Print This Post Print This Post

I want to apologize in advance for the delay in this posting, and for the questionable way I’ve used the additional time. I’m afraid I lacked the discipline to adhere to the prescribed summary/critique format, or to frame my critique as questions for group discussion. I find Chapter 4 tremendously engaging but very frustrating. After some preliminaries, I intersperse criticism with exposition. This is partly to motivate Kamm’s arguments and further distinctions, but also, I admit, out of sheer impatience. In making this criticism, I offer some case-variants of my own; this imitation, however poor, should be taken as a sincere form of flattery. Anyway, I hope that I’ve succeeded in presenting enough of this Chapter’s striking, original, and idiosyncratic claims to provoke a lively discussion. Of course, you should consult the text of Chapter 4 closely, because I can’t vouch for the accuracy of my quotes or paraphrases, let alone my interpretations. I also apologize for the numerous typos you’ll come across, but I want to get this out while there are still a few days left for discussion.

In Chapter 4, Kamm presents what she believes to be a more nuanced account of what it means to intend evil, and of the relationship of intention to the permissibility of actions, than that found in the Doctrine of Double Effect (DDE), or its revision, DDE(R), which condemns not only intending evil, but intending others’ unconsenting involvement in that which the agent foresees will lead to evil for him (93; Kamm uses “evil*” as shorthand for evil/ such involvement in evil; I won’t.). This account, which she calls the Doctrine of Triple Effect (DTE), is developed through a series of cases and variations, but its core is a seemingly simple distinction, introduced in Chapter 1, which Matthew describes as that “between doing something because an effect will occur and doing it in order that it occurs”. DTE, which relies on this distinction to state conditions under which causing evil is justifiable, appears to supplement or qualify DDE (R), which states conditions under which it is not.

Chapter 4, which clarifies, defends, and refines the because/in-order distinction, is based on an earlier paper by Kamm but plays an integral role in the development and defense, in Chapter 5, of her own principle of permissible harming (PPH). PPH, as we will see, tries to make intentions irrelevant for permissibility. Chapter 4 can be seen as setting the stage for PPH by developing the most plausible account of the relevance of intentions. For those who find DTE implausible, however, claiming that PPH is superior to DTE might seem like damning with faint praise. I will focus on the first third of the chapter. In much of the second third, Kamm applies the because/in-order distinction to the claim that a rational agent who intends an end must intend what she believes to be the means to it. I will not comment on this discussion of instrumental rationality, but will return in closing to Kamm’s intriguing suggestion in the last third of the chapter that the because/in-order distinction applies to bringing about good as well as bringing about evil.

The DTE case our group is most familiar with is Munitions Grief, to which I’ll briefly return. But it’s useful to start where Kamm does, with the Original Trolley Case (OTC) and its Loop variant. In OTC, where the agent kills the one by diverting the trolley onto a spur to save the five, most nonconsequentialists (NCs) seem to agree (1) that diversion is OK, and (2) that however clearly the agent foresees that the trolley will hit the one, she does not intend it. In Kamm’s version of Loop, where the diverted Trolley will come around and hit the five unless blocked (but would not hit the one if the five blocked it), an “extra bit of track” makes the second conclusion much less obvious: the agent intends to save the five from the trolley and knows that the only way he can do so is to hit the one, since otherwise the trolley will loop back and crush the five. Most NCs who find the diversion acceptable in OTC also find it acceptable in Loop. Most also hold that because the agent knows that hitting the one is a necessary means to saving the five she intends to hit him and uses the one as a means to saving the five. Loop thus presents an apparent exception to DDE (R), since it permits the agent to intend evil as a means to achieving a greater good. It is this conclusion Kamm resists. (Those NCs who find it impermissible to turn the trolley in Loop may hold that an “extra bit of track” can make a difference. Or, like me, they may deny that it is permissible to turn the trolley in OTC, and claim that the acceptance of the diversion in OTC rest on the kind of suspect intuition that should be rejected in a robust exercise of reflective equilibrium. We will hear no more from these outliers here).

Kamm accepts (and imputes to the agent the belief) that it would be pointless to turn the trolley if the one were not on the tracks to stop it; the additional minutes or seconds this would buy for the five would hardly be worth the effort of turning the switch (it might be worthwhile if it would buy them hours, or if getting hit from behind would spare them the awful fear of imminent death, but we can assume that there are no such benefits in Loop). Kamm thus agrees that the agent turns the trolley in Loop only because it will hit the one; that she would not turn it if it would not. But she denies that the agent intends to kill the one; this is the basis of her distinction between “doing something because an effect will occur and doing it in order that it occurs.” She claims “a general conceptual distinction” (95) between the two, and tries to elicit intuitive support for that distinction with Party. In that case, the agent wants to throw a party for her friends, but not if there’s going to be a big mess to clean up afterwards. She thinks, though, that her friend’s indebtedness will cause them to help clean up, and she would not throw the party if she did not believe it would produce such a feeling of indebtedness. She only throws the party because it will produce indebtedness, but it doesn’t seem correct to say that she throws the party in order to make her friends feel indebted, or that she intends to make them feel indebted. Kamm notes disanalogies between Party and Loop (while the good of having the party would be realized even without the bad – the guests’ feelings of indebtedness and clean-up efforts, the good of saving the five would not be realized without the bad of the trolley hitting the one; while the guests are still better off with the good and the bads than with neither, the one on the tracks clearly is not). Party can be altered, though, to make it more analogous to Loop without making the host appear to intend to do evil.

