise the Action on Undefeated Authorized Preference account. Clarke’s challenge to my (and to Mele’s (1995) and to Dennett’s (1981)) event-causal indeterminist accounts is that a free act on those accounts can be determined by something that is not up to the agent—which does not fit with the motivating considerations for libertarianism in the first place. To explain: the Ekstrom (2000/2003) account allows the formation of intentions to perform particular overt acts to be causally determined by the agent’s preferences. A freely made decision to act, on that account, is self-determined in being caused and justified by reasons of the agent, particularly by her preference to so act. But holding fixed the prior deliberation and holding fixed the formation of preference, the theory allows for it being the case that the agent could not have done otherwise than form that decision to act. I saw the formation of preference as being, in fact, up to the agent because done for reasons of the agent’s and indeterministically caused by those reasons without coercive influence; the agent could have done otherwise at the time than form that preference, in a categorical sense. Nonetheless I see the problem Clarke presses for my account, which is that the active formation of preference cannot itself be a free act, on pain of regress—a regress threat I noted (in 2000) and responded to by saying that the preference formation at issue need not be preceded by a prior formation of preference. The target preference formation could have been otherwise, but it cannot be self-determined, on the proposed account of that notion, or else it would require a preference in its favor, and so on backward prior to each formation of preference leading to free action.

So on the Action on Undefeated Authorized Preference model there is one act that could have been otherwise (the preference formation) and another act (decision to act or the overt bodily act) that can be deterministically caused by the preference, so that it is self-determined. But the problem is that, on this picture, there is no single act that is both self-determined and could have been otherwise.

There’s a separation in stages of the two requirements. This seemed to me at the time to be a strength of the view. It depicted our actions as flowing directly from our selves (without the liability of our actions being wayward, contrary to our preferences) and it allowed that we could have become different sorts of people, since the sorts of people we are is understood as the sorts of characters we have. Who we are, on the model, is not a deterministic outflowing of the past and natural laws. But it was wrong to say (as I did in 2000) that intentions to act that are indeterministically caused by preferences, rather than deterministically caused by preferences, are ones with regard to which the agent has to “wait to see” if they occur. When made for and caused by reasons of the agent’s in deciding, intentions are not passive states but are actively formed. And intention formations that were indeterministically caused by the agent’s formation of certain preferences would not be accidental or lucky (2000, p. 105).

We need to account for directly free actions—those the freedom of which does not derive from prior free actions—and our directly free actions should be both self-directed and ones with respect to which we have alternatives. Pereboom (2014a) also points out this problem. He writes:

...an infinite hierarchy of preferences is impossible for us, and so we must at some point arrive at fundamental preferences. On Ekstrom’s event-causal picture, such fundamental preferences can’t themselves be freely formed in the sense required for moral responsibility for them, since any such formations can’t be backed by preferences. Thus the agent will not be morally responsible for the fundamental preferences, and also not for any preference formation or decision backed by them, and this undermines the position. (2014, p. 38)

He is right. How easy is it to fix my account so that the same act both could have been otherwise (securing the alternative possibilities condition of freedom, the power and the opportunity to do other than as one does) and was self-directed? On the alternative account I wish to explore here,¹⁷ in a case of directly free action (that is, an action the freedom of which does not derive from an earlier free action), the act is caused non-deviantly and indeterministically by certain kinds of agent-involving events, namely, attitudes of hers—such as preferences, acceptances, desires, values, intentions, and beliefs—which provide a reasons explanation of the act, and, furthermore, the act is not the result of compulsion, manipulation, or coercion for which the agent herself has not freely arranged. A decision to act, conceived as the active formation of intention to act, could be free in virtue of having this sort of causal history. A decision to form a particular preference is free if it has this sort of causal history. The directly free act, so conceived, counts as self-directed in virtue of its being caused by the agent’s reasons—which on this model could include a variety of agential attitudes, not only certain ones such as preferences—and it counts as meeting the alternative possibilities condition of freedom in virtue of the fact that, at the time the agent performed the act, she could have done otherwise in a categorical sense. (Now we have no regress of preference formations, having widened the conception of the self so that a preference formation may be preceded by the occurrence of other sorts of agential reasons.)

If this sort of account is right, then an agent can act freely in acting on a desire to take a drug that she prefers she did not have and that she prefers she did not act on, so long as acting on that desire counts as a reason and does not count as an instance of compulsion, manipulation, or coercion. And the agent could act freely in taking the drug even if his taking it conflicts with his higher-order desires and his values and his convictions, so long as his taking it has the right sort of indeterministic causal history involving reasons of his. Perhaps this is the right result, if what we are concerned about in characterizing free action is action for which an agent is morally responsible in the sense grounding deserved praise and blame, rather than action that expresses the agent’s innermost self or favored conception of the self, or action from which the agent is not alienated. Perhaps the notions of autonomy, identification, ambivalence, and alienation are better suited to exploring such cases.

