Print This Post Print This Post

The purpose of this chapter is to show that concerns focused on the agent can supplement concerns focused on victims in bolstering the conclusion that agents ought not to infringe negative rights, even when the consequences of doing so are clearly better than the consequences of not infringing negative rights. That is, agent-focused considerations can bolster the case against consequentialism.

The chapter focuses on examples originally due to Bernard Williams; Williams himself advanced them to a similar end. The first is the well-known case of Jim and the Indians:

Jim finds himself in the central square of a small south American town. Tied up against the wall are a row of twenty Indians, most terrified, a few defiant, in front of them several armed men in uniform. A heavy man in a sweat stained khaki shirt turns out to be the Captain in charge and, after a good deal of questioning of Jim which establishes that he got there by accident while on a botanical expedition, explains that the Indians are a random group of the inhabitants who, after recent acts of protest against the government, are just about to be killed to remind other possible protestors of the advantages of not protesting. However, since Jim is an honored visitor from another land, the captain is happy to offer him a guest’s privilege of killing one of the Indians himself. If Jim accepts, then as a mark of the special occasion, the other Indians will be let off. Of course, if Jim refuses, then there is no special occasion, and Pedro here will do what he was about to do when Jim arrived, and kill them all. Jim, with some desperate recollection of schoolboy fiction, wonders whether if he got hold of a gun, he could hold the captain, Pedro and the rest of the soldiers to threat, but it is quite clear from the set up that nothing of the kind is going to work: any attempt at that sort of thing will mean that all the Indians will be killed, and himself. The men against the wall, and the other villagers, understand the situation, and are obviously begging him to accept. What should he do?

The right answer for a consequentialist (allegedly) is obvious: since the consequences of Jim’s killing the Indian are clearly and significantly better than the consequences of his refraining, he ought to kill one of them. Williams thinks that this conclusion isn’t obvious. Kamm doesn’t take a clear position on the issue: her concern is to argue that the permissibility of Jim’s killing is permissible is a function, inter alia, of to whom the responsibility for the negative consequences of the action attaches.

Kamm thinks that this is a consideration that the consequentialist must be blind to, since the consequentialist is committed to thinking that who ought to be held responsible is a function of the consequences of holding responsible. This is a mistake: being a consequentialist at the level of normative theory does not commit one to being a consequentialist with regard to the metaphysics of agency. Moreover (and independently of this point), a system of non-consequentialist responsibility-norms can be justified on consequentialist grounds (think of Vargas’s revisionism). Kamm makes the mistake because she thinks that moral responsibility is a topic of normative ethics. I think this is not merely a mistake but an important one: her claim that assigning responsibility properly helps us settle the normative ethics is dependent on this initial mistake.

Kamm limits her discussion to cases which are Pareto optimal: no individual is worse off as a result of Jim’s killing than would have been the case had he acted. Thus she excludes from consideration cases in which innocent bystanders are killed in order to save many lives. Even cases which have the Pareto feature might involve impermissible killing, she argues, when the responsibility for the killing attaches to Jim and not the Captain. When does the responsibility attach to the Captain and not Jim? When Jim collaborates with the Captain (this is a terminological oddity; we normally take collaboration with evil to be an extra bad feature of a case, not a better-making feature. But so far as I can tell, this is merely a terminological point). When Jim collaborates with the Captain to bring about the death of one Indian, the responsibility for the consequence that that Indian is killed attaches wholly to the Captain, and not Jim; whereas when Jim acts without collaborating with the Captain, the responsibility for that state of affairs attaches partially to him. The fact that the responsibility attaches partially to him is a reason against Jim’s killing one Indian.

Kamm compares two versions of Jim and the Indians, standard Jim (W) and C(Scan). In W, the Captain makes an offer to Jim (“if you shoot one Indian, I will release the others”). In C(Scan), no offer is made; rather Jim has an infallible brain-scanning device that tells him that the Captain will kill all the Indians unless Jim kills one. Kamm thinks that even if it is not clearly impermissible for Jim to kill one Indian in C(Scan) it is much worse for him to act in C(Scan) than in W, and the reason is that in W the responsibility for the death attaches to the Captain, whereas in C(Scan) it attaches partially to Jim.