Kamm sees the Party and Loop cases as sharing two features that block the ascription of intent to cause evil to the agent. First, the role of the evil is not to bring about the end the agent pursues, but to prevent something from defeating the attempt to achieve it, or more broadly, from reducing or eliminating the value of achieving it; second, the agent is not committed to doing everything possible to make sure that this evil befalls the individual; indeed, he may be permitted or even required to try to avoid or mitigate that evil after she involves the individual in a way she expects to expose him to it. These two conditions are explained and defended in several further cases. Their plausibility is critical to evaluating the triple-effect doctrine Kamm formulates at the end of the chapter: “A greater good that we cause and whose expected existence is a condition of our action, but which we do not necessarily intend, may justify a lesser evil* that we must not intend but may have as a condition of the action” (118).

The first condition requires an individuation of the threats or problems that the agent is trying to avoid. This can be clearly illustrated by Party, more uncertainly by Loop. In Party, the agent wants to have good time, an end will be defeated (whether subsequently or concurrently) by her having to clean up the mess all by herself. Her end in throwing the party is not to avoid messes, even if that is a long-term goal, so inducing feelings of indebtedness in her guests is not a means to achieving her end: even if avoiding messes is another end, it is not one she pursues by throwing a party. (But can’t her end be described as “throwing a hassle-free party”, for which inducing indebtedness does seem a means? Kamm addresses this challenge, but not with this example; I will return to it later.) But what of the end in Loop? Here, Kamm describes the structure the case in a manner that ascribes narrow ends to the agent, by differentiating the threats or problems posed by the undiverted and diverted trolley. First, there is the threat of “the trolley heading toward them in one direction (94). The diversion takes care of this problem. But this creates a second threat and problem: “the trolley coming at the five from a different direction”. This “only arises because I redirect the trolley away from the five. One way to see this new problem is as a second threat facing the five because I have taken care of the only threat that faced them to begin with”. Kamm addresses the question of why the problems should be differentiated this way by treating it as a question of whether different threats must have different objects doing the threatening. She responds that Loop “is the same, for moral purposes,” as Wagon, in which turning the Trolley sets off a different vehicle around the loop, which can only be stopped from hitting the five by first hitting the one.

Now it might be thought that the Wagon example merely showed that the individuation of threats/problems, and correspondingly, of ends, was so flexible that it could be easily manipulated to suit different “structures” in the Loop and other cases. Why should the agent’s end not be more broadly framed as keeping the five safe from getting run over, and the one problem she faced that of averting the various ways that various vehicles could run over the five, given the trolley’s present path, the current location of vehicles, the existing configuration of tracks, etc.? It may seem more natural to distinguish the current operation and the rebuilding of the factory in Munitions Grief than the front and rear approaches of the trolley in Loop, but no principled basis has been offered for making this judgment. In comments on this issue presented at an APA panel, Henry Richardson pointed out the problem with using threat-individuation as a basis for determining the morality of Trolley diversion. (I quote from an earlier draft of his comments):

To straighten out this problem, let me straighten out the tracks. Specifically, let me propose the following, essentially equivalent case:

Bypass Case: Although the track is elsewhere single, there is a brief doubling of the tracks between the trolley and the five. If we redirected the trolley onto the right-hand track, it will hit the one who is on it and stop. If the one were not on the right-hand track, the trolley would rejoin the main track and hit the five from the front.

Given Kamm’s amendment of Thomson’s case, the Bypass Case seems to me to be morally equivalent to Loop. Turning around Thomson’s rhetorical question, how can it make a difference whether the threat that dogs the redirection is due to a loop or to a rejoining of tracks? Yet in the Bypass Case, we can no longer say to ourselves, abstracting from the further consequences that include the evil*, that we have saved the 5 from being hit from the front. We can say that we have saved them from being hit by a trolley that has traveled on the left-hand track; but we might as well paint the trolley green and say that we had saved them from being hit by a red trolley [Richardson credits John Mikhael with the case on which Bypass is based].

If it the distinction between doing-because and doing-in-order-to depends on the elastic, if not arbitrary, individuation of threats, problems, and ends, it would be hard to sustain. But Kamm has another way of sustaining it: an agent who does something because an effect will occur and not in order that it occurs, is not committed to any further actions that will ensure that this effect occurs, and may even be committed to further actions to prevent its occurrence.

Kamm herself does not present this as a distinct condition, and I’m not sure what she intends the relationship to be between her account of “because” in terms of the role of evil in preventing or resolving second/derivative threats, and this account of “because” in terms of a highly-circumscribed commitment to bringing that evil about. One indication that Kamm sees threat/problem individuation and qualified commitment as independent ways of sustaining her core distinction may be found her remarks on Prize Party, where a harmful effect is required only to achieve a second end, not the first (102). In that case, the agent throws a party intending to have fun and to win a prize for the cleanest house, where the fun would not be reason enough for her to throw the party, and the guest’s feelings of indebtedness are necessary conditions only for the second end. Kamm holds that the agent can pursue her two ends this way without intending to make the guests feel indebted. Even though inducing those feeling is a means to, and not just a condition for achieving, one of her ends – the prize – the agent may be committed to dispelling those feelings before they can induce her guests to do the necessary clean-up. Her willingness or commitment to preventing or undoing the evil she has attempted to cause acquits her of intending that evil.

