In Mele (2006), I developed a problem about luck for a standard libertarian thesis about free will (described shortly) and I offered a solution. Levy (2011, ch. 4) and Pérez de Calleja (2014) have argued that there is an analogous problem about luck for compatibilists, theorists who—unlike libertarians—maintain that free will is compatible with determinism.Footnote 1 In this article I explore similarities and differences between the alleged problems.

I conceive of free will as the ability to act freely and treat free action as the more basic notion. My primary interest in free action is, more specifically, in what I call moral-responsibility-level free action—“roughly, free action of such a kind that if all the freedom-independent conditions for moral responsibility for a particular action were satisfied without that sufficing for the agent’s being morally responsible for it, the addition of the action’s being free to this set of conditions would entail that he is morally responsible for it” (Mele 2006, p. 17).Footnote 2 By “free action,” in this article, I mean moral-responsibility-level free action.

The Problem of Present Luck

What I called “the problem of present luck” (Mele 2006, p. 66) or, more fully, “the problem of present indeterministic luck” (p. 201) is a problem for typical libertarians. I have devoted a lot of ink to making the problem salient (Mele 2005, 2006, pp. 5–9, ch. 3, ch. 5). In my brief presentation of it here, I get by with a few introductory remarks, Pérez de Calleja’s sketch of the problem, and some comments on her sketch.

I start with a complication. Libertarians have the option of distinguishing between directly and indirectly free actions (Mele 1995, pp. 207–9). For example, they can claim that A is a directly free action only if its proximal causes did not deterministically cause it and claim as well that an action that was deterministically caused by its proximal causes may be indirectly free, provided that the agent earlier performed some relevant directly free action or actions. A more specific example is provided by the claim that an agent might have freely pressed a button even if his pressing it was deterministically caused by proximal causes that included his deciding to press it straightaway, if his deciding to do that was a directly free action. According to libertarians who countenance indirectly free actions, only agents who have performed directly free actions can perform indirectly free actions. This gives directly free actions a special importance: without them, there are no free actions at all.

Typical libertarians endorse what may be termed a time-of-action alternative possibility requirement for directly free actions.Footnote 3 They can reasonably disagree about the content of the strongest acceptable such requirement. Here is one candidate: An action A performed by an agent S at a time t in a possible world W is directly free only if there is another possible world with the same laws of nature as W and the same past up to t in which (O1) S instead performs some other directly free action at t.Footnote 4 Libertarians who regard O1 as too demanding have other options. They include, but are not limited to, the following substitutions for O1:

O2. S instead performs some other action at t.

O3. S does not do A at t.Footnote 5

O4. Either (1) S does not do A at t or (2) S A-s at t, but not on his own.

(Readers who know the literature will see that O4 is of special interest to libertarians who are impressed by Frankfurt-style cases.)Footnote 6

Stories used to illustrate the problem of present luck may be expected to target whatever particular time-of-action alternative possibility requirements are under discussion. The story about luck that I discussed at greatest length in Mele (2006) featured an agent who, in one world, acted at t as she promised to act and, in another world with the same past and laws of nature, decided at t to break her promise (pp. 58–73, 123–28). Another story, featured an agent who, in one world, decided at t to cheat and in another world with the same past and laws of nature decided at t to do something else instead (pp. 73–75, 118–19). Targeted problems of present luck grow directly out of the time-of-action alternative possibility requirements that they target. Typical libertarian requirements for—that is, proposed necessary conditions for—directly free actions are claimed to generate the problems, a point to which I return in Sects. 3 and 5.

After summarizing an example of mine (from Mele 2006), in which an agent, Bob, decides at noon to cheat and in another possible world with the same laws of nature and the same past decides at noon to do something else instead, Pérez de Calleja writes:

What Mele calls “the problem of present luck” (2006, 66) is that Bob’s deciding to do the wrong thing instead of deciding to do the right thing seems to be just a matter of chance, and hence – given the decision’s significance for Bob – just a matter of luck for Bob, because the causal antecedents of both decisions are exactly the same. Bob is just unlucky to decide to cheat instead of deciding to do the right thing, because after all he did (indeed after all that happened) up to noon, each alternative decision was compatible with the past and the laws, and it was mere chance that one was made at the time instead of the other. (2014, p. 114)

Where she writes “What Mele calls ‘the problem of present luck’,” I would add “as that problem is raised in this particular story.” At any rate, readers should understand that the problem of present luck is not only about competing decisions.

Bob’s story is about an agent in a situation in which both weak-willed action and its contrary seem to be possible. “He judged it best on the whole to do what he agreed to do” (Mele 2006, p. 74)—namely, toss a coin at noon to start a football game. And he was tempted by the offer of fifty dollars to help a gambler win his bet that the coin toss would occur at 12:02. In the actual world, at noon, Bob decides to cheat and to toss the coin at 12:02 after pretending to be searching for it in his pockets. “In a possible world that does not diverge from the actual world before noon, Bob decides at noon to toss the coin straightaway and acts accordingly” (p. 74).

Someone may assert that the worlds diverge as they do at noon because, in both worlds, it is up to Bob what he decides and he decides differently in these worlds. But, one may ask, is it any more up to Bob at noon whether, right then, he decides to cheat or instead decides to flip the coin than it is up to a genuinely random number generator whether the number it outputs at noon is 7 or 11 in a scenario in which it has only a few possible outputs at the time (see Mele 2013, p. 244)? In any case, this is among the questions raised by the problem of present luck. One who poses the problem may hope for a persuasive defense of a plausible answer.

As Pérez de Calleja notes, my discussion of the problem of present luck includes no luck-based argument against libertarianism (2014, pp. 122–23 n. 3). I have never argued for the falsity of libertarianism, and, as I mentioned, I offered libertarians a solution to the problem of present luck. As I observe elsewhere, I regard my central question about present luck as an analogue of a request for a theodicy in response to the problem of evil—an explanation of why a perfect God would allow all the pain and suffering that exists and, of course, has ever existed and will ever exist (2013, pp. 241–42). Philosophers who offer an argument from evil for the nonexistence of God should expect rebuttals of the argument and should not be surprised if the rebuttals include no theodicy. Philosophers who present the problem of evil as vividly as they can and then ask for an explanation of why a perfect God would allow all the pain and suffering at issue may hope to receive an attractive explanation. For my own analogue of a theodicy regarding present luck in the spheres of free will and moral responsibility, see Mele (2006, ch. 5).

A novice may claim that because the problem of present luck is generated by a typical libertarian requirement for directly free actions it cannot be a problem for libertarianism. An obvious problem with this claim is that something that someone asserts to be a necessary condition for X can be incompatible with X. Consider, for example, the idea that free will requires determinism, which has had some advocates. If incompatibilists are right, that alleged necessary condition for free will is incompatible with free will.

Closing in on Close Luck

Return to Bob’s story. Obviously, it is set in a world in which determinism is false. If, in a deterministic world, Bob decides at noon to cheat, then in any world with the same past and the same laws of nature he decides at noon to cheat. Levy and Pérez de Calleja recognize this, of course, and they maintain that the luck featured in Bob’s story—what I call present luck—is analogous to luck at or near the time of action in some deterministic scenarios (Levy 2011, pp. 89–93; Pérez de Calleja 2014).Footnote 7 I dub what they have in mind close luck—or C-luck for short.

Consider the following from Pérez de Calleja (2014):

a decision performed at a time at which the agent is psychologically able (and suitably skilled and placed) to refrain from deciding that way in the circumstance is subject to cross-world luck, whether the world is deterministic or indeterministic. Bob, in particular, is lucky that he decides as he does rather than otherwise because, in a nearby possible world where the salient causes and background conditions which are relevant to his deciding one way or the other in the circumstances are the same (including his reasons, his character traits and even his way of deliberating), he decides otherwise instead. If we make Bob’s world deterministic, we don’t thereby eliminate the nearby possible worlds where Bob’s counterpart does otherwise in conditions which don’t significantly differ from Bob’s. (Though, of course, since the worlds at issue are deterministic, either the past or the laws must differ in some way to produce different outcomes.) (pp. 114–15)

In the same vein, she writes: “the fact that Bob* is determined to decide to cheat rather than deciding to do the right thing isn’t grounded in the attitudes, tendencies, skills and features of the circumstance which are relevant to what he is motivated and disposed to do in the circumstance” (p. 115) and “the nearest possible world where Bob*’s counterpart voluntarily refrains from cheating (let’s say it’s Bob**’s world) is one where nothing significant changes in his relevant motivations and condition, and Bob** refrains for the moral reasons for which Bob* was about to refrain when he decided to cheat” (p. 115).

Pérez de Calleja’s way of getting at C-luck in deterministic scenarios involves holding certain features of a story fixed and varying others when moving from the world in which the original story is set to another possible world in which the story is modified. In the passages just quoted, she holds fixed “the salient causes and background conditions… relevant to [the agent’s] deciding one way or the other in the circumstances,” where these things are understood to include the agent’s “reasons,” “character traits” and deliberation. Also held fixed here are “the attitudes, tendencies, skills and features of the circumstance… relevant to what [the agent] is motivated and disposed to do in the circumstance” and “relevant motivations.” The word “relevant” occurs three times in the preceding two sentences, and it would be good to have something approaching an account of what it is for a factor to count as relevant to an agent’s “deciding one way or the other” and to what an agent is “motivated and disposed to do,” in the intended sense of the word. (The same is true of what it is for a cross-world difference to count as significant.) But it is possible to proceed without further guidance about this than Pérez de Calleja offers. I refer to whatever is to be held fixed as belonging to the collection of privileged factors, and I leave the question open exactly what range of factors should be privileged.

What sorts of thing might it be permissible to vary in composing stories about C-luck in deterministic possible worlds? The situationist literature in psychology is a potential source of inspiration. I offer just one example. Robert Baron found that “passersby in a large shopping mall were significantly more likely to help a same-sex accomplice (by retrieving a dropped pen or providing change for a dollar) when these helping opportunities took place in the presence of pleasant ambient odors (e.g., baking cookies, roasting coffee) than in the absence of such odors” (1997, p. 498). The presence of pleasant odors significantly increased helping behavior. And if the helping was preceded by decisions to help, the presence of pleasant odors significantly increased the number of decisions to help. Obviously, the people who helped did not take the pleasant odor to be a reason to help. The influencing role at issue was played by non-reasons.

Consider a story about Bob and the coin set in a deterministic possible world, W1. I call the story Hungry Bob. In W1 Bob decides at noon to cheat. But there is a deterministic possible world, W2, in which the wind is just a bit stronger shortly before noon and Bob gets a whiff of some of the delicious food cooking in the vending area. This boosts his mood just a bit, which in turn slightly boosts the strength of his desire not to cheat people out of their money. This brief sequence of events tips the scales: at noon, Bob decides to toss the coin straightaway. Although, Bob’s getting a whiff of the pleasant odor is causally relevant to the strength of his motivation to refrain from cheating and to what he decides to do, it might be thought not to have the kind of relevance that Pérez de Calleja has in mind. In short, it might be thought not to be what I called a privileged factor. (Notice the words “might be thought.” I return to this issue in Sect. 4.)

Under what conditions does at least some of the luck involved in a story about a decision to A made at a time t in a deterministic world W count as C-luck? In the absence of an account of the collection of privileged factors, I cannot give a precise, detailed answer. The following answer may do for present purposes. It is sufficient for at least some of the luck involved in such a story to count as C-luck that at least some of the luck is associated with there being another possible world in which the collection of privileged factors is held fixed and, even so, at t, the agent does not make that decision, and this difference is explicable partly in terms of a difference near t in peripheral scale tippers.

We can say that in W2 Bob had the good luck to smell the pleasant odor (“good” because it tipped the scales in favor of the “better” option). When we compare W1 with W2 and zoom in on noon and the preceding second or so, we may wonder about the control Bob has over what he decides at the time. If the scales are tipped by something seemingly as trivial as whether or not he gets a pleasant whiff of food, does Bob in these worlds really exercise all the control over what he decides that he needs to exercise in order to make a directly free decision? Does he exercise freedom-level control in either world in making his decision? Can what an agent decides turn on something seemingly so trivial if he is exercising freedom-level control over what he decides? Perhaps, if an agent decides among morally insignificant options in a case in which he correctly regards the options as being of equal value, we would not worry much about control. But Bob’s case obviously is not of that kind.Footnote 8

The questions just posed resemble questions raised by some stories that feature present luck. But, as I explain shortly, my discussion of C-luck has not yet captured a certain important feature of the problem of present luck.

The Problem of Global Close Luck

Typical libertarians hold that an action A performed at t is directly free only if there is another possible world with the same past and laws in which, at t, the agent does not do A. (This is compatible with their also endorsing stronger time-of-action alternative possibility requirements.) This requirement has an upshot that is important for the purposes of this article: in the case of typical libertarian views, the problem of present luck is a problem about every action that is a candidate for being directly free. Call this the global aspect of the problem of present luck. Pick an agent—Bob, say. If it is claimed by a typical libertarian that he has performed thousands of directly free actions in his lifetime, then the problem of present luck is a problem about each and every one of these actions.

To produce a problem about C-luck that is more similar to the problem of present luck than any problem posed by a non-global case of C-luck, we should imagine an agent in a deterministic possible world who is such that every candidate for a directly free action that he performs occurs in the context of C-luck. (I dub such an agent a globally C-lucky agent—or a GCL agent, for short.) This would bring the global aspect of the problem of present luck on board. Levy makes it clear that he does not view C-luck as posing a global problem (2011, pp. 93–94). That leaves Pérez de Calleja to carry the ball.

For purposes of discussion, a directly free action in a deterministic world may be roughly characterized as a free action that the agent was able to do otherwise than perform, on whatever compatibilist conception of “able to do otherwise” the reader finds most attractive.Footnote 9 We should also imagine that the putative directly free actions the agent performs are about as frequent as those of a typical allegedly free agent on a typical libertarian view. This will strengthen an analogy between C-luck and present luck. And when an action at issue is a decision, we should imagine that the putative directly free decisions the agent makes are about as frequent as those of a typical allegedly free agent on a typical libertarian view.

A bit more precision in my working characterization of a directly free action in a deterministic world would be useful. For some purposes, it is important to include temporal indices in ability claims. Here is an illustration. At noon, Joe is able to finish reading his homework assignment by 3:00. In fact, he has an hour to spare; he is capable of completing the assignment in two hours of continuous reading. It is now 2:00, and Joe has not yet started reading. He is no longer able to finish reading the assignment by 3:00, and he has not been able to do this for quite a while. How does this apply to C-luck? Because C-luck is located very close to the time of action, the ability to do otherwise that is at issue needs to be located there too.

Consider a GCL agent in a possible world Wc. The world is deterministic, of course, given what it is to be a GCL agent.” Every time this agent makes a decision that purports to be directly free, something is at work around that time that falls outside the collection of things that Pérez de Calleja wishes to hold fixed and outside the collection of privileged factors. This thing is a scale tipper. In Hungry Bob, the scale tipper was the presence or absence of a certain pleasant odor. (In one world, the presence of the pleasant odor tips the scales one way; and in another world, the absence of the odor tips the scales the other way.) And in every relevant case, the scale tipper is roughly as peripheral as this to the matter the agent is making a decision about. Our GCL agent, Chuck, always decides for reasons, but different reasons win out—for example, either Chuck’s reasons to cheat or his reasons not to cheat—in some different worlds in which everything that Pérez de Calleja wants to hold fixed (and everything in the collection of privileged factors) is present, owing to differences in peripheral scale tippers—henceforth, PSTs.

How does Chuck strike you? Some incompatibilists may say that they have no interest in Chuck, because his inability to make free decisions is already ensured by his world’s being deterministic. Some compatibilists who are committed to sufficient conditions for free action that are satisfied by decisions Chuck makes in the presence of C-luck may say that they have absolutely no doubt that Chuck decides freely in these cases, because they are certain that their view about sufficient conditions is correct. But many people may find reflection on the current story interesting. Some incompatibilists may find in the story the basis for an objection to some proposed compatibilist sufficient conditions for free action. Some compatibilists may experience an urge to consider whether sufficient conditions for free action that they have offered should be revised. And, of course, there are agnostics about compatibilism and compatibilists who have not yet officially (or unofficially) endorsed any particular set of sufficient conditions for free action.

Attention to a pair of differences between the problem of present luck and what I dub the problem of global C-luck is in order. One difference is obvious: whereas the problem of present luck is set in worlds that do not diverge before the time of action, no problem of C-luck is set in such worlds. For example, where we find C-luck across a pair of worlds in which an agent makes different decisions at t, there is a difference in PSTs before t.

A second difference is that whereas the problem of present luck is generated by an explicit requirement of typical libertarian views of directly free actions, the problem of global C-luck is not generated by any explicit compatibilist requirement for directly free actions.Footnote 10 As I reported, typical libertarians hold that an action A performed at t is directly free only if there is another possible world with the same past and laws in which, at t, the agent does not do A; and this claim itself raises the problem of present luck. But no compatibilist has claimed that an action A performed at t is directly free only if (1) there is another possible world in which the collection of privileged factors is present and the agent does not do A at t and (2) this difference in what happens at t is explicable partly in terms of a difference near t in PSTs.

Someone may contend that compatibilists are committed to the claim just mentioned about directly free actions, even if they do not realize it. That contention is among the topics discussed in Sect. 4. Parts of the remainder of the present section prepare the way for that discussion.

In a passage quoted earlier, Pérez de Calleja asserts that “the nearest possible world where Bob*’s counterpart voluntarily refrains from cheating (let’s say it’s Bob**’s world) is one where nothing significant changes in his relevant motivations and condition, and Bob** refrains for the moral reasons for which Bob* was about to refrain when he decided to cheat” (2014, p. 115). Perhaps this is true in some cases; but no compatibilist should make it a requirement for the direct freedom of a decision to A made at t that in the closest possible worlds in which at t the agent (or his counterpart) voluntarily refrains from making that decision, there is no significant difference in the agent’s “relevant motivations and condition”—or in the privileged factors.Footnote 11 The following certainly seems conceivable. At noon in deterministic world W, Bob decides to cheat, and some other possible world in which he instead voluntarily refrains from making that decision then, was offered less money to cheat (say, forty dollars instead of fifty), and no PST is involved in his story at the time is closer to W than any possible world in which he voluntarily refrains from making that decision then and a PST is at work around noon. The view that an agent’s decision at t in a deterministic possible world may be directly free only if the closest possible worlds in which he voluntarily refrains from making that decision at t feature a PST around the time of the decision, and not if they feature a difference in incentives or other privileged factors, seems to be untenable. It should not be accepted in the absence of a convincing argument for it.

To forestall confusion, I distinguish between two different questions about a case in which at noon in deterministic possible world W, Bob decides to cheat.

  • Q1 What are the closest possible worlds like in which Bob voluntarily refrains from making that decision then?

  • Q2 What are the closest possible worlds like in which Bob voluntarily refrains from making that decision then and all the privileged factors are present?

Q1 will be posed by someone wondering how things would have been if, at noon, Bob had voluntarily refrained from deciding to cheat. Q2 will be raised by someone in search of C-luck and therefore of a PST. Only when the closest possible worlds to W in which Bob voluntarily refrains at noon from deciding to cheat include the presence of all the privileged factors will correct answers to these two questions be the same.

Return to the first option for a libertarian time-of-action alternative possibility requirement on directly free actions stated above: An action A performed by an agent S at a time t in a possible world W is directly free only if there is another possible world with the same laws of nature as W and the same past up to t in which (O1) S instead performs some other directly free action at t. Notice that the alleged requirement here is that there be some possible world with the same past and laws of nature as W in which, at t, S performs some other directly free action—not that the closest worlds with the same past and laws in which S does not do A at t have this feature. Perhaps, in many cases, no world in which S performs some other directly free action at t is closer to W than some world in which S A-s just a bit later. The problem of present luck that arises out of this alleged requirement is not framed in terms of the closest possible worlds in which the agent does not do A at t; and if there is a closely analogous problem of C-luck, the same is true of it (by virtue of being closely analogous).

What is the range of possible worlds that bear on the problem of global C-luck? That is, what is the range of possible worlds that it is permissible to appeal to in looking for global C-luck in a story. When spinning a story about present luck in a case in which an agent A-s at t in W (an indeterministic world, of course), we know that possible worlds with different laws than W and possible worlds with a different past than W are out of bounds. And all possible worlds with the same laws as W and the same past up to t in which the agent does not do A at t are in bounds.Footnote 12 But where do we draw the exclusion and inclusion lines when spinning a story about global C-luck? This difficult question should become salient once it is clear that a close analogue of the problem of present luck is not framed in terms of the closest possible worlds in which the agent does not do A. Instead, it is framed, one may say, in terms of whatever possible worlds are relevant. In the case of the problem of present luck, as I have explained, we have a good grip on criteria for relevance. The matter is not so easy in the case of the problem of global C-luck.

Recall Pérez de Calleja’s claim that “a decision performed at a time at which the agent is psychologically able (and suitably skilled and placed) to refrain from deciding that way in the circumstance is subject to cross-world luck,” whether or not the world is indeterministic (2014, p. 114). By “a” at the beginning of this quotation, she means “any” (not “one”). But is it true that each and every decision made under the stated conditions in any deterministic possible world is subject to C-luck? Is it true that a PST is at work around that time in the case of every decision made under these conditions? Justified confidence about how these questions are to be answered depends on, among other things, confidence about how PSTs are to be conceived. I take up that issue in the next section.

Peripheral Scale Tippers Revisited

In Sect. 2, I suggested that Bob’s getting a whiff of a pleasant odor might be thought not to be a privileged factor—that is, not to be something that should be held fixed when looking for C-luck in his story. However, I am not sure that Pérez de Calleja would agree that it should not be held fixed. In Hungry Bob, the pleasant odor of cooking food slightly boosts the strength of Bob’s desire not to cheat people who bet on the coin toss. And, as I mentioned, Pérez de Calleja looks for C-luck in worlds in which “nothing significant changes in [the agent’s] relevant motivations” (2014, p. 115). The desire at issue is a relevant motivation. But would Pérez de Calleja count the slight change in its strength as significant? I am unsure.

Given my uncertainty about this, I look for an alternative kind of peripheral scale tipper (PST), something that has no effect on the strengths of the agent’s relevant desires or motivations in addition to not being in—and having no effect on—what has already been placed in the collection of privileged factors. In his discussion of C-luck, Levy mentions considerations that come to mind during deliberation, wandering attention, “chance features of our environment” that affect deliberation, chance variations in “the force with which considerations strike us,” and moods, along with the influence of moods on “what occurs to us” and on “what weight we give to the considerations that do cross our mind” (2011, p. 90). But a good list of things that tend to affect desire strength would include all these things. Also, Pérez de Calleja holds the agent’s deliberation fixed in her search for C-luck (2014, pp. 114–15). That itself excludes some of the items just mentioned.

If PSTs can have no effect on the strength of relevant desires, where should we look for them? Some philosophers have claimed that the following thesis or something similar is a truism: (T) Whenever we act intentionally, we do, or try to do, what we are most strongly motivated to do at the time.Footnote 13 Randolph Clarke has mounted a challenge to theses like T that bears directly on the present issue. He invites his readers to imagine a scenario in which a ping pong ball is situated on a smooth surface and two jets of air strike it from precisely opposite directions (1994, p. 5). If the surface is an incline, the ball might move in the direction of the weaker jet of air. What might play the role of the incline when an agent’s desires compete? Clarke writes: “There are apparently possible cases where we could have very good evidence that neurological conditions causally affect which action is performed and yet no change whatsoever in the balance of motivational strength has occurred” (p. 8). Elsewhere, I argue that Clarke’s attempt to substantiate this claim about good evidence is unsuccessful (Mele 2014, pp. 372–75). But my concern now is the mere possibility of PSTs of a certain kind.

What I am after are possible PSTs outside the collection of privileged factors that have no effect—either directly or indirectly—on anything in that collection, a collection that now explicitly includes (in addition to all the factors mentioned earlier) the strength of relevant desires. Of course, the PSTs are supposed to be at work around the time of action, given Levy’s and Pérez de Calleja’s descriptions of the problem of C-luck (see n. 7). It can be imagined that the PSTs are neural processes that can have a biasing effect on the production of decisions and other actions without having any effect on the agent’s reasoning about what to do, his reasons for action, what he desires, the strength of his desires, his beliefs about the value of his options, and so on. It can be imagined as well that differences in such PSTs across deterministic worlds can account for such things as the difference between Bob’s deciding at noon to cheat in deterministic world W1 and his deciding at noon to toss the coin in deterministic world W2.Footnote 14

If PSTs of the sort at issue now were to play a role in the case of every putative directly free action in some deterministic possible world, how great a threat would that be to the occurrence of directly free actions? This is a question for compatibilist readers, other readers with no commitment to incompatibilism, and any incompatibilist readers who are willing to set aside their incompatibilism temporarily. More information about the PSTs would prove useful for theorists looking for an answer. Are they shaped in any interesting way by the agent’s past behavior? For example, does a history in which one successfully resists temptation much more often than one succumbs to it normally help to generate a scale tipping system that favors a high rate of successful future resistance? Does a lot of caving into temptation normally result in a scale tipping system that favors an increased frequency of caving? Do efforts to resist temptation have any effect on potential PSTs at the time the efforts are made?

For dramatic effect, PSTs can be represented as always casting the deciding vote, as it were, in cases of C-luck. If the answer to my last four questions is yes, agents have some influence on which way these votes go. And if we were to discover that our world is one in which these PSTs are always at work in putative directly free actions, some of us might work hard to shape our PSTs in our preferred direction, in the belief that we can mold them to improve the chances of, for example, achieving our long-term goals. In light of this belief, some of us might see ourselves as morally responsible, to some substantial extent, for the votes our PSTs cast, and we might associate that moral responsibility with free will.

Suppose that our PSTs are not subject to our influence in any of the ways mentioned. In that case, we may put forward candidates for election—for example, such candidates as deciding to cheat, keeping one’s promise, and thinking a bit longer about what to do. But when a candidate is victorious, the outcome is settled by a PST whose vote we did not influence.

A world in which agents have significant influence of the sorts I have mentioned on their PSTs is analogous to worlds in which my solution to the problem of present luck is set (Mele 2006, ch. 5). More specifically, the analogy is between shaping the antecedent probabilities of future decisions through past conduct in indeterministic worlds and shaping through past conduct the ways in which one’s PSTs are disposed to work in deterministic worlds. A version of this solution may be worth considering in the case of the problem of global C-luck, but this is not the place for that undertaking. (On kinds of conduct that have special importance in the present connection, see Mele 2006, ch. 5.)

Imagine a possible world Wd in which it is a law that nothing influences what any agent decides to do unless it influences—directly or indirectly—what options he is considering when he makes his decision or how motivated he is to pursue one or more of them. (Deliberation, information gathering, and many other things influence what agents decide to do in this world.) Imagine that lots of agents make lots of decisions in Wd. Now, a change in the options an agent is considering is a change in his deliberation; and Pérez de Calleja holds agents’ deliberation fixed across relevant possible worlds for the purposes of locating C-luck. In short, PSTs are not supposed to do their work by influencing deliberation. Suppose that PSTs are conceived as in the present section. Then they have no influence on motivational strength either. It would seem, then, that PSTs never play a role in Wd in settling what an agent decides to do.Footnote 15 And if they never play a role in Wd, the direct freedom of decisions made in Wd is not threatened by C-luck. (Recall that C-luck is present in a story only when some PST or other is playing an explanatory role.)

It can be stipulated that some of the decisions made in Wd satisfy sufficient conditions for directly free actions offered by prominent compatibilists, including the ability to do otherwise. (The law about decisions that I described does not preclude satisfaction of these conditions.) So we have candidates for directly free decisions in a deterministic world that are not subject to C-luck. Pérez de Calleja claims, as I mentioned, that “a decision performed at a time at which the agent is psychologically able (and suitably skilled and placed) to refrain from deciding that way in the circumstance is subject to cross-world luck, whether the world is deterministic or indeterministic” (2014, p. 114). The scenario just developed falsifies this claim on the construal of PSTs at work now. So, at least on this construal of PSTs, we have another disanalogy between the problem of present luck and the problem of global C-luck. Whereas the problem of present luck applies to all possible candidates for directly free action, on a typical libertarian view of such actions, the problem of C-luck does not apply to some possible candidates for directly free action, on a representative compatibilist view of such actions.Footnote 16

This is not to deny that there is luck—including decision-influencing luck—in many deterministic possible worlds. What is at issue now are the nature of C-luck and how we are to understand the problem of global C-luck and the PSTs that play an important role in statements of that problem.

Questions and Answers

The present section poses some questions and offers some answers. This is an efficient way to cover the remaining ground.

Question 1 Why is it so much easier to know what to hold fixed in formulating the problem of present luck than in formulating the problem of global C-luck?

Any time-of-action alternative possibility requirement for directly free actions advocated by a libertarian tells us what to hold fixed in thinking about whether the agent of an action A in a possible world W satisfies the requirement—namely, all of W’s laws of nature and the entire history of W up to the time of action. Precisely this is what is held fixed in formulating the problem of present luck. Instructions about what to hold fixed are simply read off from the time-of-action alternative possibility requirement. This is easy.

If one were to try the same strategy with compatibilist requirements for directly free action, one would find that it is permissible—and, in fact, encouraged—to vary such things as an agent’s reasons for action (a point to which I return shortly). Holding fixed the things that compatibilists tend to hold fixed—for example, the agent’s capacity to recognize reasons and to act on the basis of reasons—does not straightforwardly generate worries about global C-luck. So theorists who wish to formulate an analogue of the problem of present luck for compatibilists need to ponder what to hold fixed. Other things being equal, pondering is harder than reading simple instructions.

Question 2 Do compatibilists have just as much of a commitment to solving the problem of present luck as typical libertarians do?

We must distinguish between two very different theses: (1) agents can act freely in indeterministic worlds; (2) agents can act freely in scenarios featuring present luck. The overwhelming majority of compatibilists accept thesis 1. But the problem of present luck is not a problem about indeterministic scenarios in general; it is a problem about scenarios in which at no time is it determined what the agent will do at t. (An event causalist about free actions may represent the problem as being about scenarios in which there is an indeterministic causal connection between the proximal causes of an action and the action.) A compatibilist who worries or believes that decisions made under these circumstances are not produced in the right way to be free decisions need not seriously entertain the idea that the truth of determinism is necessary for free decisions. The truth of the proposition that there are local deterministic causal connections between the proximal causes of many of our decisions and those decisions does not depend on the truth of determinism. A compatibilist who has the worry or belief at issue can make a case for directly free decisions outside the sphere targeted by the problem of present luck and do so without claiming that such decisions are possible only in deterministic worlds. (This is not to say that I recommend arguing that all directly free decisions are outside the sphere at issue. Reminder: I believe that there is a solution to the problem of present luck.)

The problem of present luck is generated by an alleged necessary condition for directly free action embraced by typical libertarians—that an action A performed at t is directly free only if there is another possible world with the same past and laws in which, at t, the agent does not do A—and compatibilists do not embrace this condition. When the problem of present luck is a problem for particular compatibilists, it has another source. For example, it may be that they endorse sufficient conditions for directly free action that generate counterintuitive verdicts in some cases of present luck. The bottom line is that compatibilists (qua compatibilists) and typical libertarians do not have the same commitments regarding the problem of present luck.

Question 3 The stories I use to illustrate the problem of present luck are not framed in terms of the closest possible worlds in which at t the agent does not do A. So why is the luck in those stories a problem?

Suppose that at t in W Bob decides to cheat and there is some possible world with the same laws of nature and the same past up to t in which at t Bob decides not to cheat. Reflection on this supposition can lead pretty quickly to worries about the control Bob has over what he does. Timothy O’Connor, for example, contends that an upshot of typical event-causal libertarian views is that agents do not directly control what they choose: “There are objective probabilities corresponding to each of the [possible choices], 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” (2000, p. xiii; see p. 29). O’Connor believes that agent causation solves the problem (2000), and I do not (Mele 2006, ch. 3). I will not revisit that disagreement here. The point now is that a worry about Bob’s control is not difficult to motivate even if the closest possible worlds in which at t Bob decides not to cheat are no closer to W than some worlds in which at t Bob continues to think about whether to cheat. The difference between deciding at t to cheat and deciding at t not to cheat is more gripping than the difference between the former and continuing to think at t about whether to cheat, even if both differences are matters of luck. More gripping differences are better than less gripping ones for the purposes of making the problem of present luck salient. So stories about pairs of worlds with the same past and laws in which what the agent does at t is dramatically different are more useful for this purpose than stories that lack this drama.

Question 4 If I were to pose a problem about C-luck for compatibilists about free will, how would I go about it?

I would start by looking carefully at sufficient conditions compatibilists have offered for directly free action. Then I would try to construct cases in which those conditions are satisfied by an action A and, even so, owing to the particular role played by C-luck in the cases, an intuitive reaction is that A is not a directly free action.

Notice how different this strategy is from my strategy for developing the problem of present-luck. In the latter connection, I take my lead from an alleged necessary condition for directly free action endorsed by typical libertarians. If that condition is in fact necessary for directly free action, and if the problem of present luck is an insoluble problem for it, then the existence of directly free action is in dire straits: either such action is a mystery (see van Inwagen 2000) or it is impossible. If the existence of directly free action is in dire straits, then so is libertarianism, given that one of libertarianism’s essential planks is the assertion that there are directly free actions. Of course, as I said, I believe that the problem of present luck has a solution.

Suppose that some case featuring C-luck falsifies a set of sufficient conditions for free action offered by a particular compatibilist. How strong a blow would that be against compatibilism? That depends on a variety of things. For one thing, the alleged set of sufficient conditions might have been implausible to begin with, even before C-luck entered the picture. But suppose the proposed set of sufficient conditions is very attractive from a compatibilist perspective. Then one would want to know whether the counterexample featuring C-luck can be satisfactorily dealt with by beefing up the proposed sufficient conditions in a way that is not ad hoc. This sort of thing has happened in connection with some threats to compatibilism that feature manipulation: for example, some philosophers have responded to such threats by weaving a historical condition into a proposed compatibilist set of sufficient conditions for free action (see Fischer and Ravizza 1998; Mele 1995, 2006). And it might happen in response to well-formulated, powerful objections featuring C-luck that may emerge.

An illustration will prove useful. In my opinion, the most detailed compatibilist account of free action is to be found in John Fischer’s work, including work co-authored with Mark Ravizza. Fischer and Ravizza report that “the focus of… attention” in their coauthored book is “the freedom-relevant condition on moral responsibility” (1998, p. 13). Elsewhere, Fischer writes: “We place value on acting freely, or, in other words, exercising a distinctive kind of control: guidance control” (2006, p. 21). Here, acting freely is identified with exercising guidance control, and Fischer and Ravizza identify “guidance control of some act A” with A’s resulting from “the agent’s own, moderately reasons-responsive mechanism” (1998, p. 101). They say the following about such mechanisms:

A mechanism of kind K is moderately responsive to reason to the extent that, holding fixed the operation of a K-type mechanism, the agent would recognize reasons (some of which are moral) in such a way as to give rise to an understandable pattern (from the viewpoint of a third party who understands the agent’s values and beliefs), and would react to at least one sufficient reason to do otherwise (in some possible scenario). That is, a mechanism is moderately responsive to reason insofar as it is “regularly” receptive to reasons (some of which are moral), and at least weakly reactive to reasons. (pp. 243–44)

Fischer and Ravizza note that “regular receptivity to reasons, in the sense that is required for guidance control… requires at least that the agent not be substantially deluded about the nature of reality” (1998, p. 73). They also report that mechanisms are an agent’s own if and only if he makes them his own, and an agent makes a mechanism “his own” by “taking responsibility” for it (p. 241). Taking responsibility requires that the agent “accept that he is a fair target of the reactive attitudes as a result of how he exercises [a certain kind of] agency in certain contexts” (1998, p. 211).

In the preceding paragraph, readers find an alleged set of sufficient conditions for performing an action freely. It can be assumed that directly free action (construed in a semicompatibilist way, given that Fischer and Ravizza are semicompatibilists: see n. 9) is at issue. Elsewhere, I offer counterexamples to the set of conditions at issue and suggest that some conditions be modified to deal with the problems (Mele 2006, 146–57). But the counterexamples do not feature luck. The issue before us now is what a powerful counterexample featuring C-luck would look like. Given Levy’s and Pérez de Calleja’s suggestion that there is a problem of C-luck that is highly similar to the problem of present luck, a counterexample featuring global C-luck would be especially interesting. The global aspect of the problem of present luck is a prominent feature of that problem

Imagine a story about global C-luck that features PSTs of the kind described in Sect. 4. They are neural processes that have a biasing effect on the production of decisions and other actions without having any effect on the agent’s reasoning about what to do, his reasons for action, what he desires, the strength of his desires, and so on. And, for dramatic effect, they are represented as always casting the deciding vote when agents perform actions that are candidates for being directly free. Imagine also that these PSTs cannot be shaped or influenced by agents’ past behavior. Call these PSTs the ubiquitous inflexible real settlers.

I have absolutely no doubt that there are neural influences on our actions. But for all we know, neural processes play a role in settling whether, for example, we decide to do one thing or another at a time only by influencing—directly or indirectly—such things as (or the neural or microphysical realizers of such things as) what we desire, the strength of our desires, how we reason about what to do, which options we consider, our beliefs about the value of various courses of action, or the like. The hypothesis that whenever it seems to us that we are acting freely, the deciding vote is cast by some member of the collection of ubiquitous inflexible real settlers and the most we can do is present the settlers with candidates to vote for is seriously at odds with how we think of ourselves as agents. So Fischer and Ravizza might argue that an attempted counterexample to their proposed sufficient conditions for free action that features the ubiquitous inflexible real settlers violates the requirement that “the agent not be substantially deluded about the nature of reality” (1998, p. 73). Agents are parts of reality, after all. If this requirement is unsatisfied, then given that it is part of the sufficient conditions they propose for acting freely, the attempt at a counterexample fails.

Suppose that to forestall this response one sets one’s attempted counterexample in a world in which typical human agents believe in the ubiquitous inflexible real settlers. Then Fischer and Ravizza might contend that another of their requirements is violated (so that this plank in their proposed sufficient conditions for acting freely is unsatisfied). Arguably, agents who believe in these PSTs would not accept that they are fair targets of the reactive attitudes. If it is part of the story that they do accept this and if it is made clear that when agents in this imaginary world perform actions that are candidates for being directly free, mechanisms that are “moderately responsive to reason” are at work, Fischer and Ravizza might consider beefing up their proposed sufficient conditions for free action to deal with the story. (The beefing up might include an account of how reasons-responsive mechanisms produce actions when those actions are directly free.) Alternatively, they may argue that these agents do sometimes act freely, despite the role played by the settlers.

Developing a problem of C-luck that closely resembles the problem of present luck is a project for other philosophers (and possibly for me on another occasion). A major part of that project will be working out what should be held fixed across worlds and what it is permissible to vary across worlds in searching for or identifying C-luck. What I want to emphasize in my response to question 4 is the difference in the potential roles played by the problem of present luck and the problem of global C-luck. If, as typical libertarians claim, an action’s being directly free depends on there being another possible world with the same past and laws in which, at t, the agent does not do A, and if there being such a world precludes A’s being directly free, then directly free action is impossible and libertarianism is false (because there are no directly free actions). This is where the problem of present luck points. The worry is that something that typical libertarians require for directly free action actually precludes free action and therefore renders libertarianism false. (Again, I believe that the problem is soluble.) Matters are very different with the problems of C-luck (global and otherwise). If some compatibilist set of sufficient conditions for directly free action is falsified by a story featuring C-luck, then that set should be rejected. But this alone cannot topple compatibilism. Whether compatibilism is toppled by some problem of C-luck depends on whether, in principle, compatibilists are in a position to propose a set of sufficient conditions for directly free action that avoids the problem in a way that is not ad hoc. At bottom, the dialectical difference between the problem of present luck and problems highlighting C-luck stems from the fact that the former problem targets a necessary condition for directly free action endorsed by typical libertarians and the latter problems do not target any necessary condition for directly free action endorsed by compatibilists.

Question 5 Are you claiming that luck poses no problems for compatibilists?

No, and I believe that working out the details of a problem of C-luck that closely resembles the problem of present luck would be an interesting project. In Mele (2006), I mentioned a more familiar problem about luck for compatibilists: “Compatibilism also is challenged by a kind of luck. Incompatibilists want to know how agents can be morally responsible for actions of theirs or perform them freely, if, relative to their own powers of control, it is just a matter of luck that long before their birth their universe was such as to ensure that they would perform those actions” (p. 77). (Of course, to say that something is a problem is not to say that it has no solution.) One may search for problems about luck for compatibilists at various points along a time line that begins with a stretch of time labeled long before the agent’s birth and ends with a tiny stretch of time labeled shortly before the time of action. Problems about C-luck, as I construe them, following Levy and Pérez de Calleja (see n. 7), are located in this tiny stretch of time.

Question 6 Is there something fishy about my “deciding vote” metaphor?Footnote 17

Yes. A PST is one causal factor among many, as a voter in a normal presidential election is one voter among many. A PST casts the “deciding vote” only in the sense that the collection of privileged factors leaves the outcome unsettled and the PST plays a role in determining the outcome. But bear in mind that, according to the problem of global close luck, this scale tipping happens in the case of each and every candidate for a directly free action in deterministic worlds. This claim about global C-luck gives PSTs a special importance.

Question 7 In offering an argument from present luck for the falsity of libertarianism in Mele (2006) without considering a comparable problem about C-luck for compatibilism, was I being unfair to libertarians?

First, as I mentioned, I have never offered an argument for the falsity of libertarianism. Instead, I posed a problem about luck for libertarianism, and I offered a solution. The solution was not a response to an argument. To return to the analogy with the problem of evil, my proposed solution is an analogue of a theodicy as opposed to an analogue of a rebuttal of an argument from evil for the nonexistence of God. Second, as the preceding discussion has made plain, it is unclear what the comparable problem of C-luck is supposed to be. (This is not to say that there is no problem of C-luck for compatibilists.) Third, in Mele (2006), I discussed what I took to be serious problems for compatibilists.

Question 8 Does my presentation of the problem of present luck include the claim—or something that entails the claim—that agents in some deterministic worlds have more direct control over what they do than any agent has in any indeterministic world?

No. Possibly, a claim I have made about indirect control has been confused with that claim (see Mele 2006, pp. 62–63). My presentation of the problem of present luck does not deal in comparative points about compatibilism and libertarianism regarding control. It takes direct aim at a standard libertarian position in its own right.

Question 9 Does the notion of contrastive explanation play a significant role in my presentation of the problem of present luck?

No. In Mele (2006), I provide the following gloss on “luck” in a discussion of agent causation:

if the question why an agent exercised his agent-causal power at t in deciding to A rather than exercising it at t in any of the alternative ways he does in other possible worlds with the same past and laws of nature is, in principle, unanswerable – unanswerable because there is no fact or truth to be reported in a correct answer, not because of any limitations in those to whom the question is asked or in their audience – and his exercising it at t in so deciding has an effect on how his life goes, I count that as luck for the agent. If luck is not the best short label for this sort of thing, I am open to correction. (p. 70)

This might have led some readers to believe that the notion of contrastive explanation somehow plays a significant role in my presentation of the problem of present luck. However, as I explain in Mele (2006), no debatable claim about contrastive explanation “needs to be made for the purposes of posing the problem of present luck” (p. 73). I might add that explanation—contrastive or otherwise—does not enter even indirectly into my presentation of the same problem under another name in Mele (2013).Footnote 18

Question 10 How would some of my reasoning in this article about PSTs pan out on various different conceptions of peripheral scale tippers?

Perhaps many interesting conceptions of PSTs—and of the collection of privileged factors in terms of which they are defined—are possible. I leave it to philosophers who wish to articulate an analogue for compatibilists of the problem of present luck to say what conception of these things is most useful for this purpose, in their opinion. When that happens I can revisit my reasoning about PSTs and see how it fares when these conceptions are in play.

Parting Remarks

I have no doubt that luck can figure significantly in some thought experiments that pose problems for compatibilists. (Bear in mind that some problems are soluble—including the problem of present luck, if I am right.) But this article’s topic has not been the connection between free will and luck in general. Instead, it has been similarities and differences between the problem for typical libertarians that I dubbed the problem of present luck and what purports to be an analogous problem for compatibilists.

The problem of global C-luck is analogous to the problem of present luck in some ways. The latter concerns luck at the time of action and the former concerns luck very close to that time. The problem of present luck, as I explained, is a perfectly general problem about directly free actions, as typical libertarians conceive of them; and the problem of global C-luck aspires to be a perfectly general problem about directly free actions, as compatibilists conceive of them.

There are interesting differences, however, between the two problems. Here, for purposes of emphasis, I summarize one important difference discussed above. What makes the problem of present luck a perfectly general—or global—problem about directly free actions for typical libertarians, as I explained, is that it targets an alleged necessary condition for directly free action. Obviously, if something is in fact a necessary condition for directly free action, it is satisfied in every case of directly free action. The worry posed by the problem of present luck is that an agent’s satisfying the condition at issue regarding his A-ing precludes his A-ing being directly free. If libertarians were to prove that this condition is indeed necessary for directly free action and their critics were to prove that satisfying the condition is incompatible with performing a directly free action, it would thereby be proved that libertarianism is false. Why? Because one defining tenet of libertarianism is that there are directly free actions (though “directly” is superfluous for libertarians who do not countenance indirectly free actions). But, as far as I have been able to ascertain, there is no comparable route to a perfectly general problem of C-luck for compatibilists. Motivating the problem of global C-luck requires considerably more creativity than motivating the problem of present luck does.

Pérez de Calleja has taken steps toward generating a serious global worry about C-luck that resembles the problem of present luck. When further details are filled in—including details about what belongs in the collection of privileged factors and what does not—the problem will become clearer and may generate increased interest. It may also motivate some amendments to sets of sufficient conditions for directly free action that compatibilists have defended.Footnote 19