Indeterministic Choice and Ability

Abstract

The problem of luck is advanced and defended against libertarian theories of responsibility-enabling ability. An outline of an account of ability is articulated to explore some features of the sort of ability moral responsibility requires. The account vindicates the luck objection and suggests a novel puzzle: Libertarianism is structurally barred from answering the problem of luck because responsibility requires, but inherently lacks, an explanation from reason states to actions that preserves reliability of connection between responsibility-grounding reasons-sensitivity and action.

1 Introduction

Determinism is the thesis that there is at any instant exactly one physically possible future (van Inwagen 1983, 3). Libertarians affirm that determinism is incompatible with free action or moral responsibility, and some people sometimes perform free actions for which they are morally responsible. An indirectly free action is a free action whose freedom derives from the freedom of other actions to which it is suitably related. A directly free action is a free action that is not indirectly free. Typically, libertarians insist on the following.

AP Your action A, which you perform at time t, is directly free only if there is another possible world with the same pre-t past and laws in which, at t, you do otherwise than perform A.

This strong could-have-done-otherwise condition largely underpins the problem of present or cross-world luck for libertarians, a topic of heated debate in the pertinent literature.¹ Roughly, something is a matter of luck if it is beyond your control. Capsulized, the luck problem or objection is that indeterministically produced choices or overt action—action that essentially involves peripheral bodily movement—are too luck-infused to be free or for their agents to be morally responsible for them. Dissenters, however, rejoin that the luck objection misses the mark altogether or it can be overcome. We wish to contribute to this dispute by focusing primarily on ability. Performing free actions for which you are either praiseworthy or blameworthy presupposes your having robust ability—you must be able to bring about appropriately such action. Our strategy for assessing whether the luck problem is, indeed, problematic unfolds in two steps. We proffer a partial, but sufficiently informative, analysis of robust ability; then, we trace its implications for whether agents in specified indeterministic contexts of choice have the requisite ability.

The paper divides into three main parts. In Sect. 2, we summarize the luck objection. In Sect. 3 we advance elements of a worlds-theoretic account of robust ability. In Sect. 4, we show that this partial account validates the view that agents whose choices or overt actions are aptly indeterministically caused lack the robust ability to bring about free actions for which they can be morally responsible.

2 Modest Libertarianism and the Problem of Libertarian Luck

Luck is seemingly a problem for event-causal—or modest—libertarians as well as agent-causal libertarians.² We confine discussion to a summary of the problem as it concerns the former.

Modest libertarian accounts make no appeal to Cartesian minds and avoid agent causation to explain free action.³ The sense of ‘free’ here is the personal one that true judgments of moral responsibility, such as Anna was morally blameworthy for lying, presuppose. In discussions of free action, it is customary to distinguish between causal chains associated with action from deviant motion-producing chains.⁴ The libertarian accounts of interest dictate that a free action be performed for reasons, and its being so performed consists, partially, in the agent’s having apt reasons nondeviantly and indeterministically causing it.

Modest libertarian views allow that an indirectly free action may be determined by its immediate causal precursors. They differ from compatibilist ones in that they imply that even the immediate causal antecedents of a directly free action do not determine that action. Given these antecedents, and the natural laws, there is some chance that that action not occur.⁵

Conceptualize modest libertarianism as adding to the best compatibilist theory of free action—the “host” theory—the requirement that an agent’s germane reason states indeterministically cause her directly free actions. The resulting libertarianism specifies that an agent’s freedom-level control—the sort of control free decision requires—consists in the following.⁶ Apt agent-involving events, such as the agent’s pertinent beliefs and desires, nondeviantly cause the decision. The factors that constitute an agent’s freedom-level control in making a free decision are about the very ones shared by such a libertarian view and its compatibilist host: deliberative processes with appropriate causal histories causing nondeviantly the decision. Label this the thesis that control is causal.

The problem of libertarian luck can be filled out in several different ways. For our purposes, it is sufficient to motivate this problem by sketching an example, followed by sample summary rationales of why the agent’s choice, or something that essentially involves the agent’s choice, is seemingly a matter of luck.