This second basis for the because/in-order distinction is first made in discussing Party, and then developed by a close application of Bratman’s three-part test for intention – all three of which, Kamm argues, are flunked by the claim that the agent in Loop intends to hit the one on the track. (96-101). The first requirement for intending X is that “we seek means of bringing it about”. Kamm plausibly denies that we seek means of hitting the one — as opposed to saving the five; we merely notice that what we would do to save the five would, without any extra action, also hit the one, and we do it because of that second consequence. The second requirement is that “if one way fails to produce X, we adopt another way”. This is where Kamm develops her position that the agent in Loop and structurally similar cases may be prohibited from doing anything extra to bring about the evil necessary to produce the good, even if declining to do so would result in the failure to achieve that good. She makes this clear with Extra Push, in which the diverted trolley will jump over the one, hitting the five, unless the agent gives it another push. She is not committed to doing this, for it would clearly involve hitting him in order to save the five and intending to hit him. (Kamm distinguishes this from an extra push that causes track vibrations which, by themselves, would prevent the fatal looping toward the five but also cause the trolley to hit the one first. Here, the agent would not give the extra push because it would hit the one; that result would simply be a side-effect. Extra Push is also contrasted with Two Tracks, in which the diverted trolley must be given a further push to get it on the track on which it will hit the one; there, the push is no different than the simple diversion in Loop, just getting the trolley away from the five, but doing this only because the one’s presence will prevent the trolley from hitting them by another route.)

It is in addressing Bratman’s third requirement that the peculiarities of Kamm’s position are highlighted. Bratman holds that if we intend X, then in so far as we are rational, we will filter out intentions that conflict with our intending X. Kamm thinks that the agent in Loop need not reject intentions to do acts that would thwart the achievement of X – the saving of the five. Indeed, she can intend to rescue the man on the tracks from the peril into which her action to save the five places him. If intending the rescue is not consistent with intending the hit, then the agent need not intend to hit the man when she diverts the trolley in Loop. Kamm calls this the Rescue Test, and argues that it can be rational to divert the trolley to save the five while intending to rescue the one as soon as an opportunity presents itself.

Even if the Rescue Test imposed too stringent a requirement on intending, Kamm could still deny that the agent in Loop intended the hit if he did not satisfy Bratman’s first two requirements. But the rejection of the rescue test on this basis may have implications for Kamm’s application of Bratman’s other requirements. Consider, then, the thought that Kamm imputes to the agent, which is supposed to make her intentions coherent:

I will redirect the trolley and then try to push the one out of the way. For if I have an opportunity to save him and do not do so merely because he would then be hit, I would intend his being hit. This I must not do whether by action or omission. If I fail in my rescue efforts, I get the advantage of the five’s being saved because one will be hit. If I succeed in the rescue, the five are no worse off and they had some chance of being saved (for I might have been unsuccessful in the rescue). I still only bother to do something else necessary to save the five (i.e., redirect the trolley) because the one can be hit (98).

This is not circular, because it merely requires the agent to believe that he is not intending to hit the one in diverting the trolley, even though he does so because it will hit the one. The problem is rather that the very course of action she adopts seems a paradigm of practical incoherence. She knowingly imperils the one because it is necessary to save the five, then races off to remove that peril, which she knows will prevent the five from being saved. If the agent’s intentions, as Kamm describes them, do not seem incoherent, then consider some variants that heighten the tension among those intentions. First, imagine that the agent can only turn the trolley by pressing Button 1. Right next to it is Button 2, which sounds a whistle that will give the man on the track plenty of time to step aside. Presumably, the agent should push Button 1, since that will “defeat the defeater”: it will to prevent the saving of the five from being defeated by the looping of the trolley. But she will then have the immediate opportunity to push Button 2, warning the one. If she omits to seize that opportunity, she will, according to her own thinking, be intending to hit the one. Knowing this, she is committed from the outset to pushing Button 2 after pushing Button 1. Can she really intend to save the five by one action if she is already committed to performing another action immediately afterwards that will prevent her first action from having its “intended” effect? The extent to which her intentions are in conflict is brought out by a second variant: she can turn the train by pushing either Button 1 or Button 2. Button 2 will also keep the whistle from sounding for an hour, eliminating the only means of warning the man on the track. Should she press Button 2, to prevent the diversion from being defeated by a course of action to which she is at the very same time committed? Such strategies of self-control make sense against subsequent actions that would be impulsive or akratic, e.g., Ulysses binding himself to the mast to prevent his jumping at the Sirens; I am not sure it does against actions that one fully endorses at the time one takes the precautions.

Even if these variations fail to reach Kamm’s very high threshold for incoherence, other variations suggest that the self-defeating character of the agent’s course of conduct may have a moral dimension. Again, the agent’s controls have only two buttons: Divert and Slow. Pushing divert will have the usual affect, pushing Slow will reduce the speed of the diverted trolley so that it will severely injure but not kill the man on the tracks. It will, however, not stop the trolley, which will gather speed and kill the five. Clearly, Kamm’s agent should push Divert. Almost as clearly, she should then push Slow. If she must do what is possible to prevent injury to the one, surely she must also do what is possible to mitigate the injury if she cannot prevent it. But this commits her to a course of action that will ensure that the man’s injury is in vain. It will not prevent the trolley from killing the five; it will merely ensure that her previous action of pushing Direct does not kill the one. Since she had the option of not pushing any buttons, letting the five die and sparing the one any injury, she thus appears to have injured him gratuitously. NCs are clearly not committed to minimizing aggregate harm, but they should be committed to avoiding harm that is necessary only to allow the agent to act on a set of intentions that allows him to make an idle gesture toward saving some from harm while avoiding the onus of intending harm to others.