Part of the explanation for why responsibility wholly attaches to the Captain in W but partially to Jim in C(Scan) is to do with the initiation of Jim’s killing: in W, the Captain initiates the causal chain of Jim’s killing, by making the offer, whereas in C(Scan) Jim initiates it. When Jim initiates a killing, the responsibility attaches partially to him, and this is the more fundamental explanation of the lesser degree of permissibility in C(Scan). Accordingly, when responsibility attaches partially to Jim in virtue of some fact other than initiation, initiation does no work at all. For instance, suppose Jim approaches the Indians with the results of the scan, making the offer to them. In that case, since the Indians have proper authority over their own lives, if Jim kills one of them with their permission the responsibility for the death attaches wholly to the Captain. That is why initiation is only part of the story.

When the responsibility for the negative consequences attaches wholly to the Captain, Jim is responsible for something, but only for the good consequences of his action. In these circumstances, Jim preserves his moral purity.

Now, recall the dialectical function of these claims about responsibility: it is to give us an additional reason not to be consequentialists. That is, they supplement Kamm’s standard anti-consequentialist strategy. I must say that insofar as that is their function, I think they fail. They are not independent anti-consequentialist considerations; they depend entirely for their plausibility on the supposition of the anti-consequentialist position. I think this is brought out when we consider how far Kamm departs from the standard views with regard to moral responsibility. Generally speaking, moral responsibility is taken to have two components, one to do with agency and the other to do with normativity. We can assign responsibility to agents independent of knowing whether their actions are morally significant. This actually seems quite obvious. Suppose you know that I have pulled a lever, but you don’t know what (if anything) pulling that lever does. Surely you can conclude that (on the supposition I pulled the lever freely and I knew what pulling the lever does) I am responsible for the consequences of pulling that lever, whatever they are? Normative theory settles the moral significance of the consequences, not the metaphysics of agency. Thus we look to normative theory not to see whether I am responsible for the consequences of pulling the lever, but to see whether (given that I am responsible) I am blameworthy, praiseworthy or neither for pulling the lever.

Obviously, Kamm cannot see things this way. If she did, she could not believe that adverting to considerations of responsibility help us to settle the normative issues. The procedure would be patently circular. But since she allows her intuitions about permissibility to settle issues about responsibility, the procedure is just as circular; it is just deceptively so.

She goes on to argue that considerations about moral responsibility affect Williams’ integrity objection in a related way: when the Captain bear all the responsibility, Jim’s integrity is intact. Once again, she is not entitled to the conclusion inasmuch as she is not entitled to her claims about responsibility. Similar remarks apply to her claims about physician-assisted suicide, with which she closes the chapter; once again we cannot settle the normative issues by appealing to her account of responsibility, since her account of responsibility presupposes that the normative issues are settled.

I note that in any case, Kamm seems wrong on the normative question. She holds that there is a lesser degree of permissibility if Jim initiates the act and fails to get the Indians’ permission. But consider this case: Jim is an opponent of the regime the Captain represents, and if he appears in the open to make anyone an offer, the Captain will shoot him. Otherwise the case is identical to C(Scan). It seem to me that here Jim’s shooting one Indian is just as clearly permissible as in any of the variations she presents.

In between the discussion of Jim and the Indians and her discussion of how the upshot of that case applies to questions of integrity and to physician-assisted suicide, there is another discussion which is related only inasmuch as it, too, focuses on a case from Williams: the case of the person who faultlessly kills a child. Williams argues, famously, that in this kind of case it is appropriate to feel agent regret. Kamm asks why it is appropriate. Her answer is that when one’s action was an option (not a duty; i.e. permissible but not obligatory), if one would have paid a cost to avoid a bad effect of one’s action had one foreseen it, then ‘one should undergo post-act costs to ameliorate the harm’ (325). She seems here to go beyond Williams (without acknowledging this fact): whereas Williams thought it was appropriate to experience agent regret and perhaps make an attempt at compensation, Kamm seems to think it is the agent’s duty to feel and act in these ways. Once again she appears to treat responsibilty as a topic in normative theory, inferring from the moral significance of an action that an agent is responsible for it. Willams’ intention (as I read him) was almost the opposite: to argue for the appropriateness of certain reactions despite the absence of responsibility. Correlatively, Kamm concludes that if the action was obligatory, agent regret is not appropriate. This seems wrong: the ambulance driver who kills a pedestrian while racing to the hospital is just as clearly subject to agent regret as is Williams’s agent.

