Supposing that the local pool is closed today, is it true that “Ivy is able to swim”? For one speaker, concerned with making plans for the afternoon, the fact that the pool is closed today is quite relevant, and the claim is false; Ivy is not able to swim. For another speaker, wondering whether Ivy should sign up for the swim team, and recalling Ivy’s skills and previous accomplishments, Ivy is able to swim. The context in which the sentence is uttered can affect the truth of the claim “Ivy is able to swim,” partly by affecting which features of the world we take into account when assessing its truth.

This is an example of what Ann Whittle names Contextualism about Agential Modals (CAM), according to which sentences containing agential modals – e.g., “S can A,” “S is able to A,” “S is a competent A-er” – have different truth conditions in different contexts. The book begins with an argument for CAM, and the suggestion that such claims have an invariant core – e.g., “‘S can A’ is true if and only if S’s A-ing is compossible with certain facts” (p. 8). On the version defended here, the context of the speaker, rather than that of the agent the modal is attributed to, determines how it is that we fill out parts of that core; e.g., which facts are the “certain” facts.

Whittle investigates the implications of this insight for our notions of moral responsibility (henceforth, simply “responsibility”), freedom, and control, ultimately arguing that the context-sensitivity of agential modals transfers to claims about (1) all-in abilities to do otherwise; (2) robust control, which is required for responsibility; and (3) responsibility itself.

The philosophical debate on freedom and responsibility is well-trodden terrain, and Whittle’s book does much to argue for a position that is off the beaten path. Her view is neither compatibilist nor incompatibilist, as these terms are often understood. Whether freedom and responsibility are compatible with the truth of determinism, on Whittle’s view, will depend on the speaker’s context.

The book is packed full of arguments and engagement with other thinkers, and covers a wide range of topics, including various forms of the Consequence Argument, both compatibilist and incompatibilist views of freedom, and the context-sensitivity of claims about knowledge, causation, and prevention. Yet Whittle is also careful, conscious of the limitations of the arguments presented, and aware of exactly what claims she needs to arrive at her main conclusions. And for seemingly every major claim, she provides multiple, independent arguments and thorough discussion of objections.

Though there is much to say about the project, I will restrict my remarks to summarizing its main argument and briefly commenting on points that I take to be particularly relevant to the further development of the project.

1 Regulative Freedom and All-in Abilities

The road from CAM to contextualism about responsibility is a difficult one. As Whittle suggests, one might agree that CAM holds in general, yet think it does not hold for the sort of ability required for freedom or responsibility in particular. For example, it is common for leeway theorists about responsibility – roughly, those who hold that the ability to do otherwise is required for (direct) responsibility for an action – to hold that the standards for claims about the abilities relevant to responsibility, and which types of facts are the “certain facts,” will be invariant across contexts. A leeway incompatibilist, for example, might hold that we always hold the laws and the past (up to a certain point) fixed, regardless of the speaker’s context; an agent who A-s at t has the ability to do otherwise only if there is a world with the same laws and past up to t in which they do not A. One of Whittle’s main goals in this book is to argue against invariantist views of this sort.

In the first half, Whittle focuses on a concept of freedom – regulative freedom – which requires the ability to do otherwise. This freedom is significant in some respects, she suggests, even if it is not necessary for moral responsibility. Significant for what, or in what way? Though others have argued that some form of freedom is importantly related to other things we value – e.g., creativity, or authorship of one’s life – Whittle does not attempt to answer this question. Instead, she motivates the claim to significance by appealing to the case of the willing addict who “positively endorses his drug of choice, but even if he didn’t, his addiction would compel him to take the drug in any case” and suggests that “in at least one respectable sense of ‘free’, it is false to say that he acts freely because he has no options open to him” (p. 38).