So what of Bratman’s second requirement that the agent be willing to try other ways of doing X when his first way doesn’t succeed? (Matthew has pointed out that this may be too strong a condition for intent, since it makes more sense to see a one-shot assassin as intending to kill, with the constraint of using only one shot, than as intending to-kill-with-a single shot. My point here suggests that the one-shot assassin may satisfy a modified version of this condition). Kamm’s agent may not try new means of hitting the man with the trolley once she has diverted it, but if her intention to save the five in sincere, she not only may but must make it as likely as possible that the diversion will succeed before she launches the trolley on a new course. Kamm concedes as much in Two Tracks, in which the agent must divert the trolley a second time from an unoccupied to an occupied loop. Now consider Many Tracks, where the track onto which the train is initially diverted is one of many parallel loops, so close together that a man has no space to stand between him. As she is about to divert the train to the track on which the man is lying, he wakes up and starts running across the tracks. In order to keep the trolley away from the five, Kamm’s agent must ensure that she brings the trolley into contact with the man. She must track him, so to speak, and perhaps even re-divert the trolley repeatedly, to maximize the odds of its hitting him. This course of action would ordinarily be called “aiming”, and it could be used to illustrate what it means to intend to hit a man with a trolley. Since these diversions may continue until the man is hit or missed, there may be no opportunity for rescuing him – the diversionary process may not be completed until the trolley is upon or beyond the man; either way, the duty to rescue may never be triggered. So it’s not clear that Kamm’s agent would fail to satisfy a suitably modified version of Bratman’s second requirement. I am uncertain about the application of Bratman’s first requirement and would be interested in other responses.

In formulating the DTE, Kamm makes the interesting claim that if the agent can justify evil that she does not intend but merely has as a condition of her action, she can justify evil in terms of greater good that she also does not intend but merely treats as a condition of her action (115-118). DTE, formulated in these terms, is very permissive, in the sense that it requires that the agent have only a “because” relationship to the good she adduces to justify the evil she causes, but prohibits her from justifying that evil only if she has a stronger “in-order” relationship to it. My own belief is that the asymmetry goes the other way: that we cannot adduce a greater good to justify evil unless we intended to bring it about that good, but that to the extent there are evils that cannot be justified by the good we intentionally bring about, they cannot be justified either as conditions or intended means (assuming that distinction can even be made) of our actions. I don’t have time to explore this issue in the depth it deserves, but I’d love to hear other views on the subject.


  1. 1. Posted by S. Matthew Liao | July 31, 2007 11:31 pm

    David, thanks for a very helpful analysis of Kamm’s Chapter 4. I have two questions for you, and some remarks regarding Kamm’s analysis of Loop.

    1. You said,

    Or, like me, they may deny that it is permissible to turn the trolley in OTC, and claim that the acceptance of the diversion in OTC rest on the kind of suspect intuition that should be rejected in a robust exercise of reflective equilibrium.

    That’s a very interesting idea. Can you elaborate on the rationale of such a view?

    2. You said,

    Matthew has pointed out that this may be too strong a condition for intent, since it makes more sense to see a one-shot assassin as intending to kill, with the constraint of using only one shot, than as intending to-kill-with-a single shot. My point here suggests that the one-shot assassin may satisfy a modified version of this condition

    and you go on to discuss Many Tracks. I share your view that Many Tracks satisfies Bratman’s second condition. Can you explain though how Many Tracks shows how the one-shot assassin can satisfy Bratman’s second condition?

    Some remarks regarding Kamm’s analysis of Loop:

    I accept that Party may illustrate the distinction between “because of” and “in order to.” But whereas one can tell a plausible story about how one is not throwing a party in order that one’s friends cleanup, I simply don’t see how one can tell a plausible parallel story about how one is not diverting in order to hit the one.

    As you have noted, Kamm tries to tell such a story in two ways. First, she seeks to individuate the threats/problems, arguing that when one initially diverts, one is merely intending to avoid hitting the five. But in Loop, there is simply no reason to divert, except in order to hit the one.

    Secondly, Kamm argues that the agent in Loop is not committed to intending to hit the one, because an agent who is so committed must be willing to give the extra push in Extra Push, and this agent need not be. But as I have noted, an assassin who gives himself only one shot at his target seems nevertheless to be intending to kill his target. If so, showing that the agent in Loop does not seek other means of hitting the one does not show that he does not intend to hit the one.

    So while Kamm may be right that there is a distinction between “because of” and “in order to,” it is not obvious to me that this distinction is applicable in Loop.

  2. 2. Posted by S. Matthew Liao | July 31, 2007 11:39 pm

  3. 3. Posted by David Wasserman | August 1, 2007 12:41 pm

    My misgivings about OTC haven’s ripened into an argument. But I think it’s striking that an intuition that is so widely shared has proven so hard to justify. Thomson’s attempts, particularly the “distribute exemption,” seem completely ad hoc, and Quinn’s attempt to treat both switching and not-switching as “positive agency” runs into the problems raised in Chapter 3. Alex Friedman has a nice paper in review somewhere that points out the equally severe problems faced by other proposed rationales. He argues, as I recall, that while it is not possible to prove that diversion in OTC can’t be satisfactorily justified, there’s considerable grounds for skepticism. The lack of any obvious moral basis for the felt differences in OTC, its variants, and kindred cases, invites the attention of cognitive and evolutionary psychologists, who have no trouble coming up with plausible-sounding accounts of why we are predisposed or wired to react so differently to cases that appear, despite philosophers’ best efforts, to be morally equivalent. I find the strongest ground for skepticism in Peter Unger’s systematic manipulation of our sense of shared risk or vulnerability that, he claims, underlies our different reactions to, e.g., OTC and Fat Man. Unger describes an imaginary adventure park of runaway trolleys, wiggly bridges, and rock slides, in which our moral judgments seem to depend to a disturbing extent on the order in which we are exposed to these lethal devices. (His article was published in Phil. Stud. and appears, abridged, in Living High and Letting Die; Kamm has a response in Utilitas which I haven’t read in a while). I think any credible theory of moral intuitions must acknowledge our susceptibility to amoral influences in a plausible error theory, and should be willing to jettison intuitions that, however widely held and strongly felt, just don’t hold up to scrutiny. I’ll respond to your second comment later.

