While most of the literature on morality and determinism focuses on threats to moral responsibility, determinism might be thought to threaten morality on separate grounds. Haji draws on the popular principle that “ought” implies “can”, in order to show that determinism undermines deontic morality (1998, 1999, 2002, 2019). Similar arguments are presented by Lockie (2018), although Lockie, unlike Haji, does not intend to defend scepticism about obligation, but rather to show that any such scepticism is inherently self-defeating.

By “deontic morality”, Haji has in mind any moral use of the terms “ought” and “ought not”, as well as moral judgements of right and wrong. While he concedes that judgements of moral “good” and “bad” may still make sense within a deterministic framework, he argues that the action-demanding normative terms associated with obligations and prohibitions would be seriously undermined. Determinism precludes moral duty.

However, as Haji himself makes explicit, in order to reach this conclusion, we need not only an “ought” implies “can” principle, but also an “ought” implies “able not to” principle (2002, 28). A similar principle is found in Lockie’s work (2018, 181). I will argue, firstly, that even if we accept the popular “ought” implies “can” principle, there are good reasons to reject any “ought” implies “able not to” principle. Secondly, without the “ought” implies “able not to” principle, such arguments are limited to establishing a much weaker conclusion; we cannot conclude that there are no moral duties at all, only that there are no unfulfilled moral duties. Thirdly, while this weaker conclusion may look similarly problematic at first sight, from a practical perspective it actually makes very little difference to morality.

1 Determinism, Ability, and “ought” Implies “can”

The principle that “ought” implies “can” has certainly seemed compelling to many,¹ although it’s not uncontroversial.² Haji originally calls his “ought” implies “can” principle “K”, and then later “Kant’s Law/Obligation”. But for present purposes, let us simply call this sort of principle “OIC” (so as to match the broader class of principles under discussion). Haji (2002, 14) formulates his version of OIC roughly as follows:

According to this principle, an agent only ought to do something if she actually can do it, and ought only to refrain from doing something if she actually can refrain from doing it.

1.1 The Analysis of “can”

Given that there are broad variations in the way that we might interpret “can”,³ there are also variations in the way that we might interpret OIC. Haji’s (2002, 23) most moderate definition is as follows:

While this is taken to be the bare minimum required for ability, Haji adds that it may also require being motivationally able, and having the right sort of “know-how” (2002, 16–24).

Physical and psychological possibility are fairly straightforward notions. Plausibly an agent is only “able” to perform actions that are consistent with their psychological characteristics and their physical abilities. The inclusion of the stipulation that the agent must be “psychologically able” may, however, seem controversial. It means that an agent with a strong aversion, say, may count as unable to do something, even if she could succeed in doing it should she choose to. One reason we might nonetheless endorse this reading, as Haji points out, is that it is natural to suppose that an agent with a serious enough phobia might be excused for her failure to do something that her phobia prevents her from doing. For instance, we would not typically consider an agent “able” to save a drowning child if a severe phobia rendered her incapable of entering the water (Haji 2002, 22).

Moreover, endorsing a relatively strong sense of “can” may prove indispensable to the argument as a whole. That is because the argument aims to establish that the ability to do otherwise is ruled out by determinism, where this involves the very same sense of ability for which it will be true that “ought” implies “can”. Any weakening of the sense of “can” utilised in the OIC principle may risk introducing a corresponding weakening of the argument for supposing that determinism rules out the ability to do otherwise in precisely that sense. For example, Haji notes that if we supposed a merely conditional analysis of “can” would do, according to which the ability to do otherwise simply requires that the agent could do otherwise if she chose to, then this would make it dubious to suppose that determinism rules this ability out (2002, 67–68).

In fact, Haji argues that even if such conditional abilities are present, determinism robs us of the opportunity to do otherwise. If any factors, internal or external, prevent an agent from exercising some skill they have, then this will constitute a barrier to their having the opportunity to exercise it (2002, 22).⁴

Finally, the “control” requirement is supposed, at the very least, to rule out having the “ability” to do things that happen purely by fluke (Haji 2002, 22). In analysing such control, Haji cites Vihvelin, who states: “We make judgments about ability on the basis of evidence of a reliable causal correlation between someone’s attempts to do a certain kind of act and the success of her attempts.” (2000, 142). This sort of control neither entails nor is entailed by possession of the other senses of “ability”. Plausibly, an agent’s phobia may make her psychologically and motivationally unable to purchase a pet snake, but doing so may not be “strictly out of her control”; were she to try, she could reliably succeed. Similarly, if a golf novice hits a hole in one on her first attempt, this certainly shows that she is physically able to hit a hole in one, but if it is an unrepeatable fluke, then it will still be “strictly out of her control”.