To recap: on the Action on Undefeated Authorized Preference event-causal libertarian account I articulated in earlier work, the decision to form a preference is unfree, simply because it is not preceded by preference; still, the decision to form that preference could have been otherwise and at the time the agent could have formed a different preference instead or could have formed no preference. It may be odd to suppose that a free act could arise from a non-free decision. But this is an artifact of the requirement for a preference to precede a free action. And so we could conclude that it is off-base to try to incorporate the early Frankfurtian and Watsonian considerations about the real self into an account of free action. Perhaps we do not need a fancy notion of self-determination and, instead, a free act can come (in the right way) from any sort of reason of the agent’s, including values, beliefs, desires, and intentions. The self-direction component of freedom is thus here broadened over that of the Ekstrom (2000/2003) account.

The newer proposed event-causal indeterminist view, then, is as follows.

Directly Free Action as Action Indeterministically and Non-deviantly Caused by Reasons of the Agent’s Own: A decision or other act is directly free just in case it is caused non-deviantly and indeterministically by reasons of the agent’s—such as convictions, desires, values, beliefs, and preferences—and other reasonable compatibilist conditions on free action are met, including that the act is not compelled and is not the result of (non-self-arranged) manipulation or coercion. An agent’s performing a directly free act requires that it be open to her at the time not to perform that action, either by performing an alternative act right then or by not performing any action at all right then.

This is what Clarke calls an “unadorned centered event-causal libertarian account” (2003, p. 71). Clarke argues persuasively that “add-on” features designed to improve an unadorned centered account—such as Nozick’s appeal to “self-subsuming” decisions, Kane’s appeal to an agent’s wanting more to act for certain reasons than for others, as well as Kane’s appeal to doubled efforts of will to decide in each of mutually incompatible ways—are not, in fact, improvements.

4 Virtues of the accounts

The event-causal indeterminist accounts of free action set out above have distinct advantages over competing accounts. Unlike agent-causal accounts, they make no appeal to primitive substance causation¹⁸. In accounting for free action, we need not appeal to agents as substances that somehow stand in a causal relation to events, in a way that is not reducible to event-causal terms. There are well-known problems in making sense of non-reducible agent-causation.¹⁹ Moreover, unlike some competing event-causal libertarian theories, neither of the above accounts has the problem of positing indeterminacy between our attempting to do something and the outcome of our attempt. In a good event-causal libertarian theory, the indeterministic causation should not be placed after an agent tries to do something; on that sort of view (for example, Robert Kane’s view), the agent tries to do something, exerting an effort of will, but then she might or might not succeed, which does not enhance but instead diminishes agent control over what she ends up doing.²⁰

Further, the proposed accounts withstand familiar objections to event-causal indeterminist theories of free action, namely problems of chance, randomness, accidentality, luck, and what Pereboom has recently posed as a “disappearing agent” problem.

In order to appreciate this, it is helpful to consider an example. Many of us are familiar with the tensions that can arise between the pleasures and obligations associated with our profession, on the one hand, and the pleasures and obligations relating to our personal lives, such as in our roles as parents or as friends, on the other. Consider an instance in which a writer has a strict deadline for getting a paper in to a volume editor, wants to meet this deadline, and feels pressure stemming from her own sense of professional identity to spend the afternoon working hard on the project. But she has a conflicting pull on her attention for that same period of time: perhaps it is a friend who is particularly lonely after a break-up and who has asked her to meet for coffee; or it may be a neighbor who is ill and who could use help in the form of getting groceries; or it could be her child, who simply wants her to spend the afternoon hours doing something fun together. We often make decisions resolving such conflicts in demands on our time. Now suppose that in this instance the writer decides to spend the afternoon working on her paper, turning down the other way in which she might have instead spent that time. And suppose that this decision was freely made, in accordance with the terms of the above account of directly free action.

When the writer elects to spend her afternoon working on her project, her doing so is not something that happens to her—rather, she freely decides to spend the time in that way for various reasons that favor her making that decision, including her preference to meet deadlines and to keep up her professional identity, and her belief that spending the time in that way will be both pleasurable and productive. We can see then, first, that a free act as depicted on the event-causal indeterminist view defended here is not random, since it is caused and justified by reasons. It is not an event that appears out of nowhere. It is not some sort of uncaused happening. It is not a miracle. It need not violate any laws of nature. And, second, we can see that a free act as defined here is not accidental, since it is done on purpose. It is done intentionally and for reasons that favor doing that act.