  4. 4. Posted by David Wasserman | August 1, 2007 5:39 pm

    Matthew, I never got back to my suggestion about your one-shot assassin, but will do so now. I intended to claim that Many Tracks did not pass Bratman’s unmodified second test – if the agent didn’t at first succeed in hitting the one with the diverted trolley, she wouldn’t try, try again. It might seem that, unlike the lone assassin, she would try again, switching tracks until she was able to hit the fleeing man. The problem is that I didn’t make clear what it would mean to try and fail in this setting. I’d treat all her diversions as a single, complex, try – like multiple moves in a pinball game to get the ball in the right place. She only fails when the trolley is past the point where it can be diverted again to hit the man. Past that point, she’s not going to send another trolley his way (unless his evasive action has really infuriated her). Unlike the assassin, who has a well-defined last act – pulling the trigger – her last act is her last diversion before the point-of-no-futher-diversion — a point that will depend on the position of the trolley and the man. Before their respective last acts, both the agent and assassin will make various moves to hit their target. Afterwards, neither will do anything more. It was in this sense that I claimed both would pass a modified version of Bratman’s test. I don’t know if this was worth spelling out – it’s just a footnote.

    The more important issue is raised by your comment on Loop, with which I fully agree. Your one-shot assassin case shows that one needn’t pass Bratman’s second test to intend X, so that the fact that the agent flunks that test in Loop doesn’t preclude her intending to hit the man on the tracks. My point in Many Tracks was only that the agent could show the same kind of circumscribed determination to hit her target as the assassin. Concerning the similarities of Party and Loop, I obviously share your view that it’s far more plausible to see the mess than the looping as a separate threat or “defeater”. I’m just skeptical that there’s a way to make this distinction clearly and apply it consistently to the array of cases Kamm discusses.

    I also think Kamm’s claim that the agent can coherently be committed to immediately preventing the very condition she acts “because of” is as implausible in Party as it is in Loop.

  5. 5. Posted by Guy Kahane | August 1, 2007 8:45 pm

    How do we decide that a strongly entrenched intuition that we fail to justify in a satisfying way is due to error rather that its true ground is just EXTREMELY DIFFICULT to discover?

    Life would be easier for non-consequentialists (NC) if they could deny the permissibility of diversion in the OTC. Of course Unger and most other sceptics think that rejecting a genuine distinction here should point us in the opposite, consequentialist direction (C). But as NC move away from even very robust NC intuitions, I wonder how this would affect the debate with Cs — that is, whether a growing distance from commonsense intuitions would prove to the NC’s disadvantage.

    I don’t yet see how the example of the one-shot assassin is meant to work. That is, you both agree that he intends to kill, with the constraint of doing it with one shot. I think he intends to kill with one shot. If by mistake he fires three, and kills his target, he’ll be disappointed. This is not what he intended to do. And of course there are many way he can fail in this intention and try again. His gun may jam. Or he may forget to remove the safety. So he’ll try again. Under one interpretation of his intention, we can imagine that by mistake he fires a burst of bullets and misses his target. So he tries again, now carefully making sure he only does it with one shot. Though perhaps I’m missing something obvious and barking up the wrong tree. In any case I think some such example can be devised, though it will require a philosophically sophisticated agent who intends to do X WITH ONLY ONE TRY, or something like that.

  6. 6. Posted by David Wasserman | August 2, 2007 1:04 pm

    Guy, your question about when to discard a hard-to-justify intuition is reasonable if it’s not rhetorical. Reflective equilibrium is not an algorithm, and decisions about moral conviction, like decisions about empirical belief, should be made with respect to one’s larger “web” of convictions or beliefs. But I’ll defer to Matthew here, who’s written on the subject. I am interested in your speculation about the consequences for NCs if they depart from their “base” – common-sense intuitions. I suppose there’s now a lot of data on prevailing intuitions about OTC and its variants, from polls and psychological studies. It would be unseemly for NCs to consult those sources like anxious candidates deciding what position to take on a controversial issue, but it would not be demeaning for them to consider the breadth and depth of moral sentiment reflected in that data. Still, like politicians, NCs should lead and not follow their constituents.

    As to the one-shot assassin (I slipped with “lone assassin” – Oswald on the mind), I agree that there’s uncertainty about the best characterization of his intent. I guess it’s a question of the extent to which the “one-shot” constraint should be taken as part of his goal. If his constraint was to hit only his target and no bystanders. I think we’d still say that he did what he intended – as well as other things he intended not to – if he killed his target but also a few bystanders. The one-shot limit does seem to be less of a constraint and more of a goal, but the best characterization may depend on missing details – is he trying to show his mastery as an assassin, give his target a sporting chance, or conserve lead? The first seems more like an integral part of his goal, the last a distinct objective. But I think that in the last case, we would say he intended to hit his target even if he refused to try again out of a preoccupation with scarce bullets – and I think that’s all Matthew needs to make his point about Bratman’s second test not being necessary for intent. As to failures that occur before the “last act” – the firing of the one bullet – the fact that he would try again suggested to me that he would satisfy a modified version of that second test – he’d try again if he failed before his completion of the last act (the last act isn’t basic in this case, so his performance of it requires things not under his immediate control as he squeezes the trigger – like a safety in the “off” position and an clear barrel).