Let me note that independent of my concerns about Kamm’s notion of responsibility, there is something very odd about her view, brought out in its application to the case of Jim and the Indians. She says, as we have just seen, that when an agent does their duty, they are exempt from agent regret (where, once again, agent regret is understood as imposing duties on the agent). She therefore asks whether Jim’s killing an Indian is a duty (in any of the variations presented). She says that we may plausibly conclude that it is only permissible, and not obligatory, because the costs to Jim are so high. Because killing an innocent person is so difficult, we cannot demand it of someone. This seems right, but notice what follows from it, on Kamm’s view: because Jim is only performing a permissible action, he ought to experience agent regret and make amends. Now this is crazy: Jim has added burdens only because he performed an action that would have been obligatory but for its cost to him. When someone performs an action that would have been a duty but for its cost to them, freely shouldering a burden they needn’t have borne, they deserve recompense, not a further burden.


  1. 1. Posted by Fiona Woollard | September 7, 2007 9:24 am

    Neil, thanks for a great summary and critique. I have a few comments.

    (1) I feel rather more sympathetic to Kamm’s normative-ethics based account of moral responsibility than you seem to be. Granted there is one sense of responsibility in which we can tell that a person is responsible for the consequences of her actions without knowing the moral significance of those actions. Kamm does say that there is some sense of responsibility (she calls it causal responsibility) in which Jim will always be responsible for the death of any Indian which he kills. Nonetheless, there may be a richer sense of moral responsibility which is sensitive to considerations such as those raised by Kamm. Certainly, my intuition is that there are cases in which one acts merely as the Agent of another and in which the consequences are not laid at one’s door.

    (2)You argue that Kamm gets the normative question wrong, claiming that in C(Scan) (or your version of it) there is no bias against Jim shooting without an offer. I’d like to comment on this in connection with Kamm’s claim that if the Indian’s are at risk as a result of a natural disaster, there is not bias against Jim killing one to save the others. I am inclined to agree that in both the natural disaster case and the C(Scan) case, it is permissible for Jim to act. However, it seems that there is less bias against Jim killing as the Captain’s agent, and even less against his killing as the Indians’ agent. I suspect that manipulation of the numbers involved may give us cases in which Jim could kill as an agent but not on his own initiative. However, I don’t think we’ll find such a difference between C(scan) cases and natural disaster cases.

    I think this suggests that the “acting as Agent” factor is morally relevant, but not for the reason suggested by Kamm. C(scan) and Natural Disaster seem to be morally equivalent, undercutting Kamm’s claim that there is a bias in favour of locating responsibility in another person if that person deserves it. I found the idea that we are required to “keep moral responsibility where it deserves to be” very strange. It seems to have odd results, such that if we can only save one out of two people, one of which will be murdered and the other die naturally, we ought to save the natural death, so that the would-be-murderer becomes a murder and gets the moral deserts he deserves.

    (3) Williams notes that agent regret “necessarily involves a wish that things had been otherwise, for instance that one had not had to act as one did. But it does not necessarily involve the wish, all things taken together, that one had acted otherwise.” (Moral Luck, 31). In “Remorse and Agent Regret” (Midwest Studies in Philosophy 1988), Marcia Baron uses William’s comments to distinguish several types of agent-regret. (summarised on 264).

    I suspect that Kamm’s claims about regret when one had done one’s duty, and the peculiar result that Neil points out in the ambulance case and with respect to Jim’s superogation, flow from failure to recognise the varieties of agent regret.

  2. 2. Posted by Neil Levy | September 8, 2007 1:25 am

    Fiona, I’m not arguing against Kamm’s claims about the relative permissibility of Jim’s actions across scenarios. I have no thoughts here at all (normative ethics is too hard for me). My claim is that insofar as Kamm takes her claims about responsibility to be an independent argument for her claims about permissibility, she is just wrong: her claims about responsibility depend upon her normative conclusions. This comes out clearly in the words you quote” keep responsibility where it deserves to be”. It would be a point in her favor if she could make the causal/moral responsibility distinction and I can’t. But I can. On my view, moral responsibility has two conditions: a control condition and an epistemic condition. When an agents cause something, but do not satisfy both conditions, they are causally but not morally responsible for the consequence (so if I was jogged by someone, or had a seizure I fail to satisfy the control condition; if I non-culpably fail to understand the nature of my act I fail to satisfy the epistemic condition).

  3. 3. Posted by Fiona Woollard | September 8, 2007 8:02 pm

    Neil, thanks for your response. I think it’s helped me get a little clearer on what the issue is, so I’ll try to re-comment.