These points concerning randomness and accidentality are now fairly well established in free will theory. For instance, they are granted by Clarke and Pereboom, despite their rejection of event-causal indeterminist accounts on other grounds.²¹ What about the charge that event-causal indeterminist accounts of freedom are undermined by an objection from chance? This charge, in my view, can be answered successfully, as well. Elsewhere I have argued against van Inwagen’s way of posing the problem, which employs the notion of chance, by distinguishing three senses of the term ‘chance.’²² And what about luck (which is distinct from chance, for at least the reason that chance events might matter to no one, whereas events that lucky or unlucky do matter)? In supposing that the various reasons the writer has for spending the afternoon working on her paper cause the decision indeterministically rather than deterministically, have we turned her decision into a lucky event?

We can answer that question only if we are first clear on what constitutes a lucky event. Pinning down the nature of luck (as is evident in the relevant literature in epistemology, agency theory, and political philosophy) turns out to be rather complex. If we understand luck along the lines of a modal account such as Duncan Pritchard’s, then luck is not threatening to event-causal libertarianism: such luck is harmless with regard to free agency.²³ On many conceptions of luck, a lucky event is in some sense uncontrolled by the agent with respect to whom it is lucky, and this sense is pertinent to our discussion (Levy 2011). The sense of control pertinent to luck is controversial. An incompatibilist may well say that an event’s being non-lucky with respect to an agent requires that she have a kind of control over it that involves her being at the time both able to make the event occur voluntarily and rationally and able to prevent it voluntarily and rationally, as well as the opportunity for each of these abilities to be exercised, given past events and the laws of nature. When our writer decides to work on her paper for the afternoon, her decision is then arguably non-lucky.

Perhaps we can address the worry in different terms. Pereboom has recently written: “I think the objection in this family that reveals the deepest problem for event-causal libertarianism is what I call the disappearing agent objection” (2014a, p. 32). He describes the problem as a matter of the agent’s being or going missing, on the event-causal libertarian view, when it comes to settling whether or not a particular decision will occur. The event-causal-libertarian free agent, Pereboom thinks, cannot select which of the potential decision outcomes is the actual outcome of her deliberation over what to do. If some of her reasons incline toward one decision outcome and other of her reasons incline toward a different decision outcome (or toward not making a decision in the circumstances), then which way it turns out is not up to her and so is lucky and is matter with respect to which she is not present or is missing.

Take again the case of our writer deciding how to spend her afternoon. Suppose that she is in “motivational equipoise,” as Pereboom puts it. We may now ask this: with this motivational equipoise in place, what is it that settles whether the outcome supported by considerations of professional interest occurs or a different (friendship-based or relaxation-based) outcome occurs? It may seem that the event-causal libertarian can only say that when one of the outcomes occurs, it does so without anything about the agent settling that it did, since the extent to which the agent is involved is, at this point, exhausted by the preceding considerations, which by hypothesis are in equipoise. One might doubt that this picture allows sufficient agential control. Here, one might suggest, the agent-causalist has an apparent advantage, since she can say that the agent, as substance, settles which outcome occurs.

If the notion of settling that Pereboom relies on is meant to be causal determination, then it is true that nothing settles which outcome occurs, on the event-causal libertarian model. If the allegation is that it takes causal determination of a choice in order for that choice to be free, then one may reasonably, in the context of the free debate, simply deny the allegation. Notice, as was brought out in the first section of this paper, that indeterministic causes bring about their effects. The causal events—in this case the agent’s reasons—do not simply rest there inert, waiting for someone to give them causal power. (Event-causal libertarian accounts are sometimes caricatured, as if the agent considers what to do, then there is some “blip” or “magic” and the decision results.) The event-causal libertarian simply maintains that, when the agent makes the free decision, it is a decision she made for reasons that cause and justify it and at the time of making it she had the power to do otherwise—by making a different decision or not deciding, for reasons that would have caused and justified that outcome—and so the decision she made was up to her. Thus, it is something about the agent that settles which decision occurs: the agent exercises her will in making a decision for reasons. In the case of the writer, for instance, in making the decision to spend the afternoon working on her paper, she settles what to do with her afternoon. I do not see that an agent-causalist has an advantage in being able to say that the agent, as substance, settles which decision occurs, for that requires a substance to stand in a causal relation with an event. The event-causal libertarian has the advantage in being able to say that events—the occurrence of certain of an agent’s attitudes, which are considerations relevant to the decision—cause the event of the decision; that is what it is for an agent to exercise her ability to make a decision for reasons.