  7. 7. Posted by Jeff Huggins | August 2, 2007 9:36 pm

    David, Guy, and Matthew:
    I find the discussion in this thread very interesting and helpful. In particular, to set a context for my thoughts, I’d like to briefly list three items (a comment from David, a comment from Guy, and a consideration of current responses in The Kamm Poll), which I find particularly useful.

    First, David’s comment, referring to his misgivings about OTC: ‘But I think it’s striking that an intuition that is so widely shared has proven so hard to justify. … The lack of any obvious moral basis for the felt differences in OTC, its variants, and kindred cases, invites the attention of cognitive and evolutionary psychologists, who have no trouble coming up with plausible-sounding accounts of why we are predisposed or wired to react so differently to cases that appear, despite philosophers’ best efforts, to be morally equivalent.’

    Second, Guy’s question/response: ‘How do we decide that a strongly entrenched intuition that we fail to justify in a satisfying way is due to error rather that its true ground is just EXTREMELY DIFFICULT to discover?’

    Third, consider responses to the Munitions Grief Case in ‘The Kamm Poll.’ Of 37 people responding so far (as of the time I wrote this post), 15 have responded ‘yes’, 13 have responded ‘no’, and 9 have responded ‘not sure.’

    In my view, human faculties, as well as the social-moral intuitions they produce or help produce, provide directional, and often rough, guidance to help us ‘successfully’ navigate the social-moral dilemmas we face. (What I mean here by ‘directional’, ‘rough’, ‘guidance’, and ‘successfully’—i.e., to what end?—I won’t go into right now.) Although often deeply rooted, individual intuitions are, to a degree, subject to unstated assumptions, distortions, interferences, leeways, and downright error (or at least foolishness), and they sometimes express themselves in wide variety, which probably increases in most instances (the variety, that is) in the cases involving particularly complex, rare, contrived situations. Of course, understanding human intuitions is one way—one source of important raw material—to help us understand ourselves, i.e., humans. But it’s not the only way, of course. (As participants in this thread might agree, understanding intuitions is probably necessary but certainly not sufficient to answer the key questions of morality/ethics.) To gain a more complete and solidly-grounded understanding of humans, and of human morality, as well as to help us interpret and understand the intuitions themselves, we must also consider the following questions: Why do intuitions (including social-moral intuitions) exist in the first place? What ‘effective’ function do they serve, or help serve, in human life? Is this function sensible, given everything else that we know (today) about life? And, to the degree that we can use our broader knowledge to understand the most foundational ‘effective’ function of human social-moral intuitions, dynamics, and etc., and of morality itself, can we then use that understanding to improve our understanding of the intuitions themselves? To better see what they are there/here (in us) to try to achieve, albeit in their far-from-perfect ways? It is this understanding that can, along with other information, solidly ‘ground’ the moral understanding and ‘principles’ that can be derived (in part) from the intuitions in ways that also enable us to understand when (and why) we might be dealing with intuitive answers that are ‘on track’ as opposed to the variations that are still within the ‘understandable-but-perhaps-not-quite-as-wise’ (in some situations) segment of responses or in the ‘almost-certainly-erroneous-or-at-least-unwise’ segment of responses.

    David’s comment, Guy’s question/response, and the current responses to the Munitions Grief Case poll can all be understood, addressed, and answered by this approach (i.e., with this understanding), I believe.

  8. 8. Posted by Rebecca Roache | August 3, 2007 3:07 pm

    David, I very much enjoyed your careful summary and your introduction of some new thought experiments.

    Whilst reading this chapter I was struck by how difficult some of Kamm’s thought experiments are, in that some of the things she sees as intuitively obvious strike me as controversial or downright implausible. One of these relates to her treatment of Bratman’s requirements for intention, which some of you have been discussing here. I have difficulty imagining how a rational person can both intend to save the five and not intend to hit the one in Loop. Given that the agent intends to save the five, and given that she recognises that the only way that she can do this is by hitting the one, it seems to me that the agent intends to hit the one. Of course, the agent may regret that the only means of saving the five involves hitting the one, and that hitting the one will cause the one to die, but these are both compatible with intending to hit the one (we may intend to achieve a certain end whilst recognising, and regretting, that this cannot be done without bringing about some unwanted side-effects). The Rescue modifications that Kamm introduces to show that the agent need not intend to hit the one, which involve redirecting the trolley towards the one but then trying to rescue the one, are unsatisfactory for two reasons. The first is that, with these modifications, the case becomes, as David mentions, ‘a paradigm of practical incoherence’. The second is that it is difficult to see how someone could act in this way whilst having the intention to save the five, since it seems to me that the agent’s purported intention to save the five fails all three of Bratman’s requirements. That the agent intends all along to sabotage her attempt to save the five means that she fails to seek means to save the five, thereby failing the first requirement; she fails to try again once her sabotaged attempt to save the five fails, thereby failing the second requirement (admittedly this is complicated by the case being constructed to rule out any means of making a second attempt – but, since the agent knows from the beginning that there is only one possible means of saving the five, it seems in keeping with the spirit of Bratman’s second requirement that the agent should not knowingly sabotage her only possible attempt to fulfil an intention!); and by intending to sabotage her attempt to save the five she fails to filter out intentions that conflict with her purported intention to save the five, thereby failing Bratman’s third requirement.

