Harming People in Peter Unger’s Living High and Letting Die
Harming People in Peter Unger’s Living High and Letting Die
Abstract and Keywords
Peter Unger has tried to show that relying on intuitive judgments is a worthless methodology for finding principles, and he has also offered a novel approach to the Trolley Problem. Unger, however, deals not only with the questions of when may we harm some to help others and how we can best reason about this issue. He also considers how much we must sacrifice in order to stop strangers from suffering serious losses and whether our distance from them alters our obligations. Unger's ground for claiming that intuitive judgments in cases are worthless is that we can construct cases that generate the opposite intuitive judgments. He thinks that we must decide which intuitive judgments are correct and what to do by consulting general moral values, such as the importance of reducing suffering and death. This chapter examines Peter Unger's views on the permissibility of harming innocent bystanders and the duty to harm ourselves in order to aid others.
In previous chapters we have made use of intuitive judgments about cases to find moral principles and fundamental factors underlying them. We have especially focused on variations of the Trolley Case. Peter Unger has tried to show that relying on intuitive judgments in cases is a worthless methodology for finding principles, and he has also offered a novel approach to Trolley cases.1 Given how his views bear on topics with which we have dealt, it is appropriate to consider them in some detail.
Unger, however, deals not only with the questions of when may we harm some to help others and how can we best reason about this issue. He also considers how much we must sacrifice in order to stop strangers from suffering serious losses and whether our distance from them alters our obligations. Substantively, Unger aims to prove the following four claims in the following order: (1) We must, quite generally, suffer great losses of property to prevent suffering and death; (2) we may, quite generally, impose such losses on others for the same goals; (3) we may, quite generally, kill others to prevent more deaths; and (4) we must, quite generally, kill ourselves to prevent more deaths.
Methodologically, Unger aims to show that intuitive judgments about cases that would be presented as evidence against his four substantive claims—a standard technique that nonconsequentialists employ when arguing against consequentialists—are (p.191) worthless. He also aims to show that many distinctions presupposed by these intuitive arguments in cases, distinctions to which some nonconsequentialists attribute moral significance, do not matter morally and cannot be used to rebut his (or any) substantive claims. The distinctions whose moral significance he rejects include the distinction between harming someone so that he suffers mortally and not-aiding him when he will suffer mortally, and the distinction between harming someone by redirecting a threat toward him and harming him by using him to stop a threat.
Unger’s ground for claiming that intuitive judgments in cases are worthless is that we can construct cases that generate the opposite intuitive judgments. He also thinks the above-mentioned distinctions are useless because we can show that the factors that distinguish the cases where the distinctions seem to matter are not morally significant. He thinks, instead, that we must decide which intuitive judgments are correct and what to do by consulting general moral values, such as the importance of reducing suffering and death.
II. Unger’s Ethical Method
So far, I have provided only the briefest description of Unger’s ethical method. Considered in more detail, we see that Unger actually first proposes that general reflection (rather than reflection on cases) reveals our basic primary values, and these values leave no room for the importance of many distinctions between cases relative to the aim of preventing mortal loss. He next tries to show that reflection on particular cases, that supports the importance of nonconsequentialist distinctions, is subject to negative distortion. That is, this case-specific reflection is inaccurate because it veers us away (negatively) from the basic values revealed by general reflection. He then supports the claim that our belief in constraints on harming others is often based on such negative distorting tendencies by constructing cases so as to change the intuitive judgments we have concerning the performance of particular acts. Constructing these new cases will make the responses be more in line with basic values that deny the significance of the distinctions at issue. In these revised cases, distortion is positive, that is, it leads to judgments about cases that agree with, and hence reveal, basic values. (An example of a positive distorting factor, Unger believes, is salience. It is distorting because it is really morally irrelevant how salient an event is; salience should not affect our judgment about a case. However, salience can lead us to have intuitions about cases more in keeping with our basic values, he thinks.)2
In sum, we may say that whereas many nonconsequentialists fashion principles to match intuitions, Unger’s aim is to show that we can fashion cases and intuitions about them to match principles.3 Doing this also, according to Unger, involves showing that our intuitions about cases are inconsistent. For Unger, this means that in one case we approve of a constraint on harming; in another case, we disapprove of the very same constraint.4
Next, Unger asks us to reflect on the changes he has introduced into cases to alter our intuitions about the permissibility of acts and to see that these changes (p.192) involve factors that are not in themselves morally significant. Hence, the same acts in cases lacking these factors merit the same response, even if intuitively they do not prompt the same response.5 The upshot of Unger’s procedure is supposed to be that when there is a conflict between the theses supported by general reflection (e.g., reduce suffering) and judgments about particular cases, we should stick with the results of general reflection, for our intuitions about cases are unreliable and manipulatable by morally irrelevant factors.
Unger thus presents an error theory of nonconsequentialist intuitions in cases and of nonconsequentialist restrictions on both harming others and prerogatives not to make large sacrifices to aid. This error theory is based on the psychological effects of morally insignificant factors. Hence, he thinks, the nonconsequentialist tactic of presenting case judgments that conflict with the principle of minimizing mortal loss cannot be used to defeat the obligation to minimize mortal loss. For example, Unger suggests that one of our primary values is that people’s suffering and dying is bad, and more of this suffering is worse than less. Yet, case intuitions sometimes show that we think that we have no obligation to cut off our arm to prevent the loss of another’s life. When there is a conflict between intuitions in cases and the primary value, we should stick to our primary value.