1.2 Determinism and Obligations

Haji and Lockie use rather complex arguments to reach the conclusion that determinism rules out all obligations. Moreover, Lockie’s argument incorporates the additional goal of showing that any argument in favour of determinism would be self-defeating, and Haji’s argument incorporates his attempt to show that if nothing is obligatory, then nothing is right or wrong either. I am not going to address the latter part of Lockie’s argument,⁵ and I am not going to consider whether Haji is right to suppose that wrongness and rightness depend on obligation. While this claim has been contested,⁶ I am happy to grant it. Moreover, in what follows, it is the status of actions as obligatory (rather than right or wrong) that will be the prime focus. So for present purposes, we can work with a simplified version of the argument, which might go as follows:

1. If determinism is true, no agent is ever able to act otherwise than they do act. (basic premise)

2. If no agent is ever able act otherwise than they do act, then no agent ever has an obligation to act otherwise than they do act. (premise derivable from OIC)

3. If determinism is true, no agent ever has an obligation to act otherwise than they do act. (from 1 and 2, via hypothetical syllogism)

While 3 is an interesting conclusion, it is weaker than the the one that is ultimately defended by either Haji or Lockie. It does not entail that if determinism is true, there are no obligations, merely that that there are no unfulfilled obligations. It leaves open that agents sometimes both have and fulfil moral duties. In order to reach the stronger conclusion, that there are no obligations at all, Haji introduces a parallel principle, which he calls “CK” (2002, 28). Lockie (2018, 182) puts forward a similar principle. Elsewhere, Haji gives the same sort of principle different titles, such as “Kant’s Law/Impermissible” (Haji 2019, 8) or “Obligation/Alternate” (Haji and Herbert 2018a, 186). Let us simply call this whole class of principles “OIANT principles” (for “ought” implies “able not to”). Haji (2002, 28) defines the relevant sort of principle, omitting the temporal indices, as follows:

If we grant OIANT, we can also establish that there are no obligations to do what we actually do, given our inability to do otherwise. A simplified argument of this form runs as follows:

1. If determinism is true, no agent is ever able to act otherwise than they do act. (basic premise)

2. If no agent is ever able act otherwise than they do act, then no agent ever has an obligation not to act otherwise than they do act. (premise derivable from OIANT)

3. If determinism is true, no agent ever has an obligation not to act otherwise than they do act. (from 1 and 2, via hypothetical syllogism)

4. If determinism is true, no agent has an obligation to act as they actually do act. (from 3, an equivalence through double negation)

The final step from 3 to 4 is valid provided we grant that “not acting otherwise” entails “acting as one actually does”. For present purposes, “acting as one actually does” should be understood broadly, so as to be fulfilled if the agent does not act otherwise; hence it should include the agent’s inaction, if the agent in question is not actually doing anything. Granted this broad reading, it should be uncontroversial that “not acting otherwise” directly entails “acting” as one actually does. It should be similarly obvious, granted this broad reading, that premise 2 is entailed by OIANT.

The first argument shows that, given determinism, no agent has an obligation to act otherwise than they do act. The second argument shows that, given determinism, no agent has an obligation to act as they actually do either. Between the two arguments, this rules out all moral obligations.

While the first argument appears compelling, the second argument seems considerably weaker. The principle upon which it rests, OIANT, seems more dubious than the principle invoked by the first argument, OIC. If we reject the argument from OIANT to the conclusion that if determinism were true, no one would be obligated to do what they actually do, then we are left with a weaker conclusion: that if determinism were true, no one would be morally obligated to act otherwise than they do act.

2 How Plausible is OIANT?

Haji offers various lines of argument in favour of accepting OIANT: the first is a simple appeal to symmetry between OIC and OIANT. Lockie’s work also draws on the intuition that there ought to be symmetry between such principles. However, even if we doubt that there is any obvious inherent reason to suppose that the two principles are symmetrical, we might argue that we ought to accept such symmetry on the basis that both principles are taken to be motivated primarily by a two-way freedom requirement (this seems to be the supposed basis of the symmetry for Haji). Haji also offers a “theory-fuelled” argument, which appeals to a particular analysis of obligation. I will argue that OIANT is, at least on the face of it, inherently implausible before going on to deal with each of these arguments in turn.

2.1 The Prima Facie Implausibility of OIANT

It has already been noted that psychological ability is crucially included in the definition of “able to” invoked in Haji’s OIC and OIANT principles. In light of this, however, “ought” implies “able not to” has some undesirable implications. Many actions that seem obviously morally prohibited are also psychological impossibilities for most psychiatrically well-adjusted individuals. For instance, my psychology is such that I could not take a chainsaw and use it to saw off the arms of a small child. To be clear, I don’t mean a child that has gangrene, say, and needs those limbs removed urgently on pain of death, but a perfectly healthy child; one whose limbs I have no reason to remove. In fact, I could not do such a thing even if I were offered reasons, if they were of the wrong sort: e.g., I could not saw off the arms of a child for a monetary incentive (even if I were offered a very reasonable market rate). Does this entail that it is not morally obligatory for me to refrain from sawing off the arms of small children?

This conclusion seems counterintuitive. It is the fact that such an action would be morally reprehensible which may well, in this case, explain both my irresistible aversion to it and my reasons for supposing that it is morally obligatory that one refrains from such behaviour.

Unlike Haji, I think it is plausible to suppose that my inability to do such a thing entails that I cannot be held responsible for not doing it, and hence deserve no praise.⁷ The moral expectation that I refrain from dismembering small children is a very easy standard for me to meet. It seems close to the bare minimum you might reasonably expect of me, so I hardly deserve a medal. But it seems one thing to say that I don’t deserve praise, and quite another to say that sawing off the arms of small children would not be morally impermissible. We are usually quite happy to talk about being psychologically compelled to do things that we also have a duty to do. We might even suppose that it is the very fact that something is perceived as morally prohibited that (at least sometimes) explains an agent’s psychological aversion to doing it.