With this form of freedom in mind, Whittle begins by considering a version of the famous Consequence Argument, which, roughly put, proceeds from principles about the fixity of the past, the fixity of the laws of nature, and the transfer of necessity, to the conclusion that agents in deterministic universes lack abilities to do otherwise, and thus regulative freedom. According to CAM, there are contexts in which crucial premises come out false. But, as Whittle suggests, this would be too easy a rejection of the Consequence Argument. Again, one could hold that CAM does not apply to the sorts of ability that matter for responsibility. Which ones do matter, then? Whittle suggests that it is the notion of all-in ability (to do otherwise), where this can be understood as:

S has an all-in ability to φ if and only if S has the ability to φ and the opportunity to manifest φ in the type of circumstances present (p. 43).

The relevant sort of ability (to the right of the bi-conditional), and the relevant sort of opportunity (as well as how we cash out “type of circumstances”) need clarification. Whittle suggests that there is difficulty in elaborating on “ability,” insofar as one runs the serious risk of begging the question. Yet, an agent’s having the relevant opportunity can be understood as an agent’s not being prevented from manifesting φ. And, insofar as prevention is not exclusively an agential concept, clarifying this can avoid question-begging. Doing so will go some way to illuminating all-in abilities more generally.

Whittle proceeds to delve into debates on prevention and argues that it is a context-sensitive notion, and that all-in ability inherits this context-sensitivity. Whittle offers a final reformulation of the Consequence Argument, and considers a variation on the Manipulation Argument. These, Whittle suggests, are two of the key arguments for leeway freedom incompatibilism (p. 72). With respect to each of these arguments, Whittle argues that there are contexts in which they are sound, and others in which they are not. Importantly, the contexts in which these arguments tend to be presented are ones that make certain considerations salient, and thus they are contexts in which the premises are true. Consequently, Whittle can accommodate the intuitive appeal of arguments for the incompatibility of determinism and the ability to do otherwise (and regulative freedom) without thereby accepting an incompatibilist view.

I pause briefly to mention one potential worry concerning the account of regulative freedom. Recall that the account of regulative freedom was intended to capture the significant or respectable sort of freedom – which requires the ability to do otherwise – that the willing addict purportedly lacks. In defending the account, we get a case of someone who is faced with “your money or your life” and gives their money to the mugger (pp. 78–9). Whether it is true that the victim is free to refrain from handing over the money depends on the context, and there is at least one context in which it is true. Yet one might think that, whatever significant and respectable freedom the willing addict lacks, it will be one that this victim lacks as well. If so, then Whittle’s account of regulative freedom will not pick out the significant and respectable freedom lost in such cases; such freedom is not as sensitive to contexts as Whittle’s regulative freedom. Perhaps alternatively, though Whittle’s regulative freedom is significant and respectable in some contexts, it is not so in all contexts.

2 Robust Control and Responsibility

Even if a significant and respectable sort of leeway freedom is context-sensitive, it is a further move to suggest that the sort of freedom (or control) required for (direct) moral responsibility is also context-sensitive.

A common route to the denial of a leeway theory of such control is through consideration of Frankfurt cases. Very briefly, such cases include an agent, call him Frank, who decides to do something. Unbeknownst to Frank, a nefarious neurosurgeon is waiting in the wings ready to intervene and correct course were Frank to stray from the path leading to this decision. Yet since Frank never strays, the neurosurgeon never intervenes. Frank is, intuitively, responsible for his decision, despite the fact that the presence of the neurosurgeon (and his readiness to intervene) seems to result in Frank’s lacking the ability to decide otherwise.

So, if leeway freedom is not required for responsibility, then what sort of freedom, and what sort of control, is required? Here, Whittle distinguishes between different kinds of control. A drug addict, even an unwilling one, exercises some form of control when taking the drug; they intentionally take the drug. This is, roughly, what Whittle calls minimal control. This, however, is not the sort of control required for responsibility; it is not robust control. What is robust control?

Whittle makes a common distinction between regulative control – which requires the ability to do otherwise – and what is often called guidance control – which is often taken to require, roughly, that the agent has the capacity to recognize and respond to reasons. Frank’s deciding displays the latter, but not the former (assuming he lacks the relevant ability). However, guidance control is often explained, at least partly, in terms of other possible worlds – e.g., whether there is a relevant possible world (or a set of them) in which the agent has sufficient reason to act differently and does so. Thus, Whittle argues, one might think that such accounts are, in the end, not all that different from accounts that require the ability to do otherwise.