Imagine that in the actual world, w_a, Peg has reasons, at time t, to decide to A at some later time, t*, and competing reasons to decide, at t, to C at t*. Pursuant to deliberation, she forms the all-things-considered judgment that it is best for her to decide to A, and she continently decides at t to A at t*.⁷ Assume that Peg’s reasons to A nondeviantly and indeterministically cause this decision. To introduce a term of art, the causal trajectory, or a segment of such a trajectory, of an agent’s decision is smooth provided: it is free of responsibility-undermining factors, such as, for instance, the impact of pervasive manipulation; the agent does not succumb to akratic or other irrational influences in making the decision; and, in the absence of new information, further deliberation, or reconsideration, she decides in accordance with such a judgment, barring unusual circumstances (such as the occurrence of events over which she lacks any control and which would prevent her from deciding consistently with her best judgment). Assume that the segment of the causal trajectory that commences, roughly, with Peg’s deliberations about whether to A and extends to her making the decision at t to A in w_a is smooth. Assume, further, that Peg exercises self-control in deciding to A, and she indeterministically decides at t to A. Assume, finally, that there is an apt reasons explanation of Peg’s deciding at t to A in w_a: her reason states nondeviantly cause her decision. It is vital that there be such a causal explanation because modest libertarians accept the control is causal thesis: they agree that freedom-level control is necessary for responsibility, and such control largely consists in one’s reason states appropriately causing one’s actions.

Since Peg at t indeterministically decides to A in w_a, there is a contrast world, w_c, with the same natural laws as w_a, and is past-wise indiscernible from w_a, right up to t, in which Peg decides at t to do otherwise than A at t*. Suppose, instead, that she decides at t to C at t*.⁸

It’s plausible to judge that Peg’s deciding, at t, to C at t* in contrast world w_c, or the cross-world difference between w_a and w_c, is a matter of responsibility-undermining luck. One may try to support this judgment by appealing, for instance, to any of the following. (This list is not meant to be exhaustive).

• (1) Since there is no appropriate causal connection between Peg’s reason states and her deciding at t to C in w_c, the thesis that control is causal is violated in this world. So if Peg does decide, at t, to C in w_c, her so deciding is not in her control—it’s a matter of luck.⁹

• (2) With an indeterministic agent, such as Peg, it is false that the agent has some further power to influence causally which of her alternatives she realizes, a power over and above the mere chance of acting differently, and a power over and above the power to exercise proximal control in making whatever decision she makes.¹⁰

• (3) If one agent does one thing and another refrains from doing that thing, “and there is nothing about the agents’ powers, capacities, states of mind, moral character and the like that explains this difference in outcome, then the difference really just is a matter of luck” (Mele 1999: 280).¹¹

Here, we take no stance on whether any of these rationales can be persuasively developed and defended. However, whatever their differences, these summary defenses of the luck objection share a common kernel: modest libertarianism fails to secure the reliability of the connection between your reasons and indeterministically produced action. This is significant because it is success from reliable ability that excludes luck. This important point may be better appreciated by reflecting on a prominent event-causal libertarian’s—Robert Kane’s—perspicuous remarks on the type of control moral responsibility requires. Kane calls choices or actions by which we may form or reform our existing wills (our characters, motives, or purposes) self-forming actions (SFAs). Regarding SFAs, which appear to be paradigmatic examples of directly free actions, Kane writes:

[SFAs] must be undetermined by the agent’s pre-existing will and the agents must have what I call plural voluntary control (PVC) over them. That is, agents must have the power to voluntarily and purposefully perform them and the power to voluntarily and purposefully do otherwise (where “voluntarily” here means that actions are not coerced or compelled and “purposefully” that they are not done merely by accident or mistake, inadvertently or unintentionally)…. It would not suffice, for example—if these actions are to be “will-setting” and not already “will-settled”—that the agents could voluntarily and purposefully perform them, but could only do otherwise by accident or mistake, inadvertently, involuntarily or unintentionally. If that were the case, agents could never form or reform their own wills, for they would always be acting from a will already formed and set one way. And having the power to form and reform one’s own will is a precondition on my view for having freedom of will of the kind required for moral responsibility in a genuine libertarian sense, rather than merely freedom of action. (Kane 2013: 61, notes omitted)

Imagine that a putatively free libertarian agent, Peg, does A at t in w_a (the actual world) and B at t in w_c (some contrast world). But whenever she does B in w_c she does so by accident or mistake, inadvertently, involuntarily or intentionally. Then she does B unreliably. It seems not to be up to her that she does B, in the “up-to-oneness” sense, that free action and moral responsibility requires; in all these instances her ability to do B is not reliable.