The principle that “ought” entails “able not to” surely seems dubious. We ought to accept it only if we are offered very compelling arguments.

2.2 The Defence from Apparent Symmetry

The first argument appeals to the apparent symmetry between “ought” implies “can” principles and “ought not” implies “can” principles (along with, presumably, the latter’s complement stipulation, that “ought” implies “able not to”). Haji argues “that it is difficult to see why control requirements of deontic obligatoriness would differ, in this respect from control requirements of deontic wrongness” (2002, 29). He interprets OIC as postulating an alternative possibilities condition as a control requirement for obligatory actions, and supposes that similar considerations would count in favour of accepting an alternative possibilities condition on prohibited ones.

Even “ought” implies “can” is controversial, but it has a strong history of philosophical support behind it and it seems highly intuitive. “ought” implies “able not to”, in contrast, has nothing like the same standing. As Nelkin notes, the principle is not usually seen as axiomatic, and the alleged symmetry that Haji sees between these sorts of principle is hardly obvious (2011, 102).

In fact, I think there is a plausible basis for “ought” implies “can” that simply has no parallel in the case of “ought” implies “able not to”. The appeal of “ought” implies “can” principles may in fact not rest on any control requirement that involves alternative possibilities. More plausibly, their appeal may be grounded in the simple idea that it is unreasonable to demand the impossible. We may well suppose that it is unreasonable to demand the impossible without supposing that this rests on a control requirement that involves alternative possibilities.

Any demand that is impossible to meet will, by an obvious logical entailment, also be a demand with respect to which the agent lacks two-way control. But there is no entailment in the other direction. There is certainly no logical entailment from the plausible idea that it is unreasonable to demand the impossible to the far less plausible claim that it is unreasonable to demand the unavoidable.⁸

If there are cases in which we are plausibly required to do something that we also cannot refrain from doing, then we have good reason to suppose that it is the unreasonableness of the demand to do the impossible that is doing all of the work in rendering principles like OIC plausible, and that two-way control is irrelevant. Of course, we have already examined such a case: the case of morally abhorrent actions that an agent is also psychologically incapable of.

Moreover, think about cases in which it is uncertain whether or not one is physically capable of committing some wrong. For example, I think that it would be morally impermissible for me to leave the house with a kitchen knife and stab to death the first person I see. However, I have absolutely no idea whether I could physically succeed in such an endeavour, even supposing I tried my best. It seems absurd to suppose that I should first have to be in a position to know whether I could succeed in order to work out whether stabbing an innocent bystander is morally impermissible (appeal to some theory of normative ethics ought to settle that question quite irrespective of my abilities).

There is also a clear a disparity here with respect to duty and prohibition. Plausibly, I can only be morally required to save the drowning child if I am capable of it. If it is uncertain whether I will be physically able to, then we might plausibly say that I have a duty to try, even if I could not have a duty to succeed in my attempt. In contrast, it barely seems coherent to assert that it would be impermissible for me to try to stab someone to death while asserting at the same time that it would not be impermissible for me to actually stab someone to death. For one thing, I could hardly succeed in such an attempt without first making the attempt, so if the latter is prohibited, it seems the former must be too. Moreover, it seems that the very reason we are prohibited from attempting certain things is precisely because it would be wrong to actually do those things, so a stand-alone prohibition against attempting would typically make very little sense unless coupled with a prohibition against actually doing what one is attempting to do.

Moreover, there are obvious reasons why we might expect such an asymmetry. In general, having a duty to do something might be thought to depend on our having strong moral reasons to do it. One would expect moral reasons to behave in ways that parallel reasons of any other sort, such as, for instance, epistemic or prudential ones. And reasons of every other sort seem to be asymmetric with respect to our abilities in precisely the way that I claim moral reasons are. Perhaps it cannot be true that an agent ought to believe something if she is incapable of believing it. But it does not seem to follow that she could not have good reason to believe something that she is incapable of doubting (if a belief is indubitable, this is typically thought to be a point in its favour). Or consider prudential reasons. If you are starving hungry (barring any conflicting considerations) you have good reason to eat. If you are incapable of eating, this would undermine those reasons. But it’s not at all obvious that if you cannot resist eating, that would in any way weaken the reasons you have in favour of eating.⁹

There are clear grounds for supposing that our reasons are limited to those things that we are able to do, while not being similarly limited to what we are able to avoid. Our reasons are typically based on some sort of independent value that’s at stake. If a reason for performing some action or believing some proposition is based on some value (e.g., good evidence or a strong moral or prudential case), then insofar as we are capable of sensitivity to that value, we will be sensitive to the reasons it generates. But there would be no point at all in possessing a parallel capacity for insensitivity towards those same values. Here’s another way to put the point: if we are violating some core value, we had better have a good excuse for doing so. Being incapable of respecting the value certainly is a good excuse. If we are instead respecting the value, we need no excuse for doing so, so no parallel ability to do otherwise is called for in order to render our behaviour intelligible. That something is impossible is, in itself, a reason for not bothering. In contrast, the fact that we cannot avoid choosing to do something doesn’t undermine the rationale for doing it at all. In some cases, it may well be the very strength of the rationale in favour of performing some action or adopting some belief that explains why doing so might be irresistible to us.