Whittle ultimately suggests that guidance control is best understood as an action’s being performed “under one’s own steam” (p. 141). And she argues that generally, control is best understood in terms of abilities, providing an abilities analysis of minimal and robust control.

Roughly put, an agent exercises robust control over A if and only if (a) they exercise minimal control over it, and (b) they typically have the all-in local ability to recognize pertinent reasons, and in cases of suboptimal actions (including morally wrong actions), the all-in local ability to react appropriately to those reasons (p. 144).

How “local” the relevant abilities are – that is, how many features of the agent and her circumstances we hold fixed as part of the “certain facts” (p. 34) – is sensitive to contexts, as are other features like reliability (pp. 174–5). And the “typically” in (b) is intended as a ceteris paribus clause which allows of exceptions. Which exceptions? This is a difficult question, and the answer is intertwined with much of Whittle’s discussion throughout the book. Though I cannot do justice to it here, and I would expect this to be a prime target for discussion in work engaging with the book, Whittle appeals to it in distinguishing between Frankfurt cases and cases of compulsion or addiction. All of these agents arguably lack some ability to do otherwise, but they differ on whether the features that make this true are causally active in the process (pp. 155–9).

However, though the ceteris paribus clause concerns the exercise of robust control generally, the local ability to respond differently to reasons is required in cases where the agent acts wrongly (p. 148).

Making use of all of the work throughout the book, Chap. 8 develops an argument for a contextualist analysis of moral responsibility (CMR), which states that “the truth conditions of a statement, ‘S is morally responsible for (morally wrong) action A’ can vary depending upon the context of the agent attributing moral responsibility, due to the shifting agential modals constitutive of that attribution” (p. 178).

At this point, it is worth commenting on the treatment of Manipulation Arguments, which rely on cases of manipulated agents. Such cases often involve agents in deterministic universes who, without their knowledge or consent, have been subject to value engineering leading to a substantial change in values, and these modified states issue in a particular action. Since it is intuitively plausible that (1) the victim is neither free nor responsible with respect to the action, and since (2) the no-difference thesis is true – there is no relevant difference between the victim and a standard agent in a deterministic universe – the argument concludes that freedom and responsibility are not compatible with determinism.

Such cases, or variations of them, serve various purposes: they are sometimes used to argue against compatibilism, sometimes against incompatibilist views, and sometimes against specific accounts. To play these dialectical roles, and to maintain the no-difference thesis, the victims are stipulated to have, at the time of action, the features that the target view claims to be relevant to freedom or responsibility; e.g., reasons-responsiveness, or the ability to do otherwise. Some cases involve victims who are stipulated to lack the ability to do otherwise even when one is using relatively lax standards.Footnote 1 And some involve victims who are stipulated to have the ability to do otherwise, using some of the strictest standards available, like those of the leeway incompatibilist.Footnote 2 This poses a worry for both Whittle’s account of regulative freedom, and her account of robust control.

Whittle offers a different version of the argument in relation to regulative freedom. Rather than appealing to the intuition that the victim is not free, Whittle’s first premise states that the victim is not able to do otherwise, and the no-difference thesis and conclusion are modified accordingly; determinism precludes regulative freedom. This argument, Whittle suggests, is sound in some contexts but not in others, since the truth of the first premise varies with context. But this does not seem to work with victims who have the ability to do otherwise, using some of the strictest of standards. Such victims, though intuitively not free, have all-in abilities to do otherwise across contexts. If one thinks that the freedom such agents retain is not significant, then this again suggests, at the very least, that there is more to the significant form of freedom that motivated the account of regulative freedom.