3 A Worlds-Theoretic Account of Ability

We said previously that if you are to make free choices for which you are morally responsible, you must have the requisite performative ability. In this section, we introduce some essential features of this sort of ability. In the next section, we discuss how these features bear on the luck objection.

Beginning with four ability distinctions, first, some abilities are merely dispositional, such as your Newtonian dispositions equally and oppositely to react when acted upon, or your creature dispositions of respiration and metabolism. Other abilities are performative, outputting performances when inputted reasons.¹² Performative abilities related to action and intention are conative, and those related to knowledge and belief are cognitive.

Second, some abilities are general, others specific.¹³ You have the general ability to ride a bike insofar as you have the intrinsic powers—balance, motor control, skills—to ride. Your ability is increasingly specific to the extent that you also have the agreeable state—the physiological and psychological capacities—to ride.¹⁴ Your ability is more specific still to the extent that you additionally have the opportunity—access to a bike and window for use—to ride. Roughly, you have the specific ability to ride a bike if, and only if, you have the intrinsic powers, agreeable state, and opportunity to ride.

Third, some abilities are amateurish (or unreliable, or unskillful), others competent (or reliable, or skillful).¹⁵ Amateurish abilities exhibit a low propensity for manifestation, perhaps only a single instance, across fairly uniform circumstances. Your ability to make a free throw is amateurish to the extent that you would miss the shot and require easy shooting conditions to make the shot. Competent abilities exhibit a high propensity for manifestation across difficulty-variant circumstances. Michael Jordan’s ability to make a free throw is competent because he would typically make the shot, even when the odds are stacked against him.

Fourth, some abilities are nonagentive, others agentive. Nonagentive abilities cast you in the role of subject, such as your nonagential abilities to be seen by others, be tricked by pranksters, or be the subject of grammatical sentences. Agentive abilities cast you as agent or predicate things of you. Thinly agentive ability implies causal responsibility. You manifest the thin agentive ability to break the glass when you accidentally drop it, causing it to shatter. Gusts of wind have the very same thin agentive ability to break the glass, since it is true of the gust that it knocked the glass over, causing it to break. Thickly agentive ability implies intentional responsibility: causal responsibility because, or expressing, intentionality. Intentionally smashing the glass manifests your thick agentive ability to break it.

To clarify, the thickness of agentive ability is independent of its competence. Conative ability is competent just in case your intention would produce action; and thickly agentive just in case your action would be a product of intention. Conative ability is thickly agentive without competence if you cannot but help but don’t often correspondingly act; competent without thick agentiveness if you cannot but act but don’t often correspondingly intend. An addict’s ability to resist objects of addiction is thickly agentive but not competent, and her ability to succumb to objects of addiction is competent but not thickly agentive (assuming she is an unwilling addict; otherwise, it’s thickly agentive). Ideally, conative performances are both competent and thickly agentive: you would act all and only when you so intend.

An ability is robust if, and only if, it is performative, specific, competent, and thickly agentive. Explicitly, Peg’s ability to mow Carl’s lawn is robust if, and only if, she has the intrinsic powers, agreeable state, opportunities, and reasons such that she can intend to mow Carl’s lawn out of apt sensitivity to her reasons to mow Carl’s lawn, and there is a reliable and nondeviant relation between Peg’s intending to mow Carl’s lawn (given her reasons), when she so intends, and her actually mowing Carl’s lawn. Simply, she has all the here-and-now makings to intend nondefectively to mow Carl’s lawn and she would mow Carl’s lawn if, and because, she so intended.