Demanding the impossible is unreasonable on the basis that an inability to do something may render one’s otherwise bad or irrational behaviour perfectly reasonable in the circumstances. This is not dependent on any alternative possibilities requirement for control, as evidenced by the fact that a person’s perfectly decent but unavoidable behaviour may well be entirely reasonable and explicable, even if they cannot resist this behaviour, on the basis that it is explained by their sensitivity to certain values. Such an explanation would more plausibly be weakened by introducing the additional ability to be insensitive to those values as opposed to being strengthened by it.

Moreover, whether a demand constitutes a demand for the impossible is asymmetric with respect to what the agent must do and what the agent cannot do. While it is unreasonably demanding to expect an agent to do the impossible, it is in no way similarly unreasonably demanding to expect an agent to do the inevitable. Since the requirement is so easily met, quite the opposite seems to be true; the inevitability is, if anything, evidence for the conclusion that such a requirement is undemanding. But in any case, there is certainly no parallel entailment of demandingness. This is precisely why psychiatrically well-adjusted individuals don’t deserve medals for not dismembering small children.

We cannot support OIANT then, by a simple appeal to the alleged symmetry with OIC. Moreover, it is not all all obvious that the insistence on symmetry can be propped up with the consideration that both OIC and OIANT depend on a two-way freedom.

2.3 The “Theory-Fuelled” Defence

The “theory-fuelled” defence draws on Feldman’s analysis of obligation in terms of the comparative value of the possible worlds accessible to agents (1986). More recently, Haji calls this the “doing the best we can” model (DBWC) (2019; see also Haji and Herbert 2018a).

In short, the analysis contends that we are morally obligated to actualise the best world that we can actualise of all of those “accessible” to us, where “best” is understood in terms of a ranking of the “deontic” or “intrinsic” value of worlds, according to whichever theory of normative ethics is endorsed (e.g., for a utilitarian it may be the world with the greatest sum of utility, for a Kantian it may the world in which we act in accordance with universalisable maxims, whereas for a virtue ethicist it may be the world in which we best act in accordance with the virtues).

There needn’t be a unique best world; perhaps various worlds are tied for first place. But we are obligated to actualise a best world. However, some facts may be “unalterable”; there are certain states of affairs that would occur in every possible world accessible to us (e.g., the sun will rise tomorrow, various statements about the past will be true, etc.) If those states of affairs occur in all of the worlds that are accessible to us, then it is trivially true that they will also occur in all of the best worlds accessible to us. But now we have a problem: it appears that anything unalterable will automatically be obligatory. We will automatically be obligated to actualise any world that we cannot avoid actualising. Yet this is counterintuitive; it seems intuitively wrong to say that I have a moral duty to actualise a world in which the sun rises tomorrow or to actualise a world in which certain statements about the past are true.

Haji’s solution is to appeal to an OIANT principle. That is, we assume that further to supposing that we can only be obligated to bring about states of affairs that are accessible to us, we must also suppose that we can only be obligated to bring about any particular state of affairs on the explicit condition we are also able to actualise a world in which those states of affairs do not obtain.

Perhaps this is one way to maintain a DBWC theory consistent with ensuring that the unalterable should not automatically be obligatory. But it is not the only way, and it’s hardly obvious that it is the most plausible way. For instance, instead of endorsing OIANT, we could instead add the (far more compelling) stipulation that we can only be obligated to bring about any outcome insofar as that outcome is causally dependent on our intentions.¹⁰

In fact, Haji’s claim that the relevant sort of ability for duty requires that actions not be “strictly out of one’s control” commits to precisely this. More recently, Haji and Herbert have defended the claim that the sort of ability relevant to duty ought to be robust, in the sense that requires, among other things, that it is strongly agentive, where this involves being brought about by an agent intentionally (2018a, 2018b). However, if having a duty requires that we are able to fulfil that duty in precisely this robust sense, this already rules out having the duty to bring about some unalterable states of affairs; it rules out precisely having the obligation to bring about states of affairs that will occur independently of our intentions, and hence rules out having such obligations as seeing to it that the sun rises tomorrow.

Moreover, this plausibly explains why it seems intuitively obvious that we are obligated to refrain from dismembering small children, even if not refraining from such behaviour is a psychological impossibility, consistent with the fact that it does not seem plausible that we are obligated to see to it that the sun rises tomorrow. Since the very point of moral duties is to guide our intentions, we should expect those duties to be limited in scope to those outcomes that are dependent on our intentional behaviour.

Short of having some independent reason to favour a solution that requires us to invoke OIANT over the principle that duties are limited to intention-dependent states of affairs, it seems we ought to favour the latter. While OIANT principles seem inherently problematic, the principle that one cannot be obligated to bring about a state of affairs that will happen independently of one’s intentions seems like a basic truism. Given the ready availability of this solution, a state of affairs being unalterable need not make it automatically obligatory (even if we explicitly reject OIANT). Importantly, however, the fact that some state of affairs is unalterable doesn’t rule out our having an obligation to bring it about either.