    In short, Kamm’s attempts to show that the agent does not attempt to hit the one come at the price of undermining the original intention that she wanted to explain and justify.

  9. 9. Posted by Rebecca Roache | August 3, 2007 3:44 pm

    ‘Attempt’ in the final paragraph should read ‘intend’. Sorry about that.

  10. 10. Posted by Guy Kahane | August 4, 2007 3:26 pm

    David, my question of criteria for intuitive error was not rhetorical, and although it has been in the background (and often the foreground) of the discussion in this reading group, it is especially pertinent to this and the next chapter. Familiarly, some philosophers have very little patience for following intuitions through. They lose their patience before Kamm even gets started. Implicit in Kamm’s whole work is the idea that ethics is a bit like mathematics. There is a truth out there and it is often extremely difficult to reach it, and we can do so through ever more complex constructions. My question was: if (let us suppose) the best Kamm could come up with to draw a strongly intuitive NC distinction isn’t good enough, then should we conclude that there’s really nothing there or, rather, as I’m sure Kamm would respond, try even harder? Whether we can come up with some causal explanation for the source of intuition is a separate matter — we can come up with such explanations to all moral intuitions. So it seems we need some independent sense of how difficult ethics should (or could) be. For example, you might think that given that a supposed moral distinction is a very basic one, effortlessly appealing to virtually all of us, then if it has some genuine principled grounding, this grounding must itself be pretty simple. If it’s very implausible that we are tracking the properties and principles that Kamm is appealing to, then this counts against Kamm. It seems to me that one general lesson of Kamm’s work is that the widely accepted view that meta-ethics and normative ethics are distinct enterprises is ultimately very hard to sustain.

    To turn now to my remark about the broader debate between NC and C, it is certainly true that in characterising these two views we needn’t make any reference to commonsense morality or people’s intuitions. Historically, many of Kant’s claims are highly counter-intuitive. And we can imagine, for example, a society were act utilitarianism is the ‘commonsense morality’, challenged by a tiny minority who believe in counter-intuitive side-constraints. Still, as a matter of fact C has almost always been a partly sceptical view, and much of modern NC an attempt to defend commonsense beliefs and intuitions. I asked whether, if NC find themselves retreating from these commonsensical intuitions, this would affect the shape of the larger debate. Intuitively, I think it is easier to retreat from commonsense in the C direction rather than in the opposite one. Notice that although Kant was not always friendly to commonsense, it is pretty easy to imagine various forms of extreme counterintuitive NC views — the view for example according to which we are NEVER allowed to harm other rational beings. But it’s not an accident that nobody holds or defends such views. And on the face of it, it seems that if the debate is conducted between two theories that are both distant from commonsense morality, then the Cs theory would have more of the virtues we look for in a theory — simplicity and economy, for example. Of course, I’m not saying that all of this follows from giving up one intuitive distinction.

  11. 11. Posted by Jeff Huggins | August 5, 2007 12:14 am

    Guy and David,

    The discussion on this thread is very interesting and helpful, and Guy’s recent post makes some great observations and poses and/or implies some great questions.

    I offer some thoughts that seem, to me, to apply to the present subject (of highly intricate cases and highly intricate principles), beginning with one of Aristotle’s observations from his Ethics, in which he writes: “… for it is a mark of the trained mind never to expect more precision in the treatment of any subject than the nature of that subject permits …”.

    In my view, intuitions, science, logic, and other means or sources should all be used, in conjunction with each other, to illuminate our understanding of morality/ethics and our pursuit (or development) of good principles, ideals, and dynamics. And, they should all be applied to different avenues of the pursuit, and these avenues should inform each other, frequently.

    By “avenues of the pursuit”, I mean the following, among others:

    • Exploring intuitive responses to cases and, using intuition and logic (and sometimes semantics), trying to derive basic considerations and ethical principles from them. (We all seem to note that pursuing this avenue for great distances, as Kamm does, can begin to require increasingly intricate cases, leading (often) to highly diverse sets of intuitive responses, requiring increasingly intricate logic to derive increasingly intricate ethical principles.)

    • Understanding—through intuition, science, logic, and other means—the foundational function of human social-moral dynamics in human life. As I think I mentioned in an earlier post, trying to understand a fish (for example), solely or predominantly through microscopic examination, is difficult (and can easily lead one astray) if one doesn’t first know (and/or frequently remember) that a fish is an aquatic organism that swims (rather than, for example, a machine for digging ditches, or a radio, or etc.). Thus, in my view, it helps if one constantly reflects back on the question, “What is morality for?” (I use the words loosely here just to illustrate the current point.)

    • Frequently “tying back” to what is obviously important to humans in life as well as to human life itself. For example, some degree of perceived “control”, “security”, and “stability” are necessary for healthy human life, sanity, and daily functioning, even if these are somewhat fleeting or even imagined. Thus, referring to OTC and variations, most people can imagine and (to a degree) understand that, if they are standing on a track, and if someone else diverts a train to save five other people, they might get killed if they don’t get off the track in time. But, most people probably can’t as easily, intuitively, or comfortably imagine a world in which they might be walking anywhere (on a pedestrian bridge, or on a sidewalk, or in a park, or etc.) and be suddenly pushed by a complete stranger in front of something to stop a threat they had nothing to do with. Nor can people easily and comfortably imagine living in a world where one might go into a hospital with a sore throat and find that her or his five most vital organs are being taken out to save five other people. Although the world we live in is far from perfect, most people certainly do not want to live in a world where people in general think it is fine (or even good?) to push each other in front of trains, to steal vital organs, or so forth. In my view, people don’t want to contribute to the creation of such a world, not because there is an intricate principle against or for it, but because they want to feel at least some degree of “control” and “security” and they don’t want others to do such things to them. Call it the human social reciprocity principle (or the Golden Rule), driven (at least in part) by our “mirror neurons” and our other intuitive mechanisms. The two latter notions (i.e., being pushed into death, or having vital organs stolen) threaten, intuitively and/or consciously, the very sense of “control” and “security” a human being must have in order to go outside and function each day.