Consider this way Pereboom expresses his worry:

...a way must be found to enhance an agent’s control [over that provided by compatibilist accounts] to the appropriate degree. The relevant sort of control involves two factors. The first is the absence of external determination, which both event-causal libertarianism and agent causation provide. But the second is the capacity of the agent to be the source of his decisions and actions, and this is the factor that event-causal libertarianism is missing. (2001, p. 56)

My response is that it is unclear to me what is missing in the event-causal libertarian account, and that it is not merely an absence between the agent’s reasons and her directly free decision. What is alleged to be missing is “the capacity of the agent to be the source of his decisions and actions.” But the agent is the source of his decisions and actions when they are free in an event-causal libertarian way. What is the agent? As I have suggested above, it seems to me that the agent is, functionally or psychologically speaking, his intellect and his will and his collection of character-defining attitudes. In terms of our practical agency, we ourselves are involved by way of thinking, aiming, desiring, valuing, preferring, believing, and deciding.

In fact, Pereboom continues the passage just cited in this way: “To be the source of one’s decisions and actions is plausibly to be their cause” (2001, p. 56). Exactly. And the way one causes one’s decision (or other action) is by exercising one’s abilities to consider various reasons and to decide what to do (or to act) taking into account these reasons, so that the decision (or other action) is caused and justified by reasons of one’s own. That’s what it is to cause one’s decision. Pereboom seems to suggest that one can only cause one’s decision if one is causally related to the event of one’s decision as a substance. But I do not see why this is so. It seems to me that there is no lack of the presence of an agent on the event-causal libertarian account of freedom under discussion.

Notice that compatibilists who endorse a causal theory of action think that an agent is present when making a decision for reasons that cause the decision and when acting overtly for reasons that cause the action. On the event-causal libertarian theory here defended, the agent is, I say, present when making a decision for reasons that cause the decision and when acting overtly for reasons that cause the action. It is not clear why an agent should be agreed to be present in a case in which a decision is caused deterministically by reasons, while an agent is alleged to be not present in a case in which a decision is indeterministically caused by reasons. Conversely, if an agent “disappears” or is absent in a free decision as it is described by the present event-causal libertarian account, then it would seem, by parity, that an agent “disappears” or is absent in a free decision as it is described by a compatibilist account of free action according to which decisions are caused by and made for reasons. The difference between the two types of account is that, in the incompatibilist account, the prior reasons are indeterministic causes of the decision, whereas in the compatibilist account, the prior reasons are (or can be) deterministic causes of the decision. But why should indeterministic causal relations between agent-involving events (reasons and decisions or other actions) make an agent disappear, yet deterministic causal relations between agent-involving events (reasons and decisions or other actions) preserve the presence of an agent?

If it is an agent-causal libertarian who is lodging this charge against event-causal indeterminist accounts of freedom (or someone, like Pereboom, who takes the agent-causal libertarian account to be the best account of freedom, but who is not a libertarian, since he thinks it is unlikely that the terms of the agent-causal libertarian account are satisfied in fact), then such a person ought to level the same charge against compatibilist event-causal theories of free action. For if there is the disappearance of an agent in the first case, then there ought to be a disappearance of the agent in the second case. But I take this to be a reductio of the position. An agent who acts freely as an event-causal compatibilist account depicts that action is not, in my view, plausibly said to disappear. Likewise, an agent who acts freely on an event-causal incompatibilist account is not, in my view, plausibly said to disappear.

Finally, one might ask, what is the improvement in the libertarian account over the compatibilist account? Why should it be an enhancement of some sort in human agency that reasons of an agent’s cause but not necessitate her decisions? The answer, which I have alluded to above, is this: If sometimes reasons cause without necessitating, then sometimes, in the exact circumstances, holding fixed the natural laws, an agent can make one decision and can do other than make that decision (by making an alternate decision or by making no decision at the time in question). And this is so not just in the sense that the agent possesses the requisite skills for making one decision and the requisite skills for making an alternate decision, but also in the sense that the universe leaves room for the agent to exercise these skills, giving her the opportunity to make one decision or another (or no decision at all). This account thus secures the sense many of us have of ourselves as free in part because sometimes we can extend the past in one way and we can extend it in an alternate way instead: before us there are forking paths into the future. Compatibilist accounts of freedom do not capture this aspect of the sense many of us have of ourselves as free agents and the corresponding attitude we have concerning the openness of the future.²⁴

5 Conclusion

I have presented here two event-causal indeterminist accounts of free action. Both are preferable, in my view, to agent-causal libertarian theories, and they share a number of virtues, which I have outlined. The first has a more stringent conception of the self, and it depicts a valuable form of human agency in which we ourselves can direct what we do. The second loosens the bounds of the self to include a wider variety of agential attitudes, thus expanding the range of actions that will count as free. The second has attractions, such as avoiding a regress of preference formations, as I have indicated. On neither of these views are free actions merely random, accidental, or matters of chance or luck, and on neither of them does the agent “disappear.” I recommend these accounts for further discussion.