Haji and Herbert further note that if we explicitly presume that if something is unalterable, then it cannot be obligatory, this would also provide a basis from which to argue in favour of OIANT principles (2018a, 188). But I am arguing precisely that we have no good independent reason to accept such a presumption. The fact that I am not robustly capable of committing certain morally heinous acts may well establish that my avoidance of such acts is unalterable. But the point is precisely that we have no good reason to suppose that this is inconsistent with it being obligatory that I refrain from committing those acts. So while the presumption that unalterability rules out obligatoriness could certainly provide a basis for accepting an OIANT principle (via a fairly obvious entailment), such a presumption is itself no more plausible than the OIANT principles it is invoked to establish and is no less in need of independent justification.

In sum then, it seems that we have no reason to accept OIANT. Recall, however, that OIANT was a crucial component of the argument to the conclusion that determinism entails that nobody ought morally to do anything. Without it, we are entitled only to the weaker claim that, given determinism, no one ought morally to act otherwise than they do. We must now assess whether, from a practical perspective, this weaker conclusion turns out to be just as destructive.¹¹

The following section assesses the implications of embracing just the weaker conclusion entailed by determinism and OIC, given a rejection of OIANT. In particular, the aim is to question whether this weaker conclusion alone should be regarded as destructive to deontic morality, even if we follow Haji in supposing that no one has a duty to do otherwise.¹²

3 The Lack of Obligation to Act Otherwise

The conclusion that nobody is obligated to act otherwise than they actually do may seem problematic enough. Let us call this claim “Unfulfilled Obligation Scepticism” (UOS):

This means that only our actual choices and actions could possibly count as obligatory. We may sometimes both have and fulfil moral obligations, but we can never have a moral obligation that we contravene. Perhaps this alone undermines deontic morality. UOS may seem to threaten moral deliberation, obligation, or motivation, rendering them practically unintelligible. Let’s examine these potential threats in turn.

3.1 UOS and Moral Deliberation

Firstly, it might be argued that UOS renders moral deliberation practically impossible. By “moral deliberation”, I mean reasoning about what to do in advance of deciding, rather than reasoning about how to appraise an action that has already occurred.

There are several reasons why UOS might look problematic. We always know in advance that there is no way that our actions will possibly count as “forbidden” at the time that we perform them. Moreover, whether we are obligated to perform any action seems closely dependent on whether we choose to, so we might suppose that UOS robs us of any intelligible way to give rational weight to our purported duties prior to actually making a choice.

Suppose that Ada is a highly rational moral agent, who has recently become convinced of the truth of UOS. She believes that she can only be morally obligated to do something if she does in fact do it. She now faces the following situation: Ada’s uncle has arranged in his will for her to receive all of his fortune should he die. However, he is planning to change his will when he visits the solicitor’s office later today. Her uncle has two small children and had previously supposed that his wealthy wife’s ample income would stand them in good stead should he suddenly die, so he had planned to leave his fortune to Ada, his favourite niece. However, his wife has just died in a freak accident (leaving her fortune to her husband). If he should suddenly die too, his children would now be left orphaned and destitute, while Ada would receive all of his wealth, including that of his late wife. In contrast, Ada has a decent job and a reasonably high income of her own. She will be fine without a substantial inheritance. He is therefore planning to change his will, leaving the bulk of his fortune to his children and a much more modest sum for Ada. She can appreciate the reasonableness of her uncle’s decision.

However, while she is alone visiting him, he collapses unconscious, and appears to be dying of a heart attack. No one else knows that Ada is visiting. She could easily walk away without calling an ambulance. She would then be rich enough to buy that Ferrari she always wanted. As a rational moral agent, Ada certainly would have supposed that she had a moral obligation to call an ambulance prior to being persuaded of the truth of UOS. But she must now work out what bearing this principle has. Should it change the way that she morally deliberates?

I endorse the idea that we ought to do the best we can, where this involves being obligated to bring about the best of the intention-dependent states of affairs accessible to us. So Ada ought to actualise the best intention-dependent state of affairs she can. This only seems to require two abilities: firstly, she must be able to compare the deontic value of the worlds that would result from various rival intentions, and secondly, she needs to suppose that she can actualise the best of them. We ought to ask whether UOS poses any obstacle to her doing either of these things.

Firstly, let’s think about her ability to assess the value of the intention-dependent states of affairs between which she is deliberating. On virtually any theory of normative ethics, the world in which she calls an ambulance will look superior to the world in which she does not call an ambulance. If she doesn’t call an ambulance, she will perform no morally admirable actions, and her greed and cruelty will result in an innocent man dying, and his children being left orphaned and destitute. If she does call an ambulance, she will have done a good deed, and through her fairness and kindness, she would ensure that he survives to care for his children. For deontologists, virtue ethicists and consequentialists alike then, the world in which she calls an ambulance will be ranked morally superior to those in which she refrains.

Does she need assurance of her duty in advance? It seems not. On any plausible DBWC analysis, moral duties are not going to be stand-alone considerations that exert their moral pressure on us independently of the other facts about the situation. A world \(w\) that we might actualise does not count as morally superior to some other world \(w'\) on the basis that we are morally obligated to actualise \(w\) instead of \(w'\) (that supposition would render the DBWC account entirely vacuous). The explanation is always the other way around: we are morally obligated to favour actualising \(w\) over \(w'\) precisely because we have some independent basis to suppose that \(w\) is superior to \(w'\). The obligation arises because one of these worlds has a higher “intrinsic value”. Values are conceptually prior to obligations: duties are the conceptual outputs of values.

But the point needn’t rest on accepting a DBWC analysis either. Quite independently of whether one accepts that analysis, it is a mistake to think that duty is conceptually prior to moral value. Consider Kant. There can be few theorists who afford duty a more fundamental status. Yet even for Kant, duties are not independent additional substantive reasons for acting; they are derived from considerations about the rational wills of other agents, which confer on them a status as ends in themselves. While Kant encourages us to act “from duty”, as opposed to merely “in conformity with duty” (1998, 10–11), he certainly doesn’t suppose that duties exist and exert pressure independently of the values that give rise to them; respecting duty is simply the same thing as respecting other rational beings. It’s hard to imagine any plausible system of ethics according to which duties are not derived from some prior moral value.

Perhaps it will be accepted that Ada (as a rational agent with some theory of normative ethics up her sleeve) knows that the world in which she calls an ambulance for her uncle is better than the world in which she refrains from calling an ambulance (i.e., she knows that there are substantive moral considerations in favour of calling an ambulance). Granted that she knows this, she must also know she is obligated to call an ambulance insofar as she can. But given determinism, we may worry that she has no reason to think that she can.

This concern is misguided. Firstly, we must dispense with any idea that if her intention is determined, then her actions are fixed no matter what she intends. To reason like this would be to commit the “fatalist’s fallacy”: even if her action is predetermined, this does not entail that it isn’t conditional upon her intentions. If she is determined to call an ambulance, this will be because it is determined that her deliberative process culminates in her forming an intention to call an ambulance, and this brings it about that she calls an ambulance. Determinism does not make our attempts to act causally ineffective.

Secondly, she has no reason to suspect, in advance of making up her mind, that she cannot call an ambulance. While it is possible that determinism robs her of the ability to call an ambulance, it might just as easily rob her of the ability to refrain. She has no reason to favour the presumption that her calling an ambulance is impossible over the presumption that it is inevitable. The only way that she can find out which of these she is determined to do is by reaching a decision.

From an epistemic perspective, both decisions remain open. As Pereboom (2001, 147–148), Fischer (2006), and Jeppsson (2016) have all argued, such epistemic openness is all we need in order for it to be rational to make a value-driven choice. As Fischer puts the point, if one were asked to choose which of two doors to walk through, and told that behind one them is a million dollars while behind the other there is a den of rattlesnakes, it would be ludicrous to suppose that the truth of determinism might weaken the rational case in favour of choosing the door with the money, or that one would be forced to just “wait and see what happens” instead of making a value-driven choice (2006, 329).

Moreover, suppose we grant that determinism introduces a doubt about whether Ada can call an ambulance (we should not grant this, given the deliberatively irrelevant nature of the “doubt”, when both options remain epistemically open, but suppose we grant it anyway). Doubts about whether we can do things do not usually weaken our rationale for trying when there is something morally significant at stake. Obviously, sometimes failure comes with other off-putting risks; you may be reluctant to dive into the river to save the drowning child, but it is usually the risk to your own life rather than the possibility of failing in your attempt that causes such reluctance. There is always some risk of failure, even with the simplest actions, regardless of determinism. One is “always at the mercy of the world”, as O’Shaughnessy famously notes (1973, 370). But it would be very strange for anyone to suppose that this should stop us from even attempting to bring about better outcomes.

Suppose that Sofia is in a hospital when the main power supply fails. Luckily, there is a short-term emergency power supply that will keep the electricity going for five minutes, during which time the back-up generator can be activated, saving the lives of hundreds of patients whose life-support machines will otherwise fail. Now suppose that Sofia is the only person with access to the button that activates the back-up generator. There would be something seriously wrong with Sofia if she reasoned as follows: “I only ought to activate the back-up generator if I can. But there is no guarantee that this button works, so I don’t know that I can. I therefore see no reason to bother pressing it”. Ordinarily, we do not need a guarantee that we can do something before we attempt to do it when there are morally significant outcomes at stake.

There seems to be no reason to suppose that UOS poses any serious obstacle to moral deliberation. Nonetheless, something emerges from this picture that might seem troubling. Essentially, we can escape being duty-bound to do things simply by choosing not to do them. If Ada does not call an ambulance, it will turn out, once her choice has been made, that she has done nothing wrong. Her choosing not to call an ambulance conveniently establishes that she had no moral obligation to call one. Moral obligations become easily escapable.

On the one hand, it may be argued that there is something conceptually amiss about the idea of a moral obligation that could easily be escaped; we might think that inescapability is an essential condition of moral duty. Hence we would still have a serious threat to deontic morality if it turned out that all of our purported “duties” were easily escapable. On the other hand, the worry may be about motivation; perhaps it will be accepted that we could have duties that were easily escapable, but we might wonder why anybody would comply with them.