    I find it very interesting that your epistemic condition includes the normative condition that you should not be culpable for failing to understand the nature of your act.

    In fact, it seems as if in many cases we do not lay a bad consequence of an agent’s action at his door because he is not culpable for failing to foresee it. Whether this happens depends on how bad the consequence was, how difficult it would be to foresee it and what kind of presures the agent was acting under. If a consequence is very bad and the agent’s aim trivial, we may hold him responsible even if it would have been fairly difficult for him to foresee. So we do need to something about the consequences of an action before we assign moral responsibility.

    Normative considerations can affect responsibility by determining when an agent meets the epistemic condition. It seems to me that something similar happens with respect to acting as the Agent of another. A person is not morally responsible for consequences that he brings about in justifiably acting as another’s Agent. We need the normative theory to tell us when acting as another’s Agent is justified.

    As Kamm does use her intuitions about cases to inform her judgements about responsibility, you’re right that this doesn’t provide a completely independent argument in support of non-consequentalism. However, surely it does provide some support. It seems plausible (to me)that Jim is not morally responsible for the death of the Indians when he (justifiably)acts merely as the Captain’s Agent. It also seems plausible that it should be easier to justify acting when you will not be morally responsible for any bad consequences. These two plausible claims explain our intuitions about the permissibility of cases.

    I didn’t intend to imply that you were unable to give an account of simple causal responsibilty. I simply wanted to suggest that once we have recognised that there is another type of responsibility that can be undermined by factors such as non-culpable ignorance, it might seem less unlikely that other factors, such as that the agent has non-culpably become a mere Agent, can also undermine this stronger type of responsibility.

    Perhaps there are three things at work here: Causal responsibility; Moral Responsibility 1 = causal responsibility + the epistemic condition; Moral Responsibility 2 = moral responsibility 1 without non-culpable mere Agency (and maybe some other conditions). You might argue that we should not call Moral Responsibility 2 by that name. Indeed, I am not sure if what is at stake here is a lack of responsibility or merely a lack of culpability. However, I think Kamm is on to some factor that affects which consequences we can lay at an agent’s door morally in something like the same way non-culpable ignorance does.

  4. 4. Posted by Neil Levy | September 9, 2007 12:25 am

    A quick response, before I leave for the US and my internet access becomes sporadic. I simply deny that we assess whether an agent is responsible for something by reference to its moral seriousness. The non-culpable ignorance clause sounds irreducibly normative, but it’s not: it can be cashed out in non-normative term (an agent is culpably ignorant regarding A if she is relevantly ignorant with regard to A and there is no opportunity in the past for her to acquire the relevant knowledge which she knowingly and intentionally passed up). Once again, I claim that responsibility is a topic in the metaphysics of agency. Or at very least Kamm owes an argument for why it is not (not only does she not offer an argument, she simply ignores the literature on the topic). I think her normative claims also enter into her intuitions about whether someone acts as an agent. The best thing to be said in her defense is that her view of responsibility coheres with her normative account, not that it offers any independent confirmation of it (coherence is a virtue, and to that extent might be seen as offering some support to the account).

  5. 5. Posted by Thom Brooks | September 21, 2007 1:47 pm

    My sense is that this is right: her view of responsibility is consistent with the normative account she offers. The trick then is finding independent confirmation of this view. I have noticed this worry creep in at several points.

  6. 6. Posted by S. Matthew Liao | September 24, 2007 6:51 pm

    Thanks very much for your summary and comments, Neil!

    I think that Kamm may be missing (or simply just denying) the point of Williams’s Integrity Objection. The point of this objection, as I’ve always understood it, is that Jim is someone who would feel agent-regret in killing the Indian, whether it is in W, C[Offer], or C[Request]. Suppose this is the case, according to Williams, there is something morally amiss to ask Jim to kill the Indian, even if this turns out to be the right thing to do (a la [simple] consequentialism), or (as I think Williams would argue) even if Jim would be completely faultless, as e.g. in C[Offer] or in C[Request] (a la Kamm).

    As a related point, I don’t think that Williams is claiming that Jim ‘ought’ to have agent-regret, as Kamm seems to interpret him (p. 325). That is, I think that Williams can accept that there might be other people who do not feel agent-regret when killing the Indian, and in such cases, Williams need not claim that these people ought to feel agent-regret. Contra Kamm then (p. 329), it is not a problem for Williams that there might be cases in which the Integrity Objection does not apply.

Post a comment

Name: (required)

Email Address: (required) (will not be published)