What does ‘ought’ implies ‘can’ imply about moral dilemmas?

I've had moral dilemmas on my mind lately, and I'm troubled by a common argument given against the possibility of genuine moral dilemmas. I'm hoping people can help diagnose what's troubling me. And I apologize in advance for the disorganized thinking.

Here's the argument: Suppose an individual S is obligated both to perform act A and obligated to perform act B. S is therefore obligated to perform (A&B). Assuming that for S to be obligated to ø entails that S ought to ø, then S ought to A, ought to B, and ought (A&B). Applying 'ought' implies 'can,' then S ought (A&B) entails S can (A&B). But the circumstances of the world are such that S is metaphysically precluded from performing (A&B), so she cannot (A&B). Hence, it is both true that S ought (A&B) and false that S ought (A&B).


Skeptics about the existence of genuine moral dilemmas take this argument to be a reductio of the possibility of such dilemmas: It must, appearances to the contrary, be the case that S is not obligated to A OR S is not obligated to B. And the intuitive insight is simple enough: If I ought to do only what I can do, and I can only perform one of two acts, it can't be true that I ought to perform both. Thus, I can't be obligated to perform both.

But this argument has always struck me as fishy somehow. First, one might have doubts about 'ought' implies 'can' (OIC). Both Charles Pidgen and Geoffrey Sayre-McCord have papers that have convinced me that OIC is probably not a theory-neutral principle of deontic logic but a substantive normative commitment.  As Sayre-McCord notes, it seems suspect that a principle of deontic logic can rule out certain conceptions of morality (e.g., Calvinism) on the basis that they reject OIC.

My concerns lie elsewhere though. Contrast the role OIC plays in this dilemmas argument versus its role in arguments about moral blameworthiness . Here's a familiar sort of argument: Suppose I fail to perform some act A, like rushing out into the street to save a toddler who's wandering into the path of an oncoming bus. But as it turns out, I was much too far away to reach the toddler before the bus arrives. There is a sense then in which it was not true that I could have saved the toddler. Perhaps if circumstances had been very different, I could have. But in the actual circumstances, I could not have saved the toddler. To say that I ought to have saved the toddler seems misplaced then. Furthermore, this in turn shapes what reactive attitudes I and others rationally ought to have toward my action. For instance, I ought not to subject myself to self-reproach, and others should not hold me blameworthy.

But suppose the situation instead presents an apparent moral dilemma (You don't need to accept that this really is an intractable dilemma for me to illustrate my claim.): I am near enough to the wandering toddler to be able to save her, but to do so, I would have to leave my elderly grandmother, whom I'm escorting to her doctor's appointment and who suffers from dementia and confusion, at the street corner. If I choose to save the toddler and my grandmother ends up lost somewhere in the city, it would seem reasonable to hold me blameworthy for that outcome. Alternatively, if I remain with my grandmother and I opt not to save the toddler, it would seem reasonable to hold me blameworthy for that outcome.

What does this show? It reminds us that the 'can' in typical discussions of responsibility or blameworthiness is categorical: I ought to be held blameworthy for not having done A only if could have done A (or I ought not be held blameworthy for A unless I could have refrained from A). In contrast, the 'can' of 'OIC' in the dilemmas argument is conditional: I can't do A if I do B (or I can't do B if I do A). This is why we are more inclined to react to the dilemma situation with our characteristic reactive attitudes of praise, etc. than we are to react moral blameworthiness argument in the same way. That is, let us suppose that A violates a moral obligation. If I do A because I cannot, my act does not seem open to rational appraisal in the same way if I do A because I opted for A in order not to violate some obligation I would violate by doing alternative B. In the dilemmas case, I have endorsed A in a more robust sense than in the moral responsibility case. In opting for A over B, I have, in effect, invited the criticism that will come for failing to do B.

But this causes me to wonder if there aren't two OIC principles in play here. Moreover, I wonder if there isn't something question begging about the original argument that uses OIC to deny the possibility of genuine moral dilemmas. Help?!

10 Replies to “What does ‘ought’ implies ‘can’ imply about moral dilemmas?

  1. “But the circumstances of the world are such that S is metaphysically precluded from performing (A&B), so she cannot (A&B). Hence, it is both true that S ought (A&B) and false that S ought (A&B).”
    I think this is a common misreading of the deontic principle. Suppose you take obligation to be closed under conjunction. In general, moral dilemmas won’t show that the closure principle is faulty, since semantically O(A & B) amounts to no more than the claim that in all morally ideal worlds A & B is true. Since there is such a world, it is possible that A & B. There is no worry for ought-can. The worry, of course, is that we are not in a morally ideal world and (probably) don’t believe that what we ought to do in sub-ideal worlds is anything like what we ought to do in ideal worlds. In other words, the moral principle you’re assuming–say it’s some form of utilitarianism–won’t have the logic described in standard deontic logics unless it is an odd form of ideal utilitarianism. If it is that form of utilitarianism, you cannot generate a moral dilemma (well, almost can’t). If you can generate a moral dilemma, then you’ve got the wrong logic.