3.2 UOS and Moral Obligation

The problem of easy escapability arises because we seem to have some power over whether we do certain things: even if causal determinism entails that we are unable to do otherwise, it does not entail that our actions are “strictly out of our control”; there is often a reliable causal correlation between our attempts to do things and the success of those attempts. UOS thus seems to give us a further power that might seem unpalatable; the power to escape being duty-bound to do something merely by choosing not to do it.¹³

We may well be aware of the fact that in forming the intention to act as we do, we will also be conjuring up proof that we lack any ability to do otherwise, and will therefore be actualising a situation in which we have no duty to do otherwise. This may appear to leave our moral duties precariously at the mercy of our wills. I see two reasons why this implication might look problematic; the first appeals to a Kantian notion of obligation, and the second rests on a broader conceptual concern about the inescapability of duty.

Firstly, philosophers influenced by Kant may suppose that moral duties are necessarily “categorical imperatives”. Kant distinguished hypothetical imperatives, which depend on our contingent aims and desires, from categorical ones, which apply to us necessarily regardless of our contingent aims and desires (1998, 25). When one is morally obligated to do something, the obligation is inescapable in the sense that one ought to do it (insofar as one can) regardless of whether one wants to do it.

Kant’s claim that moral duties are categorical imperatives is controversial. While this claim is plausibly at the core of any objectivist analysis of meta-ethics, many philosophers favour subjectivism. If moral duties are grounded in our subjective aims and desires, they will not be “inescapable” in this Kantian sense.¹⁴ But I am inclined to side with Kant here, so I will not pursue this line of argument. I doubt that anything without the character of a categorical imperative could seriously count as a “moral obligation”.

UOS is, however, perfectly consistent with the claim that moral duties are categorical imperatives. The DBWC notion of moral obligation certainly does not entail that moral duties depend on an agent’s subjective aims and desires (with the possible exception of certain duties towards oneself, if there are any). The reason why we are morally obligated to actualise certain possible worlds is because they are the most valuable of the ones that we are able to actualise, according to our favoured theory of normative ethics. And the reason why determinism, given Haji’s argument, entails that we are never obligated to actualise alternative worlds is not because we do not want to actualise those worlds, but because we cannot actualise them.

Ada should call an ambulance if she can. This has nothing to do with whether she wants to call an ambulance, and everything to do with the fact that the world in which she calls an ambulance is more valuable than the world in which she does not. It is not more valuable because her own subjective aims and desires deem it to be (perhaps she prefers the world in which she inherits a fortune and buys a Ferrari). It is more valuable because of the comparatively high worth of her character, her actions, and/or the likely outcome of those actions. More generally, whatever your favoured analysis of obligation, I maintain that it is these sorts of substantive moral considerations that ground Ada’s duties, and these need not leave her duties precariously contingent on her subjective aims and desires.

While it may be an essential feature of moral obligations that they are categorical imperatives then, this is not inconsistent with UOS. There is, however, a stronger sense in which it might be claimed that moral duties are necessarily inescapable. We might suppose that there is something wrong with the idea that there could be duties that are opted into; duties that we could have escaped being subject to in the first place. This sense does seem plausibly to be threatened by UOS, but it’s doubtful that this really is an essential feature of duty at all.

Promise-making is a prime example of a duty that has to be opted into. We typically suppose that we are duty-bound to keep our promises, even if we could have escaped taking on such a duty in the first place. The important point is that we did not escape taking on this duty. Consider another example: it is obligatory to feed one’s children as opposed to letting them starve. Nonetheless, many of us are under no such obligation, because we have chosen not to have children. While the same means of contraception were presumably available to many of those who chose to have children, citing this fact would hardly get them off the hook for letting their children starve. Again, the fact that they could, in theory, have escaped the obligation does not usually imply that they cannot have a genuine obligation if they did not actually escape it.

There seems to be no sense of inescapability such that it both plausibly qualifies as an essential feature of moral obligation and is plausibly ruled out by UOS.

3.3 UOS and Moral Motivation

Perhaps it is not moral obligation that is threatened by UOS, but moral motivation. While we may intelligibly have duties that are escapable in the sense specified by UOS, the worry may be that this would threaten any basis that we might have for complying with them.

Return to Ada: suppose we accept that her ability to easily escape being duty-bound does not undermine her duty, so long as she doesn’t in fact escape it. We might now worry about what sort of motivational basis Ada could have to incur the duty: by merely not bothering to call an ambulance, she can ensure that she had no obligation to call one in the first place. She only has a duty insofar as she willingly opts into it. Given that she stands to gain so much from not opting into it, we might wonder what incentive she could have for opting in.

We have already noted that duties do not, however, provide extra reasons for action that exert pressure on us independently of the moral considerations that give rise to them (see 3.1). I maintain that a competent moral agent acts out of duty not merely because it is her duty, but because she cares about the substantive moral considerations which underpin the duty (in terms of any DBWC analysis, these considerations determine the relative values of the rival intention-dependent worlds that she might choose to actualise). It is only if we accept the dubious assumption that the desire not to contravene a duty is the sole basis of moral motivation (and that the desire to fulfil duties is always curiously absent) that UOS seems to seriously undermine moral motivation.

