Sunday, December 18, 2016

Illustrating the Paradox of Deontology

One who accepts a "consequentialism of rights" might hold that deliberating killing an innocent person (let's call this "murder", for short) is so morally bad that it isn't justified even to save five lives.  But deontologists go further, suggesting that one should not murder even to prevent five other murders.  This seems puzzling: if murder is so morally horrendous, why should we not be concerned to minimize its occurrence?  This is Scheffler's paradox of deontology in a nutshell.

A deontologist might respond by suggesting that our moral aims are not so impersonal: we have a special responsibility for our own (present) actions, and so must regard our not (now) ourselves causing harm / violating rights as a distinctive moral goal.  Scheffler pushes back against this idea on pp. 415-6 of his 'Agent-Centred Restrictions, Rationality, and the Virtues':

[O]n standard deontological views, morality evaluates actions from a vantage point which is concerned with more than just the interests of the individual agent. In other words, an action will be right or wrong, on such a view, relative to a standard of assessment that takes into account a number of factors quite independent of the interests of the agent. And defenders of such views are unlikely to claim that the relevant standard of assessment includes agent-centred restrictions, but that it is a matter of indifference, from the vantage point represented by that standard, whether or not those restrictions are violated. For if it is not the case that it is preferable, from that vantage point, that no violations should occur than that any should, it is hard to see how individual agents could possibly be thought to have reason to observe the restrictions when doing so did not happen to coincide with their own interests or the interests of those they cared about. In other words, deontological views need the idea that violations of the restrictions are morally objectionable or undesirable if the claim that people ought not to commit such violations when doing so would be in their own interests is to be plausible. Yet if such views do regard violations as morally objectionable or undesirable, in the sense that it is morally preferable that none should occur than that any should, it does then seem paradoxical that they tell us there are times when we must act in such a way that a larger rather than a smaller number of violations actually takes place.

It's a fairly dense passage, so when teaching this topic last term I came up with a thought experiment to help illustrate.

Suppose that five innocent people whom you love are going to be murdered, unless you yourself murder a (distinct) innocent person.  Is it wrong for you to murder an innocent person in order to save your five loved ones?

Standard deontological theories will insist that murder, even in this case, is wrong.  But this may seem a difficult verdict to uphold, given that murdering the one seems preferable from both your personal standpoint and the impersonal standpoint.

Impersonally: five murders are worse than one.  Personally: there is a special moral cost to you in committing a murder, sure, but it is not so great a cost (we may suppose) as losing your five loved ones.  So, we may wonder, from what perspective does the deontological verdict have any normative force or appeal?

To get the verdict that murdering the one is wrong, the deontologist must hold that you are morally special (to override the impersonal verdict and get that your murdering one is morally worse than allowing five other murders to occur), but you're not so special that your interest in saving your loved ones overrides your putative moral obligations.  It's an awkward combination of claims to assert simultaneously.

How do you think the deontologist might best respond to this challenge?


  1. Your framing of the personal aim in the loved ones example is in terms of personal cost. I may simply be misunderstanding, but from your original characterization of the deontologist who makes the personal aims response I thought that responsibility, rather than cost, was playing the essential role. While there is a sort of abstract responsibility for the costs arising from one's actions, it isn't clear to me that this can be identified with the kind of personal responsibility for one's own actions to which I took the deontologist to be appealing.

    Again, it could just be that I'm not fully understanding how the personal/impersonal distinction is being cashed out here.

    1. Yeah, I think something along those lines is probably going to be the way the deontologist should go in addressing Scheffler's challenge.

      Just to clarify my set-up: I was thinking of the 'personal' perspective as being non-moral in nature, just representing something like what the agent personally cares about. 'Impersonal' here is something like Sidgwick's "point of view of the universe", or a "God's eye view". Obviously neither accurately represents the deontologists moral perspective, so the challenge for them is to articulate a competing moral perspective that seems well-motivated, and brings in just the right amount of agent-relativity (neither too much or too little).

      So yeah, I think the answer is going to be to bring in a weighty concern for agent-relative responsibility without so much weight to agent-relative interests. (I don't think there's any reason to think this can't be done. But there's a question about how compelling such a story is likely to be in this context. The loved-ones example might prime people to be a bit more sympathetic towards Scheffler's challenge than they might otherwise be.)