With these distinctions in mind, we begin our analysis of robust ability with the following basic thought. If you have an ability to do something, then there is a world accessible to you in which you do it; you can make this world actual. To build on this guiding idea, apt constraints must be placed on the accessible worlds to capture different kinds and degrees of ability and avoid counterexamples. The sort of ability in question and variety of normative assessment it grounds will partly dictate these constraints.

With moral responsibility, one primary consideration is that the ‘can’ of responsibility regarding directly free actions expresses specific as opposed to general ability, or in the spectrum from general to specific ability, an ability much closer to the specific end.

A second consideration is that your specific ability is reliable. If you don’t know the combination of an ordinary safe and, amazingly, you dial the correct sequence on your first attempt, there is a world accessible to you—which you have made actual—at which you open the safe. While you do have the specific ability to open the safe because you’ve the requisite capacities, motor skills, conceptual acquaintance with safes, and so on, to open it, your ignorance about the correct combination means that you don’t have the reliable ability. You don’t have the reliable ability because you would too easily fail; in this case, your success is too luck-laden to qualify as the sort of ability that obligation requires.

A third consideration is that if you are directly responsible for doing something, your ability to do it is specific and reliable, and you must also intentionally be able to do it.¹⁶

Other constraints, besides the ones listed, are germane to the ability requirements of responsibility. But we have enough to adumbrate central features of a framework to capture these requirements:

Ability/Responsibility You can_robust (at t) do A (at t*) if, and only if, you have access, at t, to worlds with the same laws (as the actual world) in which (1) you have the skills, know-how, psychological, and physical capacities to do this thing; (2) you have the opportunity to do A (roughly, the right environmental conditions prevail); (3) you successfully do A in a relatively large number of these worlds, perhaps even worlds where it is increasingly difficult for you to do A, and (4) you do A intentionally.

What We Can Do (WWCD) is our preferred theory of robust ability. It explicates the clauses in Ability/Responsibility.¹⁷

Starting informally and with accessibility, if you have the ability, at t, to do A, at t*, there is a world accessible to you, at t, in which you do A at t*; at t, you can make this world actual. With ability, the accessibility relation is one indexed to an agent that obtains between one world and another. Intuitively, it represents the total suite of powers invested in the agent. If world w* is accessible from world w for you, then w* represents a “narrative continuation” of your overall suite of powers at w—a story about how your powers grow, shift, adapt, or shrink. As your powers change, so too do the worlds accessible to you. If a world is inaccessible to you, then you are neither the agent nor the subject of anything that occurs in that world. If, for instance, you can shuffle a deck of cards, there are relevant accessible worlds at which you shuffle. By actually shuffling, you make actual some such deck-shuffling world. If your hands are amputated, and you thereby forever lose the ability to shuffle, then you lose access to all relevant worlds at which you shuffle a deck.

Imagine that you can now ride a bike now. You have the specific, reliable, and intentional abilities to ride a bike—briefly, you have the robust ability to ride a bike. We comment on each of these elements in turn.

General/specific abilities In every world in which you occur, you have these powers. Unlike the virtuoso he now is, 6-month old Elton John does not have the general ability, let alone the specific ability, to play the piano. As we will write, he neither can_general nor can_specific play the piano. However, there is a very general ability notion, capability or powers, such that baby Elton can_capability play the piano because he has all the internal makings to play the piano; he lacks any innate handicaps that make his playing piano impossible for him. Being normal humans, both baby and adult Elton cannot_capability shoot heat beams from their eyes. The Kriptonian baby Clark Kent, who will grow up to become Superman, can_capability shoot heat beams from his eyes, though he cannot_general (and therefore cannot_specific) now do so.

WWCD analyzes general ability in terms of accessibility to worlds that share various things with the home world—from now on, the actual world or w_a—and differ in certain, pertinent respects from this world. “Now” facts at a time are, loosely, facts at this time about your capacities, skills, know-how, and your overall, general physiological and psychological state. “Here” facts at a time are, roughly, facts about the wider environment at this time. Suppose you know how to ride a bike, you’re awake and sober, you’re wearing a red shirt, the wind’s blowing your curly locks, the road conditions are good, but there’s no bike available. With general ability the relevant accessible worlds are worlds with the same now-facts (and the laws) as w_a but where here-facts might vary considerably. To clarify: an accessible world is relevant_now only if it agrees with all the now-facts (of w_a) but allows here-facts to vary. So, some worlds at which there is a bike present, you’re bald, and you’re wearing a black shirt are relevant_now (namely those at which all the capacities, skills, know-how, and general state are the same). At t, you have the general ability to ride a bike if, and only, if, there’s a relevant_now world accessible to you, at t, in which you ride a bike.