  2. A couple of options: both the applomeration principle (O(A) and O(B) entail O(A&B)) and OIC are questionable. My own probabilistic semantics for ‘ought’ invalidates agglomeration, for example (although this doesn’t explain why it wouldn’t hold in the case of DUTIES). And there’s a good case to make that ‘ought’ only pragmatically implies ‘can’ (Sinnott-Armstrong 1984): if you’re trying to guide somebody, it’s not very helpful to consider impossible courses of action.

  3. This might be an overly simplistic way of responding to the problem(s) you pose, but I take the central one to be about the nature of OIC. Maybe you can duck the dilemma by noting the commonality between the two options-(A:I [morally] ought to save the toddler and B: I [morally] ought to stay with grandma) is that they are both MORAL oughts (as opposed to other oughts, like ‘I ought to save for retirement’). If the OIC principle was a general principle, why can’t you say that for situations where you have OughtA and OughtB, and both of them are moral oughts, that the real point of the principle is that you have OughtMORAL, and you fulfill your obligation (or act morally, or whatever the language must be) so long as you do ANYTHING that is a current moral ought? If you A, you’ve fulfilled your moral obligation, but you can also fulfill is by B-ing. You can’t do both, by stipulation, but (either) one is enough to fulfill the general OIC principle.
    I guess, in a way, this is to deny this principle: (O(A) and O(B) entail O(A&B)) and replace it with (O(A) and O(B) entail O (A v B)). The open-endedness of the principle allows the second half of the equation to be replaced by anything in universe of ‘moral oughts’. I admit, it would be weird to claim that someone who failed to both A (save the toddler) and B (keep Grandma safe) because they suddenly decided to do C (help with Haitian earthquake relief) was acting well, but I think that’s the devil in the details about what makes a moral ought an ‘active’ ought at a particular time. I don’t think a theory would be unreasonable to give priority to time-sensitive moral oughts (like A and B) over less time-sensitive ones (like C).

  4. I think Steve and Eric are on the right track regarding the source of the problem, which is the principle: O(A) and O(B) implies O(A&B).
    Consider an analogy: It is possible for me to turn right at the next intersection and it is possible for me to turn left at the next intersection. This does not imply that it is possible for to turn both left and right and the next intersection. In other words, “1. P(L); 2. P(R)” does not imply “3. P(L&R)”, but rather only implies “3. P(L) & P(R).”
    I think something like this is going on in the moral dilemma case.

  5. Consider an analogy: It is possible for me to turn right at the next intersection and it is possible for me to turn left at the next intersection. This does not imply that it is possible for to turn both left and right and the next intersection
    That’s wrongly assimilating deontic obligation to alethic possibility. Deontic obligation is the counterpart of alethic necessity. Deontic permissibility is the analogue of alethic possibilty. Agglomeration does not hold for permissibility for the reasons you cite.

  6. Thanks everyone for the comments.
    Mike, I can see the force of moving to a possible world semantics for ‘Ought’, but as you note, this seems orthogonal to the anxieties many moral theorists have about the possibility of genuine dilemmas. That there is some world in which S can (A&B) hardly seems an argument that answers to worries about action guidingness, consistency, etc. in this world.
    Eric and Steve both want to deny agglomeration: Eric opts for a disjunctive ‘ought’ or disjunctive obligation. Eric, I’d be curious to know how that disjunction would play out in terms of the sorts of concerns I raised in my post: blameworthiness, reactive attitudes, etc. My own sense is that in fulfilling a disjunctive obligation, I’d still be rightly subject to criticism, etc. for ‘the road not taken,’ so to speak, in a way that I’m not in standard applications of OIC where I couldn’t have (categorically) done otherwise. What do you think?
    Steve: Remind me again why your semantics doesn’t yield the denial of OIC in the case of duties. I’m interested because the OIC argument meant to show the impossibility of genuine dilemmas could simply be recast in terms of ‘Obligation implies can,’ with standard OIC omitted. Is that right?

  7. Michael,
    I was saying that while my semantics explains the invalidity of agglomeration, this doesn’t extend to duties (the point being that duties can also be expressed with “must”, being a matter of deontic necessity). But I was providing it only as an example; there may be other reasons why agglomeration fails for duties. (Suppose, for example, that we define duties as things that we owe to specific persons. It wouldn’t follow from the fact that we owe A to a specific person and that we owe B to a specific person that we owe A and B to a specific person).
    I wasn’t linking my semantics with OIC at all. But I think the same argument that ought implies can only conversationally also extends easily to ‘must implies can only conversationally’. The typical purpose of telling somebody what they ought/have to do is to influence their behaviour, so it is pointless to tell them they ought/have to do something impossible.