And a similar point can be made with respect to Whittle’s account of robust control, assuming it is a relatively synchronic feature. There seems to be nothing preventing us from offering a case in which the victim has robust control over his action, yet is intuitively neither free nor responsible (nor acting “under her own steam”).Footnote 3

3 Responsibility and How We Treat Each Other

The final chapter will likely be the one of most interest to readers of this journal. It is here that Whittle discusses how CMR might relate to how we treat each other – how we blame, punish, etc. – though issues concerning criminal sanctions are largely set aside. This chapter is centered around a defense of CMR from the problem of unfairness. Given CMR, “the very same agent could count as morally responsible for an action in one context of utterance, but not the other” (p. 198). This could result in unfair consequences, especially if one thinks that “if it is true to say that S is morally responsible for wrongful act A, then it is appropriate to subject them to blame” (p. 199).Footnote 4 At least this is so for expressions of blame, which can be harmful for the target (pp. 203–4).

Whittle provides multiple responses; though I’ll focus on the first main response, which involves attempting to divorce harm from moral responsibility. Whittle points to at least two ways of avoiding this unfairness, and argues for the disjunction. First, one could offer a conception on which responsibility for a wrongful action is not defined in terms of the appropriateness of blame. Second, one could offer an account of blame on which blaming does not essentially involve harming the blamed. Either way, moral responsibility for a wrongful action is not sufficient for the appropriateness of harmful blame.

I conclude with some brief comments on this move, and some further questions about the implications of the view. First, part of the argument for CMR is that it captures our responsibility practices better than alternative views. Yet, even if one grants that CMR best captures our practices of attributing responsibility, our responsibility practices plausibly include more than this; they plausibly include having reactive attitudes, expressions of praise and blame, and perhaps sanctions and punishment. If the argument is successful in divorcing responsibility from harmful blame, then one might worry the view will be unable to capture much of the rest of our responsibility practices, and CMR will not have this purported advantage over competitors.

But second, and granting that Whittle has correctly identified the source of unfairness, the attempted divorce is not clearly successful. Even if we accept that responsibility for wrongdoing is not sufficient for the appropriateness of harmful blame, it remains plausible that responsibility is necessary for it. Consequently, in cases where the further conditions for the appropriateness of harmful blame are met, the appropriateness of harmful blame can vary with context. Consider an example.

After an intervention, Pedro makes a promise to both Theresa (his therapist friend) and Bess (his best friend) that he will not smoke another cigarette. He then smokes a cigarette. Given her context, Bess speaks truly when she says “Pedro, you are responsible for smoking,” and given that he broke a promise in doing so, we can suppose that he is responsible for a wrongful action. Now further suppose that whatever further conditions are needed for the appropriateness of harmful blame, they are met; Bess can appropriately express harmful blame to Pedro, and does so. Yet given Theresa’s knowledge of Pedro’s psychological make-up, different features of Pedro are salient to her, and she speaks from a different context. Theresa truthfully says that he is not responsible for smoking. If responsibility for a wrongful action is necessary for the appropriateness of harmful blame, then regardless of whether these further conditions are met, Theresa cannot appropriately express harmful blame to Pedro (at least not from this context). The problem of unfairness arises again, insofar as Pedro is now subject to appropriate harmful blame from Bess but not from Theresa.

Of course, Whittle has multiple arguments, and even if one is not convinced that the divorce is successful, she has more to fall back on when responding to the problem of unfairness. Yet, even if these other arguments are successful, they do not avoid the conclusion that Bess can appropriately harmfully blame Pedro while Theresa cannot, and that this is due to a difference in contexts; their targets lie elsewhere. A further point of interest concerns how far this context-sensitivity spreads, given that moral responsibility and blameworthiness are often taken to be importantly connected to a variety of other aspects of our moral lives. For instance, does Pedro owe a genuine apology to Bess but not Theresa?

Another point of interest would involve developing how CMR applies in more complex scenarios. Consider a rough extension of our case. Suppose it is inappropriate to harmfully blame the blameless, and that generally, it is impermissible to do so. Bess does not flout this obligation when she harmfully blames Pedro, given his blameworthiness from her context. But from Theresa’s context, maybe Bess did flout this obligation. Could it now be appropriate for Theresa to harmfully blame Bess for doing so? Assuming she does, did Theresa now flout this obligation from Bess’s context, given that from Bess’s context, Bess does not act wrongfully, and is not blameworthy, for harmfully blaming Pedro?