Again, like general ability, the WWCD analysis of specific ability invokes germane accessible worlds. An accessible world is relevant_here-and-now only if it agrees on all here- and now-facts. Regarding the bike example, with respect to now-facts, it is a world where your capacities, skills, know-how, and general state are intact; and, with respect to the here-facts, it is a world where no overriding ability-inhibiting obstacles prevail, but there is no opportunity to ride a bike (because there’s no bike). So, all worlds at which you’re asleep, you’re intoxicated, or there’s a bike are irrelevant_here-and-now. At t, you have the specific ability to ride a bike if, and only, if, there’s a relevant_here-and-now world accessible to you, at t, in which you ride a bike. In the example, at t, you don’t have the specific ability to ride a bike because, at this time, there is no relevant_here-and-now world accessible to you in which you ride a bike.

On WWCD, the generality/specificity spectrum represents how far out into the modal space (relative to the home world) we must go to locate any worlds at which the relevant powers are manifested. All things being equal, the nearer the worlds at which the powers are manifested, the more specific the ability; the further away the worlds at which the powers are manifested, the more general the ability. Maximally specific abilities quantify over only maximally similar worlds, whereas maximally general abilities quantify over much larger swaths of varyingly dissimilar worlds.

Unreliable/reliable abilities Competent abilities exhibit comparatively higher rates of success quantified over large, high-difficulty swaths of modal space. Overall, the higher the propensity for success measured across a growing number of variably upwardly-difficult worlds, the more competent the ability. In contrast, amateurish abilities exhibit relatively low rates of success quantified over small, low-difficulty swaths of modal space. In the bike example, there is no bike present but your general ability to ride is competent. Accessible to you are manifold worlds, with the same now-facts as w_a, in which you ride the bike. Among these worlds are upwardly-difficult worlds—worlds with increasingly challenging terrain, or with lots of traffic, and so on—in which you successfully ride a bike. If you do have a bike present, accessible to you are numerous worlds, with the same here- and now-facts as w_a,—even upwardly difficult worlds—in which you ride a bike. We propose: at t, you have the specific and reliable ability to ride a bike if, and only, if, there are relevant_here-and-now worlds accessible to you, at t, in which you ride a bike. Relevance here is a function of rates of success across numerous worlds including those with variably upward difficulty.

Agentiveness of abilities Your specific ability to ride a bike is thickly agentive just in case you can intentionally ride a bike given the pertinent here- and now-facts. At t, you have the specific and thickly agentive ability to ride a bike if, and only, if, there are relevant_here-and-now worlds accessible to you, at t, in which you intentionally ride a bike.

Revisiting a previous example, you don’t know the combination of a safe, still, on the very first go you input the correct combination, and it unlocks. You have the specific—and so the general—ability to open the safe but your specific ability is not reliable. WWCD accurately predicts these results. Accessible to you is a here- and now-world in which you enter the right numbers. So you have the specific and general ability to open the safe. But it’s not true that accessible to you is a whole swath of upwardly difficult worlds in which you enter the correct combination; hence, your ability is not reliable.

Bearing in mind this informal picture of ability, we now introduce some technical machinery. A WWCD frame is comprised of a domain of possible worlds, a relevance device that outputs subsets of the domain, and a nonempty set of relations—most crucially the ability accessibility relation. From these elements, we offer the WWCD analysis of dispositional ability:

WWCD/Ability You can (at time t) be the agent/subject of a thing (at time t*) if, and only if, you have access (at t) to relevant worlds at which you are the agent/subject of that thing (at t*).

The analysis captures the exceptionally broad notion of dispositional ability, as it covers nonperformative abilities to be acted upon by laws of nature, win lotteries, or be the subject of grammatical English sentences. Before proceeding to performativity, we once again unpack elements of WWCD, but a bit more formally.