    • Exploring and understanding (via intuition, observation, science, etc.) how humans work, e.g., what we pay attention to, and how we pay attention to it; our (human) limits; our inner mechanisms; and so forth.

    • Exploring and understanding how our human social-moral faculties came about, and how our most basic (and universal) human social-moral dynamics came about.

    Information (and deep understanding) from each of these avenues of pursuit can help us understand the interrelationships between their subject matter, can help us better interpret the results from each, and can help us understand when one avenue has been largely exhausted (e.g., “exhausted” in terms of the potential rewards of pushing ever deeper into more complexities and corners of that particular avenue; there may be other senses in which a particular avenue is not exhausted, for example, with intuitions, rather than going into ever-more-complex cases and trying to derive ever-more-complex principles, one could (as is being done by some scientists) explore how the human brain-mind works internally with cases involving various factors and various degrees of complexity).

    Guy writes, “My question was: if (let us suppose) the best Kamm could come up with to draw a strongly intuitive NC distinction isn’t good enough, then should we conclude that there’s really nothing there or, rather, as I’m sure Kamm would respond, try even harder?” My answer to this question is that we should do a “cross-check” with our own understanding, and the understanding of others, associated with the other avenues of pursuit. For example, if intuitive responses to highly intricate cases are highly mixed (or, in some cases, even if they aren’t), and if we are having trouble understanding or explaining these responses without having to resort to highly intricate principles, it is possible (is it not?) that we might be ignoring or forgetting about the basic function of human morality/ethics, the ways that humans work, and/or basic human emotional and cognitive limits?

    Regarding Guy’s “question of criteria for intuitive error”, in my view, the best way (and perhaps the only way) to tell whether a specific derived intricate principle is inaccurate or whether a specific set of intuitive responses are off-base (in cases where the principle and the responses in question are in conflict) is to cross-check with understanding from the other avenues of pursuit and with the grounded understanding that comes from the integrated picture-in-progress. (This assumes, of course, that the principle has already been tested against a large sample of responses that statistically seem to support it, even though a minority of responses will typically be in disagreement.)

    My impression, as I read the ever-more-intricate cases and principles, is that this line of inquiry may be missing out on the cross-talk that could be taking place between/among the differing avenues of pursuit and, thus, missing out on the degree of light that each of these avenues could shed on and for the others.

    Guy writes, “So it seems we need some independent sense of how difficult ethics should (or could) be.” For the most part, my earlier comments apply to this. (More importantly of course, Aristotle’s observation also applies.) That said, even if a view of ethics/morality that is thought to be highly logical and true to the “ends” of morality ends up being substantially more complicated than most (99%+) human minds can even begin to comprehend, what then? I think that the answer to the question of how difficult ethics could/should be will largely come from the second and fourth “avenues of pursuit” in the list included above, that is, from an understanding of the foundational “effective” function of morality, from an understanding of how humans “work” (that is, emotionally, cognitively, and biologically speaking), and from an understanding of human limits.

    Guy writes, “For example, you might think that given that a supposed moral distinction is a very basic one, effortlessly appealing to virtually all of us, then if it has some genuine principled grounding, this grounding must itself be pretty simple.” If I understand this comment correctly, I agree with it, at least in one sense. The deep grounding to all of this (i.e., to the subject and dynamics of human morality) must itself be very simple, or at least is most likely to be fairly simple, and I believe it is fairly simple. That said, understanding only seems simple, as we all know, after it is understood, not before. For example (probably an overused example, sorry!), the Ptolemaic view of the universe began as a fairly simple Earth-centric model. Over time, in order to “maintain” this model (in at least two senses of the word “maintain”) and keep it up-to-date with more accurate observations of celestial bodies, it was complicated with many contrivances, add-ons, and etc. One could say that it was complicated with many intricate contrivances. The contrivances were necessary mainly because the underlying model was incorrect. Copernicus, Galileo, and Kepler gave us a more accurate model, at least of the solar system. The need for the contrivances (at least of the sort involved in the Ptolemaic system) evaporated. What at one early point in time seemed simple (the early Earth-centric Ptolemaic system), and what would have seemed complex to most people at that time (a solar system described by mathematical principles), have reversed their positions on the scale of simplicity-complexity.

    Of course, I’m not at all suggesting that intuitions don’t matter. They matter a lot. And, we can learn a lot from them. I’m simply suggesting that, as the cases become ever-more intricate and “rare” (and as they stretch our human ability even to imagine ourselves facing them), and as the posited principles become highly intricate, my view is that we may be pursuing this avenue of discovery beyond the point (at least in terms of intricacy) were it is most fruitful and even to a point where it risks becoming inconsistent with valid and robust learning coming from the other avenues.

    Sorry for the long post. I hope at least some of it is helpful or thought-provoking.

Post a comment

Name: (required)

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