Putting aside the possibility of determinism and UOS, let’s think about ordinary cases that parallel the sort of escapability of duty that we are contemplating. Suppose that Aisha believes that she ought to give blood so long as she is eligible to. She also knows that she has a blood donation appointment in one month’s time. Now suppose that she is planning to go on holiday before the appointment, and she is trying to decide where to go. She suddenly remembers that if she opts for the destination in sub-Saharan Africa instead of the destination in Europe, this will stop her from being eligible to give blood for at least a year. If it stops her from being eligible to give blood, it will also remove any moral duty that she has to give blood. Should we expect this to motivate her to opt for sub-Saharan Africa instead of Europe? Insofar as Aisha counts as a competent moral agent, I very much doubt that we should expect this. She may even regard it as a reason not to opt for the destination in sub-Saharan Africa.

Competent moral agents typically care about their duties because they care about the moral pressures that give rise to them. The reason why Aisha may be willing to incur the duty, even though she has been given an easy way of escaping it is because she cares about people who need blood transfusions. It is because of those people, after all, that she even takes herself to have a duty to give blood if she is eligible to; she thinks that the world in which she contributes to the supplies of blood banks is better than the world in which she does not. All those car crash victims and children with leukaemia are not going to just go away because she is not personally duty-bound to help them. If she didn’t care about these people, she might just as easily contravene the moral duty as escape it.

This brings us to the crux of the issue: the very same considerations that count in favour of fulfilling the duty, should you have it, count just as strongly in favour of opting into the duty, if you need to do so in order to fulfil it. And the very same considerations that count in favour of opting out of the duty, if you can, count just as strongly in favour of contravening the duty, if you cannot. In no case then, does the fact that the duty can only be fulfilled if opted into (i.e., UOS) change the agent’s reasons for deciding either way. Just like Aisha, the reasons that Ada has for fulfilling her duty to call an ambulance for her uncle (should she have such a duty) also count in favour of incurring the duty if she needs to incur it in order to fulfil it. And the same reasons she has to opt out of incurring the duty would count in favour of contravening the duty if its existence did not depend on her opting into it. In no case does it appear rational for her to arrive at a different decision, given UOS, than she would have arrived at without it.

It’s unclear why anyone would be keenly motivated not to contravene a duty, while at the same time caring so little about fulfilling one. Such a mindset seems to be directly inconsistent with the sort of sensitivity to value that characterises competent moral deliberation. What exactly is the imagined psychology of an agent who is highly motivated by an aversion to contravening duties while also trying to avoid fulfilling them? Such an agent, despite her thorough commitment to not contravening duties, would be completely indifferent to the moral pressures that actually give rise to duties, as well as being positively averse to fulfilling a duty if it’s possible to escape it. Even if it were possible for an agent to have this bizarre attitude towards moral pressures, this certainly does not capture the way most of us morally deliberate.

A competent moral agent typically reasons from considerations about the respective values of the courses of action between which she is deliberating to conclusions about what she ought morally to do. The moral landscape for anyone who reasons in this way seems to be largely untouched by UOS. So it does not appear to pose a serious threat to moral motivation.

4 Conclusion

Haji argues, similarly to Lockie, that there could be no moral obligations at all if determinism were true. In order to establish this conclusion, however, we must invoke both an “ought” implies “can” principle and an “ought” implies “able not to” principle. In section 1, I argued that without OIANT, we could establish only the weaker conclusion that there are no unfulfilled moral duties. In section 2, I argued that we ought to reject OIANT, and hence that only the weaker conclusion has been plausibly established. Finally, in section 3, I argued that while this weaker conclusion may initially look just as damaging, it actually has surprisingly little practical importance for morality. While I believe (contra Haji, and in agreement with Lockie) that determinism plausibly threatens moral responsibility, I deny that it poses any serious independent threat to deontic morality.

I admit that aspects of this thesis seem paradoxical. It seems odd to suppose that if determinism is true, this entails that nobody ever violates a moral duty. The air of paradox arises, I think, from two sources. Firstly, from the fact that we do not know in advance what we are capable of doing, since we do not know in advance which actions are impossible and which are inevitable. This means that acting otherwise remains an epistemically and pragmatically live option when we contemplate our potential moral duties in advance. Secondly, it may well be that the sense of “can” typically used in relation to principles like OIC is actually distinct from the sense of “can” according to which determinism robs us of the opportunity to do otherwise.

I have granted for the sake of argument that OIC is true and that we can use a single sense of “can” both in formulating OIC and in defence of the incompatibilist claim that nobody can do otherwise if determinism is true. This has the upshot that nobody can be obligated to act otherwise if determinism is true, and hence that there are no unfulfilled duties in a deterministic world. If that conclusion seems too counterintuitive to accept, then an alternative strategy would be to question whether we should accept all of the following three theses: (1) that OIC is true, (2) that determinism may well be true, and (3) that no one can do otherwise if determinism is true in precisely the same senses of “can” according to which “ought” implies “can”. My goal has been to argue that even if we accept all three, the threat to morality might not be as all-encompassing as it seems. Whether we should accept all three is another question entirely.¹⁵