The relevance device sorts the domain into subsets per relevance criteria we call a parameter. Relevance is a categorical notion defined by appeal to the set facts that constitute a parameter. A world is relevant if, and only if, it totally satisfies the parameter; irrelevant otherwise. A world might be relevant under one parameterization, and irrelevant under another. For example, a world can be relevant given a here-and-now-parameter, which is comprised of all the prevailing here- and now-facts, but may not be relevant given simply a now-parameter. Roughly, parameters in the WWCD framework play the role of dividing accessible worlds into those that are relevant and those that aren’t when it comes to analyzing various abilities, such as general or specific ability. With respect to modeling abilities, parameters have the result of holding constant, and allowing variance in, for instance, the here- and now-facts in worlds accessible from the home world.

Performative ability is a richer ability notion than dispositional ability. A performance is an output of a set of reasons-responsive powers given a set of reasons as inputs. Performativity is performing-for-reasons. Hence:

WWCD/Performativity You can (at t) perform-A-for-reasons (at t*) if, and only if, you have access (at t) to relevant worlds at which you perform-A-for-reasons (at t*).

We use lettered variables ‘A’, ‘C’,…, ‘Z’, throughout to designate the propositional content of performance; that is, what the performance effects. So, the locution ‘perform-A-for-reasons’ might designate your bringing it about that your hand is raised (for reasons r). For simplicity, we omit parenthetical indexing to reasons throughout.

We said before that WWCD captures the ability distinctions. Here’s a quick look how. First, dispositional and performative abilities are distinguished by different analyses such that performative abilities are conditionalized on reasons, dispositional abilities not. Second, we postulate that the generality/specificity spectrum should be understood by appeal to how well-specified the modal space over which the abilities are exercised is. We have introduced as examples the notions of relevance_now and relevance_here-and-now. Roughly, then, general abilities are defined over relevant_now accessible worlds; and, specific abilities defined over relevant_here-and-now worlds. A defined ability can be made more or less general, or more or less specific, by including or excluding various now- or here-facts in its parameter. Third, the amateurish/competence spectrum is homologously handled, for we posit that the spectrum should be understood by appeal to rates of success measured across a swath of difficulty-salient worlds. Roughly, comparatively low rates of success over relevant_easy accessible worlds define amateurish abilities; and, comparatively high rates of success over relevant_challenging worlds define competent abilities. In a word, assessing an ability’s reliability involves surveying the modal space of upwardly-difficult worlds and determining how often the agent succeeds in those worlds. Finally, appeal to certain conditional relationships capture the nonagentive/agentive distinction. Nonagentive abilities are not conditional upon efforts of the agent, and so tend to have a logical form of pure possibility (or, in some cases, raw dispositional ability). By contrast, agentive abilities, especially thick agentive abilities, have conditional dependencies upon reasons or modalities of effort. Thick agentive ability, for instance, is attributed to the extent that action is conditionally dependent upon intention. The fewer instances upon which action is conditionally dependent upon intention, the less thickly agentive it is.

Overall, WWCD offers the following picture of robust ability. Your ability to perform-A-for-reasons is robust if, and only if, you successfully perform-A-for-reasons in here-and-now worlds, as well as across diverse difficulty-salient now-but-not-here worlds, because you successfully intend-to-perform-A-for-reasons in such worlds. The conditionalization of success upon reasons ensures performativity. The appeal to here-and-now worlds ensures specificity. The appeal to successful performance across diverse difficult-salient worlds ensures competence. The appeal to successful intention-to-perform across worlds where you perform ensures thick agency. Capsulized, an ability is robust when it is specific, reliable, and intentional.

4 Robust Ability and Responsible Choice

We assume that moral responsibility requires robust ability because it is seemingly the ability picked out by paradigmatic moral choices. Suppose you’re listening to public radio, and the fact that it’s the end of the annual pledge drive grabs your attention, making morally salient to you the choice of either calling to donate a meager sum for the benefit of the public service or ignoring the ad and not. Seemingly, your abilities to call and donate, or to not, are both robust: performative because you’re aptly sensitive to reasons; specific because you have everything here-and-now to realize either option; competent because you would easily do what you intended; thickly agentive because whatever you do would be the direct product of what you intended. The robustness of ability here typifies the nature of the choices upon which moral responsibility supervenes. Therefore, it is a serious problem for event-causal libertarianism if it should fail to vindicate dual robust abilities to perform-, and do otherwise than perform-, A-for-reasons. We now argue it does so fail.