  2. You're still imposing a consequentialist framework on the question. Of course deontologists want to prevent all murders, and they will act to do so, but only through moral means. These leave them many murder prevention strategies, but murdering someone is not one of them. What you seem to be asking is: If none of the morally permissible strategies will work, shouldn't the deontologists just default to some kind of consequentialism, where they judge the morality of actions not by means but by outcomes, personal and objective? I don't think it will be hard for deontologists to answer "no". After all, if they tried their best to morally prevent the murders, there will be no blood on their hands - their own moral integrity remains intact. The *outcome* might be globally and even personally tragic, but that has bearing on the rightness of their actions only if you think that outcomes and not means determine rightness of actions. So I think the stalemate remains.

    1. Yeah, that sort of flat-footed response is available (effectively just insisting that murder is wrong and so mustn't be done, regardless). But there's a worry that it's a bit non-responsive, insofar as Scheffler's challenge prompts us to rethink our intuitive commitment to the wrongness of life-saving murder. Why is it so important for one to have "no blood on their hands" (more important even than saving one's loved ones and achieving the impersonally better outcome)? The challenge for the deontologist to explain the normative force of their proscriptions in this context, without implicitly committing themselves to (at least) a consequentialism of rights.

  3. Hi Richard,

    I think a few potential replies are:

    1. If the deontologist also believes in some sort of afterlife punishments, or karma, etc., an answer would be that it's also in a person's own interest not do commit the murder.
    2. Else, two possibilities are:
    2.1. It's not wrong to murder an innocent person in order to save your five loved ones. After all, deontology is not committed to a particular demand on murder, or lying, even if those are the most common cases.
    2.2. It's wrong to murder an innocent person always, but it may well be overall in your interest to do so; even though you morally should not do it (i.e., you shouldn't do it in the usual moral sense of "should"), it's not the case that you overall means-to-ends shouldn't do it.

    In addition, the deontologist might reply by modifying the scenario a bit, and raising a challenge to a consequentialist view:
    For example, what if it's not 5 loved ones vs. 1 stranger,, but 1 loved one + 1 stranger (which happens to be a dog) vs. 1 innocent stranger, and also the action is not only murder, but, say, rape+burned to death? (if that's not enough, one might picked worse). Would it still be morally acceptable to rape the innocent stranger, then burn her to death (or worse if needed), to save the loved one plus the dog from a similar fate? (one may assume there is enough evidence to make the probabilities of the results in question very close to one).

  4. Personally, I think that the bigger challenge lies with the agent-neutral consequentialist. She needs to explain why we should think that an agent's only moral reasons are those that she would have even if she were to abstract away from her relations to her actions and those whom her actions affect and thereby take up the "point of view of the universe" or the "God's eye view." For why think that such relations are irrelevant from the moral point of view? By comparison, the challenge for those who deny agent-neutral consequentialism is merely to explain what makes a reason a moral one as opposed to a non-moral one -- that is, what makes a reason of the sort that can generate a moral requirement as opposed to merely a rational requirement. And my thought would be that there are some relations that we bear to our actions and to those who are affected by our actions that we must care about independent of whether we take a personal interest in those relations. Thus, I must care not to bear the relation of being someone's murderer even if being their murderer is in my self-interest and even I have made committing murder my project. And I must save my daughter as opposed to some more gifted stranger even if I've made acting in accordance with utilitarianism my project.

    1. Hi Doug, I can go along with giving extra weight to the interests of one's loved ones. What I have trouble with is giving (so much) weight to non-welfarist considerations. Why must we care so much about not bearing the relation of being someone's murderer (compared to bearing the relation of not having prevented someone's murder)?

      Do you think that reason requires you to prefer that your five loved ones be murdered, in the case I describe?

    2. Why should we give so much weight to welfarist considerations? It seems to me that there's no explanation for why one moral consideration has this or that much weight. As it happens, I do think that reason requires me to prefer the world in which my five loved ones are murdered by someone else to the world in which I commit murder.

    3. Because people matter a lot, and it's really bad when they're seriously harmed? Perhaps that's not so much an explanation as just re-stating the claim in a more rhetorically convincing idiom. So in fairness perhaps I should likewise not ask for an explanation of why we should care non-instrumentally about deontic constraints (or our relations to certain act types), but just an invitation to say something that sounds more convincing in this context...? Caring (non-instrumentally) about act types just seems completely unmotivated to me.

    4. Because treating people as ends-in-themselves matters a lot, and it's really bad when they're treated as a mere means. Of course, I don't imagine that the utilitarian will find this any more convincing than the Kantian will find your explanation for why welfarist considerations matter so much more than such non-welfarist considerations will.