Possibly, however, Unger is not consistent in abiding by this recommendation. He considers a case in which, while we are visiting a poor distant country, we meet someone whose leg we can save. He says, “As we intuitively react,” it would be “morally outrageous”6 not to help this person on the grounds that if we save his leg he will reproduce at a greater rate, which will lead to an overall increase in human suffering in that poor land. This intuitive moral outrage conflicts with the general value of reducing overall suffering and death, for future suffering and death is part of overall suffering and death. Does Unger think that it is permissible to side with the intuition that we must help the person in this case rather than with the general value?
I shall assume that Unger would be willing to revise his judgment about this case and side with the general value, and thus leave the man by the road. If he would not, he risks allowing cases to drive the formulation, acceptance, and rejection of normative principles in the way they have for some nonconsequentialists. This would be contrary to his stated view that rather than preserving our original intuitions about cases—a hopeless task anyway, if they are inconsistent—we are (correctly) liberated from them as primary guides to moral truth.
Unger’s ultimate thesis is that we should make moral decisions using only primary values. But it must be emphasized that this does not mean that in proving this thesis his method ignores case judgments entirely. In his attempt to prove that we should not rely on intuitive case judgments in deciding what is morally right, Unger himself relies on intuitive judgments about cases (that yield competing intuitions). It is important to his project to show us that our responses to cases will change in the direction of what he thinks are primary values, if he alters cases in certain ways that are morally irrelevant. If our intuitions about cases do not in fact change as he says they will, his arguments against relying on case judgments in deciding what to do will not succeed. Likewise, his arguments against the existence (p.193) of nonconsequentialist constraints on preventing mortal loss will not succeed and his account of the (supposedly) morally unimportant factors prompting nonconsequentialist intuitions will fail. (It is when he asks us to consider whether the alterations he makes in cases that [supposedly] change our intuitions could possibly be morally important in themselves that he relies on reflection rather than on intuitions in cases.)
Of course, it is important that Unger, who believes that intuitions about cases are subject to both positive (i.e., value-tracking) and negative (i.e., value-veering) distortions, have some method besides judgments in cases for identifying correct values. For without knowing what the correct values are, we will not know whether what he thinks are negatively distorted cases really are taking us in the wrong direction. As noted, his method for finding the correct values is that of reflecting directly on general claims, such as the claim that mortal loss should be minimized. This too seems like an intuitive judgment, even if not a case-based one.
I do not believe that Unger is successful in proving either his substantive or his methodological claims as I have described them. In this chapter, I shall critically examine his substantive move from the permissibility of physically harming some people in order to save others in some cases to the claim that, quite generally, we have a duty to physically harm ourselves to save others. His methodological claims will be evaluated as well. Along the way, I will re-present the alternative nonconsequentialist account of some of the cases that Unger discusses, which I presented in chapter 5. However, the correctness of this alternative account is not essential to my critical argument. In conclusion, I shall discuss his project of reconciling his views with the ordinary moral judgments with which his views differ.
III. Property Losses and the Duty to Aid
Before proceeding to these issues, however, I will briefly summarize what I have elsewhere said about Unger’s claims concerning the duty to suffer losses of property (rather than physical harms) and the permissibility of imposing property losses (rather than physical harms) on others, in order to save people.7 This is to provide the necessary background for his discussion of physical harming.
Unger is eager to show that we must suffer great losses in property in order to help distant starving people. He contrasts the ordinary intuitive judgment that we should suffer a great deal in order to aid someone who is in an accident near us with our reluctance to condemn someone who does not do much for international famine relief. He thinks that it is salience—the degree to which suffering imposes itself on us to the point where we psychologically cannot ignore it—and a sense that we can take care of a problem entirely in the near case, but not in the distant case that underlie these different judgments. But both of these factors, he thinks, are morally insignificant and cannot justify our different judgments in this set of cases. Hence, in deciding what to do, he thinks that we should rely on the general value that suffering is to be reduced.
(p.194) To test his analysis, I suggest that we consider a case in which a distant person is the victim of an accident, and the problem can completely be dealt with by our aid. Also imagine that this person’s plight is salient because we have long-distance vision. I discuss a distant accident rather than distant starvation, as Unger does, in order to hold constant all factors when comparing a near accident case with helping distant people. This is in addition to holding constant salience and completely dealing with a problem to which Unger points in explaining the different judgments in near and distant cases. I think that our intuitions about some distant aid case can still be different from our intuitions about some case where the victim is near when we hold all these factors constant. This difference may show up, however, only when the cost of aiding is equally high (rather than equally low) in both near and distant cases. We have a duty to aid only the near person at high cost. (For a factor like distance to matter morally, it need only make a difference sometimes, not always.) This suggests that proximity per se plays a role in generating our intuition that we should suffer a large property loss in order to aid. If so, more needs to be said both about the way in which proximity affects our intuitions and whether we can justify its having a role in generating our intuitions and in morality.
With respect to the first point about the way in which proximity affects our intuitions, I do not think that even if proximity can be morally important, this necessarily means that we always have a weaker obligation to aid at high cost a person who is distant from us than to aid one who is near us. Indeed, though it sounds paradoxical, it is because nearness matters that we may have a strong duty to help a distant person. For example, suppose that my boat is near a drowning person from whom I am distant. I may be obligated to let my boat be used to help him because it is near him though I am not. This case bears on another of Unger’s claims: that we may, in general, impose property losses on others in order to help people. Intuitively, he claims, we think that it is morally permissible to take someone’s boat to save a drowning person, even if this will cause large, uncompensable damage to the boat, but we also think that it is not permissible to steal money from someone’s bank account to help distant people. He does not think that there is any factor of moral significance that explains this difference, and hence we may impose property losses of any sort on others generally in order to provide aid. But I have suggested that if someone’s property (such as his boat) is near a victim, she will intuitively be thought to have an obligation to let it be used to save the victim and that she lacks this obligation if her property (e.g., money in a bank account) is not near the victim. This suggests that distance is thought to be of moral importance.
Suppose we can show how proximity is generating our intuitions. The task of showing that we can justify its having this role because proximity is of real moral significance remains. Notice, however, that (contrary to what Unger thinks) if there is a duty to help because of nearness and sometimes no duty to help in the absence of nearness, this does not necessarily mean that we must do our duty to the near person rather than help the distant one. For very often, supererogatory acts—that is, aid we are not duty-bound to give—may be done instead of what is strictly our duty.8
A. Unger’s Thesis
Unger moves beyond the discussion of imposing losses of property for the sake of saving life and argues for duties to physically harm oneself and others. By using intuitions about cases in which someone is not simply giving up his life to save others, Unger arrives by steps at his radical conclusion, that I have a duty to give up my life in order to save the lives of two strangers. That is, Unger does not claim that we have direct intuitions about cases to the effect that I have a duty to give up my life to save two strangers. This is so even when their need is salient and my loss is going to completely solve their immediate problem.
The case Unger presents in which our intuitions tell us that we should kill another person to save a greater number of people is the Trolley Case, in which a trolley is headed toward killing (in his version) six people, and we may redirect it, though we foresee that it will certainly kill one. Indeed, it would not be wrong to summarize a great deal of what I shall explain in more detail below by saying that Unger attempts to use our intuitions about cases with the redirection-of-threat structure to deduce the permissibility of sacrificing people in general in order to diminish overall mortal loss. He does this by arguing that what the redirection-of-threat structure does, and what in large part accounts for our intuition that it is permissible to kill some people in this sort of case, is to group people and so overcome our tendency to separate people. Hence, whenever we come to group people (even because of a different structure), we will think (correctly, he thinks) that it is permissible to kill some people to save a greater number of others. But when we reflect on how grouping arises, he says, we see that grouping is not really morally crucial per se, and so even in cases where there is no obvious grouping, it should be permissible to kill some to save others. Hence, part of Unger’s strategy is to rely on our intuition that it is permissible to redirect a threat away from killing six, even though it will then kill someone else, and show that there is no morally significant difference between this case and others where we would have to kill one to save others, for example, by pushing someone into the threat to stop it. This should leave it open for us to decide to kill based on our reflection on general values.
The traditional Trolley Case is a two-options case: Either we let the trolley kill the six, or we redirect it and it kills one. In being a two-options case, it is like another case where we either let the trolley kill six, or we push someone in front of the trolley to stop it. In discussing the traditional two-options redirection Trolley Case, Unger claims that we think of the one and the six as grouped (he calls this “projective grouping”) because they share a connected set of trolley tracks. But in discussing other cases9—cases that, I believe, also have a redirective structure, though Unger does not explicitly draw attention to this fact—he points to other factors that produce grouping. These other factors are that the same threat will affect either the one or the six and that the threat going toward the one person is the noncausal flip side of the threat ceasing to go toward the six.10 Unger thinks that these (p.196) factors have a role in our thinking because they group the one and six, that is, we think that the one is involved with the six, shares their problems, and hence is “fair game” to be harmed so that the six will survive. This contrasts with our tendency to “projectively separate” someone from the problems that other people are facing, leading us to think it is wrong to involve her in order to save the others.
In addition to grouping versus separating, Unger argues that we respond to what he calls the “protophysical” characteristics of a case, that is, we find it easier to redirect an object already in motion rather than start one up; we find it easier to slow an object down rather than speed it up; and so on.11 These factors will also affect our willingness to harm someone in order to help others.
If we think of pushing a stationary person in front of the trolley to stop it from hitting the six, we are faced with putting an entity into motion, producing harm to someone who does not share trolley tracks with the six, and producing harm to someone when this is not the flip side of the threat to the six ceasing. Some of these factors are important, according to Unger, because they lead us to see the one person as separate and not involved in the six’s problem. Unger thinks that projectively separating someone involves treating people unequally,12 because it gives the one person’s claim not to be harmed more weight than is given to the comparable claim of the six not to be harmed by the trolley.
Unger thinks that we can overcome projective separating and so come to find it intuitively permissible to push one person in front of the trolley because we group him with the six, by constructing cases that involve several options (i.e., more than two).13 So he constructs the Switches and Skates Case, described as follows:
By sheer accident, an empty trolley, nobody aboard, is starting to roll down a certain track. Now, if you do nothing about the situation, your first option, then, in a couple of minutes, it will run over and kill six innocents who, through no fault of their own, are trapped down the line. (So, on your first option, you’ll let the six die.) Regarding their plight, you have three other options: On your second option, if you push a remote control button, you’ll change the position of a switch-track, switch A, and, before it gets to the six, the trolley will go onto another line, on the left-hand side of switch A’s fork. On that line, three other innocents are trapped and, if you change switch A, the trolley will roll over them. (So, on your second option, you’ll save six lives and you’ll take three.) On your third option, you’ll flip a remote control toggle and change the position of another switch, switch B. Then, a very light trolley that’s rolling along another track, the Feed Track, will shift onto B’s lower fork. As two pretty heavy people are trapped in this light trolley, after going down this lower fork the vehicle won’t only collide with the onrushing empty trolley, but, owing to the combined weight of its unwilling passengers, the collision will derail the first trolley and both trolleys will go into an uninhabited area. Still, the two trapped passengers will die in the collision. On the other hand, if you don’t change switch B, the lightweight trolley will go along B’s upper fork and, then, it will bypass the empty trolley, and its two (p.197) passengers won’t die soon. (So, on your third option, you’ll save six lives and you’ll take two.) Finally, you have a fourth option: Further up the track, near where the trolley’s starting to move, there’s a path crossing the main track and, on it, there’s a very heavy man on roller skates. If you turn a remote control dial, you’ll start up the skates, you’ll send him in front of the trolley, and he’ll be a trolley-stopper. But the man will be crushed to death by the trolley he then stops. (So, on your fourth option, you’ll save six lives and you’ll take one.) On reflection, you choose this fourth option and, in consequence, the six are prevented from dying.14
In this case, Unger thinks that we will find it permissible to push the one person on roller skates in front of the trolley, even though if it were the only alternative to letting the trolley kill the six, we would find it impermissible. But we should, he thinks, reflect on the psychologically powerful features that differentiate pushing someone in front of the trolley in the Switches and Skates Case from pushing someone in front of the trolley when it is the only option other than letting it kill six people. These features are not really morally important, according to Unger. Hence, we are free to use general moral values to decide what to do and should conclude (by these steps, if not by direct intuition) that it is also permissible to push someone in front of the trolley in the two-options Trolley Case.
Why does Unger think that we will intuitively find pushing the one person in front of the trolley permissible in the Switches and Skates Case? Essentially, he thinks that we will have the intuitive judgment because the intervening possible courses of action serve as bridges between what we intuitively find permissible (redirecting the trolley toward the three) and what is initially thought impermissible (pushing the person on skates in front of the trolley). The bridging is supposed to occur in the following way. Redirecting the second trolley (which is already in motion) in the Switches and Skates Case has many of the properties of redirecting the first threatening trolley, so we will think that it is as permissible as redirecting the original trolley threat. And pushing the person on skates in front of the trolley has some of the same characteristics as redirecting the second trolley (where it is the weight of the two people in it that stops the first trolley), so we will think that it, too, is permissible. More specifically, he says, like the six and the three in the first redirection option, the people in the second trolley are on trolley tracks; they are in a trolley that is already moving; and all we must do is redirect the second trolley. And so, he thinks, we will find it permissible to redirect the second trolley with its heavy people into the first trolley threat. The single person on roller skates, like the people on the second trolley, will be in motion once we start him and, like them, he is on wheels (albeit, roller-skate wheels), and it is his presence that causes the trolley to stop. Even though we do have to start the skater up (rather than redirect him) and he is not on any tracks, Unger thinks that the similarities will make us find it permissible to send him in.
Abstractly, the model is that option A has properties abc, option B has properties bcd, and option C has properties cde. Unger thinks that because B is in some ways like A, and A is permissible, we will think that B is permissible, and (p.198) because C is in some ways like B, we will think that C is permissible, too. Hence, C will be permissible, if A is. He thinks that his reasoning holds, even though the resemblance between A and C, when considered directly, is minimal.
Given that, according to Unger, this “bridging resemblance” (my phrase) is all that makes us think that it is permissible to reduce the number of people killed by pushing the one person in front of the trolley, we cannot think that it is only when the bridging resemblance is present that it is permissible to send in the one person. That is, there is nothing intrinsically morally important about this bridging resemblance, except that it helps to overcome inhibitions in a positive direction (given that it leads us to minimize mortal loss, as general reflection [according to Unger] says is correct). For example, there is nothing intrinsically morally important about being on wheels rather than not being on wheels when you are pushed in front of a trolley. Thus, Unger concludes, it is, in general, permissible to push one person into a threat in order to save a greater number.
Unger’s theses can be summarized as follows: (1) Our intuitions will be different in the Switches and Skates Case from our intuitions in the traditional two-options Trolley Case with respect to pushing someone in front of the trolley; we think that it is permissible to push the one person in front of the trolley in Switches and Skates; (2) it is projective grouping (via bridging) of the one person with the six that leads us to think that it is permissible to do this to the one in order to help minimize mortal loss to the six; (3) projective separating leads us to believe that we may not involve someone in this way; and (4) if it is permissible to act when people are seen as grouped, it is permissible to act in the same way when they are seen as separate.
Let me first summarize briefly my view about each of these four theses of Unger’s: (1) My intuitions about the impermissibility of pushing someone in front of the trolley are not changed by the introduction of several options, as in the Switches and Skates Case. Intuitively, I do not think that it would be permissible to direct the second trolley so that it stops the first, killing two people on board, if this were the only alternative to redirecting the original threat to three other people.15 (2)(a) Unger employs a highly contentious notion of grouping, which he is right to think has no intrinsic moral significance, and so it should not lead us to change our intuitions about what it is permissible to do. (2)(b) It is not true that people who are grouped, in his sense of grouped, are intuitively thought to be “fair game” to have anything done to them in order to minimize mortal loss. Hence, such grouping is not sufficient to account for the permissibility of harming when harming is intuitively permissible. (3)(a) He employs a contentious notion of “separateness,” and (3)(b) it is not true that we think that it is impermissible to do harmful things to people who are separate in his sense. Hence, grouping in his sense is not necessary to account for the permissibility of harming when it is intuitively permissible. (4) Sometimes, contexts have intrinsic moral significance, and so we cannot draw implications from them about what to do in other contexts. But since the several-options context does not change my intuitions, I think that this point is moot. (p.199) (Notice, that I have made reference to “my intuitions” rather than made predictions about what others’ intuitions will be. This is because I believe that it is primarily through an individual generating her own intuitive judgments and then trying to see what factors account for them and might justify them that we can make progress by a method employing intuitive judgments.)
Let me expand on these points (though not in the same order in which I have presented them). I do not find that my intuition changes about the impermissibility of pushing someone in front of a trolley, whether it is one of two options or one of several options. Indeed (as already noted), I do not even intuitively think that it is permissible to send in the second trolley, when the people on it are ordinary passengers, to crash into the first heavy trolley. Hence, I cannot be accused of inconsistency in thinking that it is wrong to send in a person if it is an option in a two-options case but permissible to send him in when considering a severaloptions case.16
I believe that my intuitions do not change in the Switches and Skates Case because the sort of bridging resemblance that Unger creates between the options in the Switches and Skates Case (and in other several-options cases he discusses) does not ensure that the characteristics that truly account for the permissibility of redirecting the trolley threat are present in the other options. This implies that grouping, at least of the sort created by bridging in the several-options case, is not the characteristic that accounts for the permissibility of redirecting the trolley threat originally. Further, the fact that option B resembles option A, and option C resembles option B, does not mean that they resemble each other in sharing the property present in A that makes it permissible to redirect the trolley. Suppose that A has properties abc and so is similar to B, which has bcd, and C is similar to B in having cde. This is no indication that C will be permissible, if A is permissible in virtue of a. This grouping, if it exists, seems no more explanatory than saying that because all people involved in redirection of the trolley threat are wearing red, we can show that it is permissible to push someone wearing pink in front of the trolley because we can show that it is permissible to redirect the second trolley, whose occupants are wearing half red and half pink. This is part of what I mean when I claim (in my response [a]) that Unger employs a contentious notion of how people get grouped.
But there is more to be said about the notion of grouping. If Joe is close to Jim, who is close to Tim, who is close to Tom, then Joe and Tom may be part of a group, in the sense of “not separated by much space,” whereas if there were no individuals between them, there would be no group—they would be separated by a significant amount of space. And their being part of a group, in the sense of not being separated by much space, might sometimes have significance for what it is permissible to do to them. This is one way in which we can group people. But this is not Unger’s sense of grouping. Rather, as I have said, he thinks we group the person on skates with the six in the Switches and Skates Case in virtue of a linkage of similar properties with intervening options. But, if act A is permissible because of some property it does not share with B or C, other similarities between A and B, and B and C do not imply that we may do act C because we may do act A.
(p.200) Indeed, Unger’s understanding of how to make redirecting a trolley and pushing a person in front of a trolley seem equally permissible has something like a sorites structure:17 Creating a series with (supposedly) slight differences between each member of the series gets us to equate the two extremes of the series, which we know to be distinguishable. Unlike a true sorites, however, a morally significant difference between the first act and the others should stop the illegitimate equivalence in its tracks, as evidenced by the intuitive rejection of the permissibility of sending in the second trolley (as well as the man on roller skates).
I also claim (in response to ) that Unger employs a contentious notion of “separation” in projective separating. I think that his concern with grouping and separating is a sort of response to the objection raised to utilitarianism: that it does not take seriously the “separateness of persons.”18 The point of this objection is to emphasize that when one person is harmed to benefit others, the benefits to the others do not compensate the person harmed. This is unlike the case of intra personal harm for benefit, where the person harmed is himself compensated by the benefit. The failure of interpersonal compensation is still true of people who are grouped (in any of several senses of “grouped”); the loss to the ones who die is not compensated for by any good to them. Hence, we do not eliminate the separateness of persons, in the sense in which this is the basis of objecting to utilitarianism, by grouping them.
There are at least two other senses of “separate persons.” First, persons are separate when they are not in a group linked spatially by other persons. Second, persons are separate when they are not grouped together because they do not share the same or similar properties. The latter is the sense of “separate” upon which Unger focuses, corresponding (by denial) to the sense of “grouped” upon which he focuses. (Call these last notions of separate and grouped “Unger-separate” and “Unger-grouped.”) Thus, we can put Unger’s theses (2) and (3) (summarized above) as the view that when people are Unger-separate, the fact that one person would not be compensated for harm done to him in reducing mortal loss to others will be thought morally important enough to make such harming seem impermissible, but when people are Unger-grouped, the fact that the one person will not be compensated for harm done to him in reducing mortal loss to others will not be thought morally important enough to make such harming seem impermissible.
My strategy for showing that Unger-grouping is not what makes harm done to people seem morally permissible (for example, in redirecting the trolley threat) is as follows: Show that when people are Unger-grouped, it is not thought (intuitively) to be permissible to do just anything to one person in order to minimize mortal loss to others; show that when people are Unger-separate, it is still thought (intuitively) to be permissible to do certain things to them in order to minimize mortal loss to others. This suggests that what makes acts that harm permissible or not has nothing to do with Unger-grouping or Unger-separateness. Indeed, rather than permissibility stemming from Unger-grouping versus Unger-separating, grouping or separating in a sense different from Unger’s stems (to a large degree) from judgments of permissibility. For whatever factors account for the permissibility of harming, those who may permissibly (p.201) be harmed will be potentially involved with each other (i.e., grouped), and those whom it is impermissible to harm will be uninvolved with each other (i.e., not grouped).19
To carry out this strategy, I must first point to intuitions about cases that conflict with the moral significance of Unger-grouping and Unger-separating. Ideally, I should then suggest a principle other than Unger-grouping or Unger-separating that correctly distinguishes between the cases where harming is permissible and where it is not. However, I shall not here spend much time delineating the correct principle.20 This is because Unger’s claim is that by Unger-grouping we can change our intuitions about behavior, and only if intuitions change will he be able to support the principle that we may in general harm people in order to minimize mortal losses. So, all we need to do to argue against him is show that Unger-grouping does not affect intuitions as he claims that it does. We need not provide a correct alternative principle.
I have already given my first countercases to the claim that Unger-grouping explains permissibility: My intuitions imply neither that it is permissible to redirect the second trolley in a way that threatens its passengers nor that it is permissible to send in the man on skates in the Switches and Skates Case. Here are further particular examples. As noted, Unger says that in redirection cases, the people on either side of the trolley track are seen as grouped, and grouped people are fair game to be used in order to save others. This is one reason that he thinks that our intuitions tell us that we may redirect the trolley, thereby killing three people rather than six in his case. But if this were so, then we would also think that it is permissible to do many other things to Unger-grouped people. For example, consider a new case in which three people and six people are on either side of a trolley track, and a trolley is headed toward the six. These are grouped people, according to Unger’s criteria. But in this new case, redirection of the trolley will only save four of the six people at the cost of paralyzing the three, because the trolley will roll back after four have escaped and two people still remain pinned to the first track. However, if we throw the three people who are on the second track in front of the trolley, all six on the first track will be saved, though the three will die (Push-Three Case). Although these latter three people are Unger-grouped with the six, intuitively I do not think that the three are fair game to be caused greater harm than they would suffer by redirection of the trolley, in order to save all six.
In a second new case, Push-Three 2, the trolley is headed toward six people. It can be redirected to another track where it will kill three people. However, the hitting of the three on the track will not stop the trolley, and the track loops back to the six. The three and the six are grouped according to Unger’s criteria, but, of course, we should not redirect the trolley, as we would then kill nine. May we throw the three from the second track onto the main track, if this will stop the trolley but kill the three?21 I do not believe that this is permitted. Hence, members of an Unger-group are not necessarily fair game. There are some acts of harming them that still intuitively seem impermissible. This argues against Unger’s claim that Unger-grouping is sufficient to account for our intuitions when we judge that it is permissible to minimize mortal loss by harming others.
(p.202) But what about the factor of protophysics? If we think that it is more permissible to redirect a threat or person already in motion, perhaps this will account for the judgments of impermissibility in both of the Push-Three cases, even when people are Unger-grouped. For in these cases, we must push people not previously in motion.22 Can we find a case where there is Unger-grouping, no protophysical problem, and still there is impermissibility? I have already said that if we redirect the second trolley in the Switches and Skates Case, there is no protophysical problem and the people are Unger-grouped, yet I think that it is impermissible to send in the second trolley with people in it.
Also consider the following case. A trolley is headed toward six people. It can be directed onto a side track where three people are seated, but this route leads through the three and back to the six, so it is unacceptable to redirect the trolley. We can, however, redirect a second trolley, already in motion but threatening only furniture, toward the first trolley, thereby stopping it. However, the second trolley will also run into and kill the three people on the side track (Second Trolley Redirection Case). It is intuitively impermissible to send in the second trolley, even though the six and the three are grouped in Unger’s sense and there is no protophysical problem. Hence, Unger-grouping where there are no protophysical problems is not sufficient for permissibility.
Unger also says that when people are not Unger-grouped, we think that it is impermissible to involve them in other people’s problems. (So, he has an “if and only if Unger-grouped will the act be permissible” claim.) In a two-options case, he argues, a single person who is not on the trolley tracks branching from the tracks on which the trolley is headed is considered Unger-separate from the six toward whom the trolley is headed. Yet, my sense is that if we could redirect the trolley from the six, knowing that it would run on the grass killing one person, it would be permissible to redirect the trolley. Unger also thinks that someone who is sitting in his yard beneath a mountain far away from the trolley tracks is intuitively thought to be separate from the six toward whom the trolley is hurtling (Yard Case).23 However, my sense is that it is permissible to redirect the trolley away from the six, even though we foresee that it will tumble down the mountain, killing the one person sitting in his yard.
It is interesting to observe that when Unger considers the Yard Case, he does not consider redirecting the trolley so that as a result of redirection, it kills the man in his yard. Rather, he considers the following sequence of events: We start another trolley that crashes into the trolley headed toward the six, but the second trolley then tumbles down the mountain, killing the man in his yard.24 Unger says that we intuitively find this impermissible. I agree, but this impermissibility is not due to our killing someone who is Unger-separate from those threatened by the original trolley, since it is intuitively permissible to redirect the original trolley in the Yard Case. Rather, I believe, it is impermissible because a mere causal means to helping the six people—the second trolley—will kill another person.25
Furthermore, we can find Unger-separate cases in which there is also what he would consider a protophysical problem and yet it is permissible to kill one person to save others. If this is so, the combination of separation and problematic (p.203) protophysics will not account for impermissibility. (Hence, Unger-grouping and the lack of protophysical problems will not be necessary for permissibility.) For example, consider the Lazy Susan Case V26: A trolley is headed toward killing five people who are seated on a resting Lazy Susan turntable. We cannot redirect the trolley, but we can start up the Lazy Susan, turning the five away from the trolley. However, the Lazy Susan will then start a rock slide that will kill an innocent bystander sitting in his yard far away. Here we put into motion a threat and kill someone who is Unger-separate, yet I believe that this action is permissible. It is even permissible to start up the Lazy Susan, I think, if an effect of its turning is to push a stationary bystander who is on the grass into the path of the trolley. So, contrary to what Unger says about protophysics, even in a two-options case, it is not only intuitively permissible to send a threat into someone (as when we redirect the trolley), it is also intuitively permissible to send an Unger-separate person into a threat.
So, I believe that we may harm Unger-separate people in what Unger says is a protophysically problematic way, and we may not harm Unger-grouped people in a protophysically nonproblematic way. Hence, our intuitions concerning the permissibility of harming in a two-options case do not necessarily stem from Unger-groupedness and no protophysical problems, and intuitions concerning the impermissibility of harming in a two-options case do not necessarily stem from Unger-separateness and protophysical problems.
I conclude that neither Unger-grouping nor the absence of problematic protophysical factors, alone or together, is sufficient to determine the permissibility of harming some to save others. Nor is the presence of these factors, alone or together, necessary to determine the permissibility of harming. So, it is not by altering cases to achieve grouping and by manipulating protophysical factors that one can always acquire the intuitions that say it is permissible to push someone into a trolley in order to save others. This further implies that one also cannot reflect upon these as factors that change our intuitions—as there is no change wrought by these factors in many cases we have considered—and conclude that because they are per se morally insignificant, we are free to act on general values and harm some people in order to prevent greater mortal loss to others quite generally. The several-options method does not, contrary to what Unger claims, liberate us to act on the values he claims to know are true by reflection on general theses.27 Rather, it suggests that there is some other principle not identified by Unger that distinguishes permissible from impermissible harms.
C. The Correct Principle of Permissible Harm?
As described in detail in chapter 5, I believe that there is another characterization of factors that are crucial to making it permissible to harm some people in order to prevent greater mortal loss to others. Those who may permissibly be harmed when these factors are present will form a group with those for whose sake they may be harmed. (I would call this “permissibility-grouping.”) If Unger’s description of when people are grouped did, perchance, partially coincide with this other (p.204) characterization, it is this other characterization that reveals why they are grouped. Similarly, when protophysical properties are consistent with the characteristics that are crucial to making acts that harm permissible, they will not be problematic; when they are inconsistent, they will be problematic.
Those whom we may not permissibly harm for the sake of others will not form a permissibility-group with those others. But this does not mean (contrary to Unger) that their claim not to be harmed is treated as greater than that of other people. After all, we might harm them for the sake of others if permissible means of doing so were available. The impermissibility of harming them just means that they, like everyone else, including those who need to be helped, should not be treated in a certain way for the sake of others. Suppose that a large number of people are being mistreated, but it is impermissible to stop this because in order to do so we would have to mistreat a smaller group. This does not mean that morality endorses the mistreatment of the larger number. But if it were permissible to do to some people what is now thought to be impermissible in order to save others from comparable mistreatment, morality would be endorsing such treatment. This would mean that every person may be treated in these ways that we had thought were prohibited.28
What characteristic distinguishes the permissible from the impermissible cases of harming? In other words, what is the correct principle of permissible harm? As I said above, I do not intend to dwell on this question in this chapter, because I do not believe that a detailed answer is necessary for my criticism of Unger to be successful. However, a rough characterization, that summarizes what has been said in chapter 5, is important because it will help us to better understand Unger’s method of multiple options. Roughly, the permissible cases involve lesser evil* coming about as an effect, aspect, or noncausal flip side of a greater good, its component, or means that have a greater good or its component as their noncausal flip side. (Recall that the idea of a noncausal flip side can be exemplified in the Trolley Case, where the movement of the trolley away from the six people is the absence of a threat to them, which is their being saved in the context.) The impermissible cases involve lesser evil* either (a) as a means, required given our act, to causally producing a greater good or (b) as the direct effect or aspect of means that produce a greater good as a causal effect.
When we turn the Lazy Susan, we move the six people away from the threat—this is the greater good in a context where no other threats are present—and it causes the death of the one person. When we turn the trolley away from the six people, the flip side of this is that they do not have any threat upon them, and this is a greater good. So, the turning of the trolley, which also causes the death of the one, is a means that has a greater good as its noncausal flip side. By contrast, in the Switches and Skates Case, when we send in the second trolley with people on it, this causes the trolley headed toward the six people to stop and also causes the death of the people on the second trolley. Hence, sending in the second trolley is either a means that has the greater good as a causal effect and also directly causes lesser evil*, or a way to have the harmful involvement of two people (whose weight helps stop the first trolley) be a required means that causes a greater good. When we send the man on skates into the path of the trolley, the means itself is bad (as it (p.205) involves someone in a way that will harm him), and it has a further causal effect of stopping the trolley, and so causally produces the six saved.
Unger finds it odd that we think that it is impermissible to chop off someone’s foot in order to save twelve lives while at the same time we think that it is permissible to kill twelve people in order to save a foot.29 The case he describes in which we do the latter involves redirecting a trolley that will then hit twelve people away from both killing twelve other people and cutting off the foot of a thirteenth.30 But notice that in this case, it is only because twelve people would be killed anyway if we do not redirect that we redirect the trolley to kill the twelve people who are unaccompanied by a potential thirteenth victim. If chopping off the foot to save twelve violates a principle of permissible harm and redirecting when harm is minimized does not, there is no oddity, contrary to Unger. These cases show that something besides reducing mortal losses is important. I hypothesize that this something is about the high status of persons expressed in our theory of permissible and impermissible harms.
Interestingly, Unger’s method of several options can be seen, I believe, as based on a misapplication of the principle that when greater good would be the noncausal flip side of what leads to lesser evil*, action is permissible. Consider that in the Switches and Skates Case, when one agent must decide which of several acts to perform, his performing one act that causes less harm is the alternative to his performing another act that causes more harm. Hence, if an agent sends in the trolley with two people on it instead of redirecting the trolley toward three people, a greater good of three people not being hit is the noncausal flip side of sending in the trolley with two people on it instead. The method of several options can be seen, then, as an attempt to apply the redirection-of-threat model to the agent considered as a threat. That is, once the agent decides to redirect the trolley, it may be thought that he can treat himself as a threat who can be redirected from (a) turning a trolley on three people to either (b) sending in a trolley with two people on it or (c) pushing a man in front of the trolley. In general, the idea seems to be that once an agent will be a threat to some, he may redirect himself to minimize the threat that he presents. This idea is what, I believe, underlies the method of several options as Unger uses it, though he does not seem to realize it.31
What is wrong with this use of the redirection model and the principle of permissible harm that I suggested underlies it? Why is it still impermissible to send in the second trolley (or the man on skates)? The important general point that must be explained is that an agent cannot be morally free to do just anything that would otherwise be impermissible simply because it is a less harmful alternative to his doing another act that is permissible.32
Here is a proposed explanation of this general point. Suppose that it is permissible to turn the trolley away from six people and toward three because (roughly) the greater good of the six saved—call this GG1—will be the noncausal flip side of turning the threat toward three people. Sending in the second trolley with two people on it (instead of redirecting the first trolley toward the three) has the three people not hurt—call this GG2—as its noncausal flip side. But GG2 is not the same as GG1. Sending in the second trolley has GG1 as a causal effect, and (p.206) it also results in the deaths of two people. The fact that GG2 (three saved) is the noncausal flip side of sending in the second trolley does not mean that GG1 is the noncausal flip side of sending in the second trolley. Hence, we have a choice between (a) redirecting the first trolley toward three people, which is a permissible way (according to my theory) to achieve GG1 (as it has GG1 as a noncausal flip side), or (b) doing something that harms two different people in a way that (according to my theory) is not permitted in order to achieve GG1. We should not choose the impermissible way when we can choose the permissible way.
Another way of making this point is to characterize the alternatives among which an agent may choose, when he would kill different people on such alternatives, as follows: Once it is permissible for him to do one act that threatens some, he may do only the alternatives that either (1) share the property that made the first act permissible (that is, they have GG1 as their noncausal flip side), or (2) have another property that would make them permissible if they were the only possible way to achieve GG1. For example, if it were an option for an agent to send the first trolley headed to the six toward one person rather than toward three people, GG1 would be the noncausal flip side of this act as well. Hence, this act would be morally different from one where an agent sends in the second trolley instead of redirecting the first trolley toward the three people.
Notice also that there is a moral difference between the case where one sends in the second trolley with two people on it (or pushes in the one person on roller skates), instead of redirecting the first trolley toward three people, and the following two cases: (1) A trolley is headed toward killing three people. I redirect it toward a track where a second trolley sits. The first trolley pushes the second trolley into the path of a third trolley that is headed toward six people. The second trolley stops the third trolley, saving the six people, but two people on the second trolley are killed. (2) Everything is as in case (1), except that the first trolley redirected away from three people pushes a person into the path of the trolley headed toward the six, which stops the trolley but kills the one person. Redirecting the trolley is permissible in these cases.
In these two cases, as in Unger’s, doing what would save the three people would involve pushing a trolley, or a person, into the trolley headed toward six people, resulting in fewer deaths. But in these two cases, the threat to the duo or to the single person is the effect of a means (i.e., redirecting the first trolley headed toward the three people) that has saving the three (GG2) as its noncausal flip side. Indeed, Case 2 shows that it can be permissible to do an act, even though it results in evil* being a productive cause of a greater good (saving the six [GG1]) when the evil* is the effect of another good greater than it (GG2) or a means having that good as its noncausal flip side.
But in Unger’s case, is it not also true that pushing in the trolley with two people on it (or the person on skates) has the same GG2 as its noncausal flip side? After all, the three people will not be hit because an agent does an alternative act. The crucial difference between my Cases 1 and 2 and Unger’s case is that only in Unger’s case is it possible to produce GG2 by just not sending the trolley toward the three people. They face only a possible act of an agent, rather than a (p.207) threat already headed toward them. Hence, in his case, it is not necessary to do what sends the trolley with two people on it (or the single person on skates) into the path of the trolley that is headed toward the six, in order to prevent that trolley from being redirected toward killing three people. