Direct control

Abstract

This article’s aim is to shed light on direct control, especially as it pertains to free will. I sketch two ways of conceiving of such control. Both sketches extend to decision making. Issues addressed include the problem of present luck and the relationship between direct control and complete control.

Direct control is a prominent notion in recent philosophical work on free will. Randolph Clarke writes: “Direct active control is exercised in acting, not before” (2003, p. 166). Timothy O’Connor reports that “exerting active power is intrinsically a direct exercise of control over one’s own behavior” (2000, p. 61). And Robert Kane asserts that agents exercise direct control over some of their choices (1996, pp. 143–44). In these cases, Kane says, the agent’s exercise of control is not “antecedent” to the choice; rather, it occurs “then and there,” when and where the choice is made (p. 144). All three references are to books on free will.

Direct control is sometimes mentioned in connection with a certain worry about chance or luck. For example, O’Connor refers to “a chancy element to choice that cannot be attributed to the person” in a representative event-causal libertarian view, and he deems “the kind of control that is exercised… too weak to ground [the agent’s] responsibility for which of the causal possibilities is realized” (2000, p. 40). O’Connor contends that typical event-causal libertarian views have the following upshot: “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” (p. xiii; see p. 29). He looks to agent causation for a solution to the problem.¹

Surprisingly little has been said about what direct control is. My aim in this article is to shed some light on the notion, especially as it pertains to free will. I should make it clear that I do not assume that exercising direct control in or when acting is sufficient for acting freely. To be in a position to assess that proposition one needs to have a grip on what direct control is.

1 Some background

I mentioned event-causal libertarianism and agent causation in my introduction. Libertarianism (about free will) is the conjunction of two theses:

• 1. The incompatibility thesis: Free will is incompatible with determinism.

• 2. The pro-free-will thesis: There are actions that are or involve exercises of free will—free actions, for short.

Different libertarians take different positions on what free will requires beyond the falsity of determinism. Agent-causal libertarians contend that only beings with agent-causal powers can have free will. Noncausal libertarians argue that only uncaused actions can be (directly) free.² Event-causal libertarians avoid appealing to agent causation, and they typically claim that directly free actions—free actions that do not inherit their freedom from the freedom of other actions performed by the agent—are indeterministically caused by their proximal causes.

Among the most dogged critics of event-causal libertarianism are philosophers who contend that only beings with agent-causal powers can have free will (see Clarke 2003; O’Connor 2000; Pereboom 2001, 2014). Agent causation is causation of an effect by an agent, as opposed to causation of an effect by states or events of any kind, including an agent’s motivational and representational states. Agent causation is not reducible to causation by events or states. Most agent causationists prefer their agent causation straight (Chisholm 1966; O’Connor 2000; Taylor 1966), but it may be mixed with event causation in a theory about the production of free actions (Clarke 2003).

Some background on action individuation will help forestall confusion. Consider the following from Donald Davidson: “I flip the switch, turn on the light, and illuminate the room. Unbeknownst to me I also alert a prowler to the fact that I am home” (1980, p. 4). How many actions has the agent, Don, performed?

Davidson’s coarse-grained answer is one action “of which four descriptions have been given” (p. 4; see Anscombe 1963). A fine-grained alternative treats A and B as different actions if, in performing them, the agent exemplifies different act-properties (Goldman 1970). According to this view, Don performs at least four actions, since the act-properties at issue are distinct. For example, the property of flipping a switch is distinct from the property of turning on a light, and the property of turning on a light (in a room) is distinct from the property of illuminating a room. One may flip a switch without turning on a light and vice versa. Similarly, one may turn on a light in a room without illuminating the room (the light may be painted black) and illuminate a room without turning on a light (by setting a dark room on fire). Another alternative—a componential one—represents Don’s illuminating the room as an action having various components, including (but not limited to) his moving his arm (an action), his flipping the switch (an action), and the light’s going on (Ginet 1990; Thalberg 1977; Thomson 1977). Where proponents of the coarse-grained and fine-grained theories find, respectively, a single action under different descriptions and a collection of intimately related actions, advocates of the various componential views locate a “larger” action having “smaller” actions among its parts.

When actions are individuated in Davidson’s coarse-grained way, it is actions under descriptions that are performed intentionally or otherwise—not actions period. Similarly, it is actions under descriptions that are performed for reason R (or for a reason at all). For example, under the description “flips the switch,” what Don does is intentional; but under the description “alerts the prowler,” what he does is not intentional. And under the description “flips the switch,” Don might have acted for a reason having to do with getting sufficient light for reading; but under the description “alerts the prowler,” Don does not act for a reason at all (Davidson 1980, p. 5). If free actions are possible, Don may act freely under the description “flips the switch” even if he does not act freely under the description “alerts the prowler” (Mele 2010).

Fine-grained and componential theorists make no special appeal to action-descriptions. Theorists of both kinds can straightforwardly say, for example, that Don intentionally flipped the switch and unintentionally (or nonintentionally) alerted the prowler.³ They can also say that Don freely flipped the switch in order to get enough light for reading and that, although he alerted the prowler, he did not do that for a reason and did not do it freely.

I am neutral regarding the three theories of action individuation I sketched. Henceforth, readers should understand the action variable A as a variable for actions themselves (construed componentially or otherwise) or actions under descriptions, depending on their preferred theory of action individuation. The same goes for the expressions that take the place of A in concrete examples. For example, fans of coarse-grained individuation should read “Don flips the switch” as “something Don does under the description ‘flips the switch’” and people who reject this theory should make no adjustments. This convention will enable me to avoid having to make a variety of points in two different ways.

2 Direct control: a beginning

John Bishop sets the stage for a brief but instructive discussion of direct control by sketching “a very bad argument” (1989, p. 70) that he attributes to Peter van Inwagen. The argument runs as follows:

Consider a mechanism that on the push of a button flashes either a red light (with 50 percent probability) or a green (with 50 percent probability). Then an agent who pushes the button has no control over which light flashes. Now if the mental states that constitute an agent’s reasons caused matching behavior only probabilistically, by parity of reasoning, the agent would have no control over whether that matching behavior was produced or not. Hence, actions cannot be behavior with merely probabilistic mental causes. (Bishop, p. 70)

Bishop objects that the argument draws “a false analogy” between a light’s flashing and an action (p. 70). Something true of a probabilistically caused nonaction might not be true of a probabilistically caused action. And although the agent only “indirectly controls” whether the light flashes or not, his pressing the button—an action—is “an exercise of direct control” (p. 71). In this way, Bishop contends, at least some actions (for example, John’s pressing a button) differ from outcomes of an action (for example, a light’s flashing). In Bishop’s view, whenever agents act, they exercise control (pp. 23, 25)—and, more specifically, direct control. Clarke agrees: “In every instance of action, the agent exercises some degree of direct active control” (2003, p. 76). Bishop’s event-causal theory of action is meant to accommodate this idea.

As Bishop understands direct control, it is nothing out of the ordinary. That is reassuring. But I, at least, am left with some questions. Consider the following two assertions: Joe exercised direct control; Joe’s pressing the button was an exercise of direct control. They have the ring of incompleteness. “Over what?” one wants to ask. If Joe raised his right arm (in an ordinary way) to vote for a motion at a meeting, we might say that he exercised direct control over his arm—or over how his arm moved, or over the motions of his arm. If he tied a rope around his right arm and then raised his right arm by raising the rope with his left arm, did he exercise direct control over his left arm and indirect control over his right arm? Did he exercise direct or indirect control over the rope? If Joe tied a rope around a log and then pulled the log to a woodpile, did he exercise direct control over the log? Did he exercise direct control over the rope and indirect control over the log? Did whatever he exercised direct control over extend no further than his body?

All these questions and many more would be answered by a full account of direct control. To the best of my knowledge, nothing approaching a full account of it exists, and I do not try to provide one here. Although the project of developing a full account would be interesting, I believe that many of the details would not have a special bearing on free will. Even so, I do not know how to proceed in a discussion of the bearing of direct control on free will without offering readers more guidance on what direct control might be than I have encountered in my own reading.

Suppose, following Bishop and Clarke, that whenever agents act intentionally, they exercise direct control over something or other. In an ordinary case of raising one’s right hand over one’s head, what does the agent exercise direct control over? One seemingly reasonable answer is motions of his arm and hand. What does an agent exercise direct control over when he decides (or chooses) to A? To decide to A, as I understand it, is to perform a momentary action of forming an intention to A (Mele 2003, ch. 9).⁴ When deciding to A is understood in this way, an analogous answer is his acquisition of an intention to A.

Consider an alternative answer to my question about an ordinary case of raising one’s right hand. Someone may claim that the agent exercises direct control over his raising his right hand—that action. The answer in the preceding paragraph represents the action at issue—the hand-raising—as an exercise of direct control over something. And the answer in the present paragraph represents that very action as something the agent exercises direct control over. According to one view of things, we perform our actions and whenever we act intentionally we exercise some direct control over one or more nonactions—for example, motions of our bodies or our acquisition of an intention. But according to the claim at issue now, whenever we act intentionally we exercise some direct control over some action.⁵ This latter idea, Bishop claims, is wrongly attributed to agent causationists. He asserts that a certain alleged difficulty for agent causationists dissolves once a misunderstanding of their view is exposed: “The theory is that actions consist in the causing by their agents of certain events or states of affairs. Thus, agents are not held to agent-cause their actions… but rather the events or states of affairs that are, so to say, intrinsic to their actions” (1989, p. 68). If, according to agent causationists, agents exercise direct control only over what they agent-cause, then, if Bishop’s interpretation is correct, they do not exercise direct control over their actions.

I paired the idea that when Joe raises his hand in the normal way he exercises direct control over motions of his arm and hand with the idea that when Joe decides to A he exercises direct control over his acquisition of an intention to A. Having a label for the conception of direct control reflected in these ideas will facilitate discussion. I will say that, when combined with the idea that direct control is never exercised over actions, they reflect a Bishop-style conception. The alternative idea that when Joe raises his hand he exercises direct control over that action is paired with the idea that when Joe decides to A he exercises direct control over his deciding to A—that action. These alternative ideas reflect what I dub a Knight-style conception of direct control (a name chess enthusiasts might like; Gangnam-style was an alternative I briefly considered).⁶

3 Directness, completeness, and free will

In any account of something called “exercising direct control,” the term “direct” should do important work. The following certainly seems to be a plausible requirement on an agent’s exercising direct control over X: If S exercises direct control over X, then S does not exercise control over X only by exercising control over something else (or, more precisely, something that does not include X).⁷ If this proposition about direct control is true, then if Joe exercises control over the log only by exercising control over the rope, he does not exercise direct control over the log. And if he exercises control over the rope only by exercising control over relevant bodily motions, then he does not exercise direct control over the rope. Comparable examples framed in a Knight-style way would involve an agent’s exercising control over one action only by exercising control over another action (that does not include the former action). If Joe exercises control over his moving of the log only by exercising control over his pulling of the rope, he does not exercise direct control over the former action.⁸ If asked, some proponents of a Knight-style conception of direct control might assert that agents exercise direct control only over their basic actions.⁹ Similarly, some proponents of a Bishop-style conception might assert that agents exercise direct control only over nonactions “that are, so to say, intrinsic to” their basic actions (1989, p. 68).¹⁰ (Example: The rising of one’s arm is intrinsic to one’s raising it.)

I have supposed, following Bishop and Clarke, that whenever agents act intentionally, they exercise direct control over something or other (to avoid excessive repetition, I dub this supposition S). If there are basic actions, an agent performs at least one basic action whenever he acts intentionally. In light of this, a pair of suggestions I just made may be augmented. A proponent of a Knight-style conception of direct control might assert that agents exercise direct control over all and only their basic actions and a proponent of a Bishop-style conception might assert that agents exercise direct control over all and only nonactions that are intrinsic to their basic actions. Readers should treat these “all and only” ideas as working assumptions.

How close does this get us to free will? Well, unless all basic actions are free actions, the Bishop-style and Knight-style views of direct control, as thus far developed, leave a considerable gap between direct control and free will. Of course, this is what one should have expected, given supposition S. With that supposition in place, unless it is true that whenever agents act intentionally they do something or other freely, there will be times at which agents exercise direct control and do not act freely. And there are many scenarios in which although an agent does something intentionally, he does not do anything freely. For just one kind of example among many, consider very young children who have developed the capacity for intentional action but have not developed the capacity for free action. Also, if there are possible worlds in which lots of agents frequently act intentionally but no agent has free will, then, given supposition S, there are possible worlds in which agents often exercise direct control but no agent ever acts freely. I leave it to readers to identify such worlds for themselves. An incompatibilist can pick any deterministic world in which there are beings that often act intentionally, unless he or she believes that intentional action is incompatible with determinism. People who hold that there are nonhuman animals that act intentionally but are incapable of acting freely can pick a world in which those animals are the only agents. Finding a world you regard as illustrating the point is left up to you, dear reader.

Is there a way of pursuing an exploration of direct control that gets us closer to free will? Reflection on a worry some readers may have about Bishop-style direct control may prove interesting in this connection. The worry is that direct control, so understood, leaves important facts about our control out in the cold. Don’t we have some control over whether we decide to A or decide to B, in some cases? And when we decide to B in such cases, don’t we exercise direct control over our deciding to B? These are among the questions such readers may ask.

I offer some potential answers from a Bishop-style perspective. The control we have over whether we decide to A or decide to B in typical cases of the sort at issue consists partly in our being able to decide to A for reasons that recommend A-ing and able to decide instead to B for reasons that recommend B-ing. And there is plenty of room for exercises of indirect control over whether we decide to A or decide to B. Consider a representative case of decision making that is informed by evidence-gathering and thoughtful reflection. Joe will be moving to another state soon to take a new job, and he is thinking about whether to buy a house there soon or rent for a while and buy a house later. He gathers information about houses and real estate agents, asks a real estate agent to show him around, looks at houses, gathers more information, and so on. These actions have an effect on what he eventually decides to do—and does—about his housing situation. In performing them, he exercises indirect control over what he will decide. In the end, he decides to buy house H. In performing actions of the kind mentioned, he exercises indirect control over his deciding to do that, over his doing it, over whether he decides to buy house H or decides to do something else instead, and over whether he buys that house or instead does something else.

I commented on exercising control that we have over whether we decide to A or decide to B. But my comments were about indirect control. What about direct control? On a Bishop-style view, when we decide to B, we do exercise direct control; we exercise it over our acquisition of an intention to B. And if we had decided to A, we also would have exercised direct control—in this case, over our acquisition of an intention to A. In typical cases in which an agent with the pair of abilities I mentioned continues to regard both A and B as live options at the moment of decision, a Bishop-style theorist may claim that in exercising direct control over his acquisition of the intention he acquires, he exercises direct control over whether, at the time, he acquires an intention to A or acquires an intention to B.¹¹

I suggested that a proponent of a Bishop-style conception of direct control might assert that agents exercise direct control over all and only nonactions that are intrinsic to their basic actions. The suggestion now is that some such exercises amount to exercises of direct control by an agent over whether, at the time, he acquires an intention to A or acquires an intention to B.

Recall that I am not assuming that exercising direct control in or when acting is sufficient for acting freely. An incompatibilist may observe that, according to traditional compatibilists, an agent’s simultaneously possessing the pair of abilities I have highlighted—the ability to decide to A for reasons that recommend A-ing and the ability to decide instead to B for reasons that recommend B-ing—is possible in a deterministic world. And, of course, any incompatibilist claims that there are no free decisions in such worlds. This is not the place for an assessment of the merits of compatibilism. So, to push forward, I assume that the pair of abilities at issue are what I dub O-abilities. An agent who did not do X at t was O-able at the pertinent time to X at t only if in a possible world with the same past up to t and the same laws of nature he X-s at t. Such worlds are indeterministic.

How close are we now to free will, on a libertarian conception of it? The expression “complete control” sometimes crops up in discussions of free will (see Mele 2014, 2015). Whatever exactly complete control is supposed to be, direct control should not be confused with it. Exercising direct control over X does not entail exercising complete control over X on any Bishop-style or Knight-style conception of direct control according to which, whenever agents act intentionally, they exercise direct control over something.

I illustrate my claim about the relationship between direct control and complete control with a pair of examples. Both are loosely based on a real-world experiment. There are simpler illustrations, but I am fond of the following pair (see Mele 2014, p. 554).

Sol’s instructions as a subject in a neuroscience experiment are to fully depress either the Q key on a computer keyboard with his left index finger or the P key with his right index finger (and never to press both at the same time). He is told that which key he presses is up to him. He will make over forty key presses—either Q or P, sometimes one and sometimes the other—in the course of an hour, and he is asked to refrain from planning in advance which key to press. Before he presses he will hear a tone that he is instructed to treat as a “decide” signal. When he hears the signal, he is to decide right then which key to press and then press it straightaway. Fully depressing a key requires that the key moves all the way down and makes contact with the switch under it. The keyboard Sol is using has a randomizer on it—a genuinely indeterministic one—that ensures that there is always a small chance that a key he is trying to press will stick and fail to make contact with the switch under it. So Sol never has complete control over whether he fully depresses the key he has selected and never has complete control over whether he fully presses the Q key or the P key.

Just now, Sol fully depressed the P key, as he was trying to do. On some accounts of basic action, his doing so is a basic action. And proponents of Bishop-style and Knight-style views who regard it as such (as I am interpreting such views) will say that Sol exercised direct control over the key’s contacting the switch (Bishop-style) or over his fully depressing the key (Knight-style). That Sol obviously does not exercise complete control over these things is no problem at all for their claims.

A theorist who is more restrictive about basic actions may say that Sol’s basic action in this case is his moving a finger in a certain way or his trying to depress the key fully.¹² Theorists who take this position will say that Sol exercised direct control over certain bodily motions (Bishop-style) or over the finger-moving action or the attempt (Knight-style).

Consider a variant of Sol’s story in which a genuinely indeterministic randomizer has been installed in Sol’s brain that ensures that there is always a small chance that his proximal intentions to press a key—his intentions to do so right then—will not be followed by a corresponding attempt nor by any relevant bodily motions. Just now, Sol fully depressed the Q key, as intended; and, of course, he lacked complete control over whether he would try to press that key on this occasion and over whether he would move his left index finger. He also lacked complete control over whether he would try to press the Q key or the P key and over whether he would move his left index finger or his right index finger. But these facts are entirely consistent with the truth of the claims at issue about what he exercised direct control over.

I am not claiming that the existence of directly free actions depends on exercises of complete control (over something or other). However, people who think that there is this dependence and who associate complete control with direct control should be aware of the points just made.

4 Deciding

Might it be that when we turn to deciding in particular we find a special connection between direct control and complete control? Might direct control over whether at t one acquires an intention to A or instead acquires an intention to B amount to complete control over that? Might something analogous be true if a Knight-style view is in play?

Consider the following story (from Mele 2006, pp. 73–74). Bob lives in a town in which people make many strange bets, including bets on whether the opening coin toss for football games will occur on time. After Bob agreed to toss a coin at noon to start a high school football game, Carl, a notorious gambler, offered him $50 to wait until 12:02 to toss it. Bob was uncertain about what to do, and he was still struggling with his dilemma as noon approached. Although he was tempted by the $50, he also had moral qualms about helping Carl cheat people out of their money. He judged it best on the whole to do what he agreed to do. Even so, at noon, he decided to toss the coin at 12:02 and to pretend to be searching for it in his pockets in the meantime (decided to C, for short).

Call the world in which my story is set W. In some other possible worlds with the same past up to noon and the same laws of nature, Bob does not decide to C. In some such worlds, Bob decides at noon to toss the coin straightaway, as promised. In some others, he is still thinking at noon about what to do. Readers are free to speculate about other noontime possibilities that are compatible with W’s past up to noon and its laws of nature.

Assume now that Bob lacks agent-causal powers. Then, in O’Connor’s view (to return to some claims quoted earlier), “there are objective probabilities corresponding to each of [Bob’s possible choices], but within those fixed parameters, which choice occurs… seems, as far as the agent’s direct control goes, a matter of chance” (2000, p. xiii), and “the kind of control” Bob exercises “is too weak to ground his responsibility for which of the causal possibilities is realized” (2000, p. 40). In O’Connor’s view, then, even if Bob exercises Bishop-style direct control over his acquisition of the intention he acquires at the time, he does not decide freely. It is safe to say that O’Connor does not view Bob as exercising complete control over whether, at the time, he acquires an intention to cheat or instead acquires an intention not to cheat.

The root of Bob’s problem, as O’Connor sees it, is that he lacks agent-causal powers. So, given that he lacks such powers, even if Bob were to exercise Knight-style direct control when he makes his decision, the alleged problem would remain. O’Connor’s worry is not about Bishop-style direct control in particular; it is about event-causal direct control.

I return to O’Connor’s worry shortly. Event-causal Knight-style direct control requires some attention now. If we were to ask a fan of Bishop-style direct control in what an agent’s exercise of direct control over his acquisition of an intention to A consists, the fan would have a straightforward answer. The exercise of direct-control consists in the agent’s deciding to A. In deciding to A, the agent forms an intention to A. This agent’s acquiring an intention to A is related to his deciding to A as the rising of his arm is related to his raising it. Return to the Knight-style idea that when an agent decides to A he exercises direct control over his deciding to A—that action. In what does this exercise consist, according to an event-causal proponent of this idea? Presumably, not some distinct action over which the agent exercises some control. For then the control the agent exercises over his deciding to A would be indirect: he would be exercising control over his deciding to A only by exercising control over some distinct action. Perhaps, then, our Knight-style theorist will claim that the agent’s exercising direct control over his deciding to A just is his deciding to A. If the answer is supposed to apply to all possible instances of deciding to act, it can be stated as follows: Necessarily, an agent’s deciding to A is identical with his exercising direct control over his deciding to A (notice that if exercising direct control over one’s deciding to A is understood in this way, it suffices for freely deciding to A only if all possible decisions are free). Someone who finds this answer attractive may take a parallel position on basic action: necessarily, any basic action, A, is identical with the agent’s exercising direct control over A.

I mentioned a Bishop-style idea about typical cases in which an agent who at t is O-able to decide then to A for reasons that recommend A-ing and O-able to decide instead then to B for reasons that recommend B-ing and who continues to regard both A and B as live options at the moment of decision. The idea is that in exercising direct control over his acquisition of the intention he acquires, this agent exercises direct control over whether, at the time, he acquires an intention to A or acquires an intention to B. A counterpart Knight-style idea is that, under the conditions specified, in exercising direct control over the decision he makes at the time, the agent exercises direct control over whether, at the time, he decides to A or decides to B. On the Knight-style view now under consideration, not only is an agent’s deciding to A identical with his exercising direct control over his deciding to A, but, in the conditions specified, his deciding to A also amounts to an exercise of direct control over whether, at the time, he decides to A or decides to B.

Here O’Connor would again object that within the parameters fixed by the objective probabilities corresponding to each of Bob’s options, “which choice occurs on a given occasion seems, as far as the agent’s direct control goes, a matter of chance” (2000, p. xiii). An agent’s exercise of event-causal Knight-style direct control over whether, at t, he decides to A or decides to B (as just now characterized) is not enough for O’Connor.

Because Knight-style direct control over whether one decides to A or decides to B might sound more impressive than it actually is, I will develop O’Connor’s worry—or, rather, a very similar one that I have called “the problem of present luck” (Mele 2006, p. 66)—a bit further. I have devoted a lot of ink to making the problem salient (Mele 2006, pp. 5–9, ch. 3, ch. 5). The next several paragraphs are a brief statement of the problem.¹³

Return to the story of Bob and the coin. Someone may assert that, in this story, certain relevant worlds—specifically, a world in which Bob decides at noon to cheat and a world with the same past up to noon and the same laws of nature in which he decides at noon to toss the coin right then, as promised—diverge as they do at noon because, in these worlds, it is up to Bob what he does at noon and he acts differently at noon in these worlds. But one may have one’s doubts about the extent to which this is up to Bob.

Consider a fuller version of Bob’s story in which although—right up to noon—Bob does his very best to talk himself into doing the right thing and to bring it about that he does not succumb to temptation, he decides at noon to cheat. In another possible world with the same past up to noon and the same laws of nature, Bob’s best was good enough: he decides at noon to toss the coin straightaway. That things can turn out so differently at noon (morally or evaluatively speaking) despite the fact that the worlds share the same past up to noon and the same laws of nature will suggest to some readers that Bob lacks sufficient control over whether he makes the bad decision or does something else instead to make that decision freely and to be morally responsible for the decision he actually makes (again, it is the direct versions of free action and moral responsibility that are at issue). After all, in doing his best, Bob did the best he could do to maximize the probability that he would decide to do the right thing, and, even so, he decided to cheat—precisely the opposite of what he decides at noon in another world that does not differ at all from this one before noon. One may worry that what Bob decides is not sufficiently up to him for Bob to be directly morally responsible for making the decision he makes and for it to be a directly free decision.

Given the details of Bob’s story, how can Bob have enough control over whether he decides to cheat or does something else instead at noon for his decision to be directly free and for him to be directly morally responsible for it? This is an instance of the central question posed by what I have called “the problem of present luck” (2006, p. 66). (In Mele 2006, ch. 5, I offer event-causal libertarians a solution to the problem.)

It may be claimed 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. Or consider the claim that possessing the power of agent causation is required for having free will. If agent causation is impossible, and the alleged necessary condition is true, then free will is impossible.¹⁴

An event-causal Knight-style theorist can offer something represented as an agent’s exercise of direct control over whether, at t, he decides to A or decides to B. But, as I have characterized this offering, it is just a matter of the agent’s making one decision or the other at t (for reasons that recommend the course of action decided upon) while regarding the alternative course of action as a live option and being O-able at t to decide at t to take it (for reasons that recommend the alternative course of action). I am not claiming that not enough control is exercised here for the agent to have decided freely. But one can see why someone—O’Connor, say—would worry about this. It may seem that the featured difference at t between a world in which Bob decides at t to cheat and a world with the same past up to t and the same laws of nature in which he decides at t to toss the coin straightaway is too much a matter of luck (or chance) for Bob to have decided freely—even if Bob exercised direct control over whether he decided to cheat or decided to toss the coin, as such an exercise is characterized in the event-causal Knight-style view I presented. (The same goes for Bob’s exercise of event-causal Bishop-style direct control over which intention he acquires at the time.)¹⁵

As I mentioned, in O’Connor’s view, Bob’s problem is that he lacks agent-causal powers. In principle, an agent causationist who takes a position on direct control can opt for either a Bishop-style or a Knight-style position. In the following passage, O’Connor sounds like a Knight-style theorist: “[Tim] had the power to choose to continue working or to choose to stop, where this is a power to cause either of these mental occurrences. That capacity was exercised at t in a particular way (in choosing to continue working), allowing us to say truthfully that Tim at time t causally determined his own choice to continue working” (2000, p. 74). Here, it seems, Tim agent-causes a mental action of his, his choosing to continue working. But it is clear that O’Connor’s considered view of direct control is a Bishop-style view (see p. 72, n. 11). What is agent-caused, in his considered view, in the case of an agent who freely decides or chooses to A, is “the coming-to-be of a state of intention” to A (p. 72, n. 11). This is what I have been referring to as the agent’s acquisition of an intention to A.

I have argued elsewhere that the postulation of agent-caused intentions (or acts of deciding, for Knight-style theorists), along with any direct control this involves, leaves the problem of present luck intact (Mele 2006, ch. 3). I will not rehash the argument here. But I will repeat and comment on a pair of passages from my discussion of the issue.

The first passage is about Tim in the remark I quoted in which O’Connor sounds like a Knight-style theorist. After supposing that Tim conducted himself exactly as O’Connor claimed he did, I wrote:

Why should we suppose that the following cross-world difference is not a matter of chance or luck: that Tim exercised the capacity at issue at t in choosing to continue working rather than in choosing to do something else, as he does in some possible worlds with the same past and laws of nature? Grant that Tim “causally determined his own choice to continue working.” Why aren’t the differences in his causal determinings at t across worlds with the same past and laws of nature a matter of chance or luck? Tim was able to causally determine each of several choices, whereas a counterpart who fits the event-causal libertarian’s picture was able to make—but not to causally determine—each of several choices. If it is a matter of chance that the latter agent chooses to keep working rather than choosing to do something else, why is it not a matter of chance that the former agent causally determines the choice he causally determines rather than causally determining a choice to do something else? (Mele 2006, pp. 54–55).

Notice that the central question here is about the agent’s causally determining (or agent-causing) one choice rather than causally determining (or agent-causing) another choice (readers should not assume that I regard myself as having a good understanding of what O’Connor means when he says that an agent “causally determined” his choice). Someone may claim that the cross-world difference between Tim’s choosing at t to continue working and his choosing at t to take a break (both worlds having the same past up to t and the same laws) is just a matter of luck, and an agent causationist may reject this claim on the grounds that, in both worlds, Tim causally determines (or agent-causes) his choice. But this just moves the debate back a step and opens the door to the central question in the passage just quoted—the question whether the difference at t between Tim’s causally determining one choice and his causally determining the other choice is just a matter of luck.

The next paragraph in Mele (2006) reads as follows:

Perhaps O’Connor is thinking that the conceptual relation between control and chance is such that the fact that Tim exercised direct control over which choice he makes answers each of these questions. Should his readers find this thought persuasive? I do not see why. Even if the fact that Tim exercised direct control in choosing to continue working is incompatible with its being just a matter of luck that he chose to continue working, this does not show that a relevant cross-world difference between his exercising direct control “in [this] particular way” (2000, p. 74) and his exercising it in choosing to do something else is not just a matter of luck. The reader should bear two points in mind. First, as I explained, it is not just a matter of luck that Tim chose to keep working even on event-causal libertarian views [i.e., even if Tim lacks agent-causal power]. Second, O’Connor does not place cross-world differences in agents’ doings out of bounds in the context of free will: in fact, such differences are featured in his objection from chance to event-causal libertarians. A third point also is worth making… O’Connor’s critique of event-causal libertarianism makes it plain that he does not believe that what Kane [an event-causal libertarian] conceives of as an exercise of direct control… solves the problem of cross-world luck at the time of action, and there is a parallel worry about exercises of direct agent-causal control (Mele 2006, p. 55).

As far as I can see, if the problem of present luck is a problem for event-causal libertarians—whether they go in for Bishop-style or Knight-style direct control—it is a problem for agent causationists as well and the direct control to which they appeal, whether that control is Bishop-style or Knight-style. One may stipulate that agent-causal power enables one to decide in a luck-precluding way in scenarios of the sort at issue, but stipulating this is uninteresting. Providing a compelling account of what it is about agent-causal power in virtue of which it enables one to do this—and of what exactly it is being claimed to enable one to do—would be another matter entirely. I have not seen such an account. Another option for an agent causationist to pursue is to explain why the presence of cross-world luck at the time of decision is compatible with directly free deciding (see Steward 2012). In Mele (2006), I pursue this option from an event-causal libertarian perspective.

5 Conclusion

My aim in this article, as I announced, has been to shed some light on exercising direct control, especially as it pertains to free will. I sketched two ways of conceiving of such exercises—what I called Bishop-style and Knight-style ways. Both sketches extend to decision making, and both honor the supposition, S, that whenever agents act intentionally, they exercise direct control over something or other. Given S, as I explained, there is a considerable gap between direct control and free will. How that gap is to be filled is, of course, a matter of great controversy. For my own way of filling it for libertarians, which includes a proposed solution to the problem of present luck, see Mele (2006, ch. 5).

An anonymous referee suggested, reasonably, that some incompatibilists might not accept S and that some “authors may unwittingly use the notion of ‘direct control’ more or less stringently in different dialectical contexts.” Philosophers who reject S either have or are seeking a conception of direct control that is more exclusive than anything that can encompass S. They have the option of understanding my sketches as having something they might label direct control lite as their subject matter. They may conceive of those sketches as needing to be augmented with something for which the rejection of S may clear the way—something that would secure an action’s being directly free. The mode of augmentation I have in mind would build on a connection I have explored between exercising direct control and performing a basic action. I have no objection to this way of proceeding. In fact, given that they reject S, I recommend it. (Philosophers who accept S may regard what these other philosophers are after as something that should be labeled direct control plus.) Philosophers who unwittingly use “direct control” to mean different things at different times may benefit from learning that they have done this, but I have no desire to call anyone out on this.

The same referee described this article as something that is “apt to serve as a touchstone for theorists in clarifying how they intend their own usage of ‘direct control’ (and ‘complete control’) to be understood.” If this article does serve that purpose, I will be pleased.