    5. And interpreting the deontologist as caring fundamentally about act-types as opposed to people seems a bit uncharitable to me. It's just that caring about the value that people have may call for more than just the response of promoting their welfare. It may also call for respecting their autonomy.

    6. Okay, so now Scheffler would argue that this appeal to the moral undesirability of "treating people as a mere means" implies that we should be concerned to minimize its occurrence. (It's not as though my treating someone as a mere means is morally more important than anyone else's doing so.)

      [Also, of course, it's just not true that utilitarian tradeoffs involve treating anyone as a "mere means", unless we stipulatively redefine that term to mean "contrary to Kantianism" rather than anything to do with genuinely merely instrumental valuation.]

    7. Richard,

      A comeback would be that it's morally wrong to care more about saving their loved ones (if it is; but let's simplify), so in the moral sense of "ought", they ought to care more about not being a murderer. If they in fact care more about saving their loved ones, then they're not doing what they ought - i.e., they're being immoral.

      Now, they might not say why it's immoral, but then again, in all theories, there always seems to be a bottom - i.e., you reach a point at which there is no "why" but a brute fact.

      For example, your consquentialist also does not seem to say why people matter a lot, and whether it matters to an agent depends on the agent's constitution. If the claim is that people ought to matter a lot, then your consequentialist still does not seem to explain why they ought to matter a lot to, say, the alien invader in my example (or why other people should matter more to a human person than not being a murderer, for example).

  5. Richard,

    I think that a deontologist (at least, one that gives a lot of weight to our constitution, which I grant is not a common deontologist view, but I think it's the best deontological candidate) has an alternative reply: they might reply that it's our psychological constitution to care about, say, not being a murderer, a rapist, etc., and/or about not to break moral rules, and if that is unmotivated, then so is caring about other people, about preventing bad things, etc.

    On that note, it seems to me that what an agent cares about depends on the agent's constitution, and the agent's constitution itself - or at least, some basic part of it - does not have a motivation - it just is.
    For example, you say that people matter a lot. They matter a lot to some and perhaps to most humans, and it's really bad when people are seriously harmed (unless perhaps they deserve it, but let's leave that aside).
    However - for example -, people do not need to matter at all to coherent, superintelligent moon-sized alien (perhaps an AI, perhaps a cyborg, etc.) that decides to take over the solar system and use up the resources, killing every human in the process simply because they're in the way (if you count the alien as a person, then people other than the alien do not need to matter to it; in any event, humans do not need to matter at all), and morality also does not need to matter to it. In particular, it doesn't need to either know or care that killing people is a really bad thing. It values positively the acquisition of the resources that it achieves while it kills people, and is coherent about that.

    So, we humans care about people - or about avoiding the bad - because it's our constitution as human beings, but other entities do not need to care, even if they're more intelligent than we are. The deontologist can say the same goes for caring about not breaking rules, or about not being a murderer, or a rapist, etc. - it's just our constitution. It's also our constitution to care more about not committing murder than about not preventing 2 murders (for example).

    1. It's not a normative theory if you're just describing what people do care about, rather than specifying what they ought to care about. E.g. The "deontologist" you describe would seem to have no come-back to anyone who cares more about saving their loved ones from being murdered than they do about committing murder once.

    2. Sorry, I meant to reply to your reply here, not to the other reply above (i.e., a come-back is that he ought not to care more about that, and that's a brute fact, but all moral theories have brute facts without "why" explanations).

  6. Hi everyone.

    Prof Chappell, I was wondering whether the deontologist could construe the wrongness of murder as consisting solely in forming the intention to murder (because, say, such intention manifests a very ill will).Could she then claim that all you would achieve by murdering in order to save your beloved ones would be the addition of an instance of your own ill will to the already instantiated ill will(s) of the prospective murderer(s) of your beloved ones?
    Apologies if the suggestion is very silly!

    1. Interesting suggestion! To deal with that argument, we'd need to consider a variation on the case whereby committing one murder yourself would somehow prevent five others from even forming the intention to murder your loved ones in the first place. (Perhaps you would prevent some event that is essential to the moral corruption of the potential murderers. So by preventing this event, you not only save the lives of your loved ones, but you also save the moral virtue of their would-be murderers.)

    2. Thank you! I could tell there was an obvious counterargument to my suggestion, but i couldn't pin it down.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.