Imagine Peg, who is deliberating on whether she’ll mow Carl’s lawn. Libertarians wish to say that she is morally responsible for whatever it is that she chooses, and does. If so, she must have the requisite (dual) robust abilities. (Kane, recall, insists that a decision for which you are responsible satisfies the plurality conditions.) Suppose Peg has a slew of salient pro and con reasons, which we will call her total reasons r. She weighs and considers r, eventually coming to some decision at t. Suppose that, in w_a, Peg decides at t to mow Carl’s lawn for total reasons r. On the libertarian picture, if Peg is to be morally responsible for her choice to mow Carl’s lawn, there must be contrast worlds, such as w_c, at which Peg decides at t not to mow Carl’s lawn for the very same total reasons r. Hence, if Peg is to be morally responsible for mowing Carl’s lawn, then she has the robust abilities to decide to mow, and refrain from mowing, Carl’s lawn on the basis of r.

Assume Peg’s ability to decide to mow Carl’s lawn, given r, is robust. After all, she does so decide for total reasons r in w_a. To ensure competence, we postulate that she has access to a range of difficulty-relevant worlds at which she faithfully decides to mow Carl’s lawn for r. Perhaps, for instance, in some of these worlds Peg’s allergies are acting up, or she’s ill, or a marathon of her favorite show is airing, and so forth, which make it more difficult for her to fulfill her promise of mowing Carl’s lawn, but she faithfully decides out of apt sensitivity to her total reasons r. So, across wide swaths of modal space where Peg has r, she, for those reasons, decides to mow Carl’s lawn. Similarly, to ensure thick agency, we postulate that Peg’s sensitivity to r properly explains her successfully deciding to mow Carl’s lawn, ranging across both w_a and the nonactual difficulty-relevant worlds.

If Peg’s ability to decide to mow Carl’s lawn, given her total reasons r, is robust, then her ability to decide otherwise, given r, can’t be (and, hence, Kane’s plurality conditions aren’t satisfied). Peg’s deciding to mow Carl’s lawn is a contradictory of Peg’s deciding otherwise. Since she has, and acts from, total reasons r, we hold fixed the worlds at which she has, and acts from, exactly r. Since Peg’s ability to decide to mow Carl’s lawn for total reasons r is robust, it is competent, and therefore she so decides for r across wide swaths of modal space, and thereby fails to decide otherwise for r across the very same modal space. If, for instance, Peg’s ability to decide at t for total reasons r is 90% competent, then her ability to decide otherwise at t, given r, is at most 10% competent. In other words, the competence of Peg’s ability to decide to mow Carl’s lawn for total reasons r is inversely related to her ability to decide otherwise for r: every world that counts in favor of the robustness of one ability counts against the robustness of the other. Consequently, Peg’s ability to decide otherwise than mow Carl’s lawn for her total reasons r is not reliable and, hence, not robust.

Generalizing, the competence of libertarian ability to perform-A-for-reasons is inversely related to the competence of libertarian ability to do otherwise than perform-A-for-reasons, given the same reasons. The more competent (or reliable) the ability to perform-A-for-reasons, the less competent (or reliable) the ability to do otherwise for those reasons, and vice versa.

5 Conclusion

The problem of luck, in its various guises, problematizes modest libertarianism by highlighting how its conception of freedom lacks sufficient robustness to ground moral responsibility. What we want, and what modest libertarianism fails to deliver, is an explanation from your reasons states to your actions that preserves the reliability of the connection between reasons-sensitivity and action, for it is success from competent ability that precludes luck. Libertarian ability to do otherwise is success from unreliable—not competent—ability, which is merely success because of luck. Hence, libertarian ability is inimical to responsibility-grounding ability and, therefore, precludes moral responsibility.