  8. Michael,
    I think it is true that there is some sense that my disjunctive obligation argument still leaves a moral agent open to some form of moral criticism, but the way I was trying to cast it should make a milder form of moral criticism than “you failed to fulfill your moral obligation”. The agent who A’s (or B’s) HAS fulfilled a moral obligation. In an ideal world, he would be able to do both A and B, but in this (real) world, he can’t do both, so he does (morally) right by doing at least one thing he’s morally obliged to do (either A or B).
    If there is a sense in which he can be morally critiqued (“you should have helped that child!” or “you should have stayed with grandma!”), it seems to me that it can’t be a serious, sustained critique without expecting someone to be able to do the impossible. At best, the critique can only be something like “Would that you could have helped the child” and nothing more stringent.
    I’m assuming that there is no available lexical ranking or prioritizing of the moral obligations; A and B are equal in all of the relevant senses.
    I think this disjunctive explanation also helps brush off some Singer-like arguments about samaritan duties. I may have an obligation to help the distant needy, but so long as I am fulfilling other moral obligations, I have not acted badly. Again, though, this only works when the moral obligations are equal-if helping the distant needy is lexical more important than, say, being a coach for my child’s soccer team, I could be duly criticized for failing to help the needy. But that would require, at minimum, at least an argument supporting the lexical ordering.
    I guess, to summarize my rambling, that the disjunctive explanation takes some of the sting out of the moral dilemma. I think it allows the dilemmas to be considered ‘real’ without seriously condemning someone for having to make a choice.

  9. Mike Almeida,
    Thanks, that’s helpful. It seems to me that the problem that Michael Cholbi is concerned with arises just in virtue of this commonly assumed disanalogy between possibility and obligation. If O(A), O(B), implies O(A&B), and P(A), P(B), does not imply P(A&B) (where P is the modal operator for possibility, because I don’t know how to render a diamond in a comment), and if we understand “ought implies can” to mean O(x)->P(x), then “ought implies can” is expressly connecting possibility with obligation.
    One solution is to deny the deontic principle that O(A) and O(B) implies O(A&B). I think that there is a way of making sense of this, at least with imperfect duties, more or less in the way that Eric does. So, even if performing A does satisfy some moral obligation, and performing B does satisfy some moral obligation, it is not the case that I have an obligation to satisfy all instances of my imperfect obligations at the same time. For Kant, I think, what I have is an obligation to ensure that all of my actions satisfy some imperfect obligation or other, within the constraints of my perfect duties.
    A different solution is to question this version of “ought implies can.” Kant doesn’t construe this principle as a matter of physical possibility, but rather as having to do with motivation. I think that construing it as a matter of physical possibility make the principle implausible. For instance, I have an obligation not to damage others’ rightful property (this may be the source of their claim to damages if I do cause harm). I have this obligation even when I am driving and I lose control of my car, even when I have driven well within reasonable constraints. Still, if I run into your fence through no fault of my own (at least so long as it not someone else’s fault) I can rightfully be made to recoup your loss, because of my obligation not to damage your rightful property. So even when it is not physically possible for me to avoid violating my obligation, I take it is plausible that the obligation remains in force. If that is right, then the principle O(x)->P(x) is not a good one.

  10. if we understand “ought implies can” to mean O(x)->P(x), then “ought implies can” is expressly connecting possibility with obligation.
    Pete,
    Ought-can, as it is used in deontic contexts (I mean, standard deontic contexts) just tells us that our obligations are consistent. This is perfectly analogous to the alethic modal case were (certainly) Nec P & Nec q -> Pos (p & q). So, combining moral necessity (deontic obligation) with alethic possibility does not generate the problem you cite. Or else, I don’t see it.
    One neglected feature of rejecting this or that theorem in SDL is that it can remove the inconsistency without accommodating moral dilemmas. Suppose you decide to reject agglomeration. This is pretty easily done using minimal models. But doing so semantically models someone who endorses two moral standards with two sets of ideal worlds. That can get you OA and O~A consistently, but it does not model a moral dilemma. All that tells you is that, relative to one standard, OA, and relative to another O~A. It tells you nothing about what you ought to do “all in”. Genuine moral dilemmas have agents facing situations in which they are all-in obligated to do A and all in obligated to do ~A. It’s really hard to model that situation properly. I’m not sure it can be done.

Comments are closed.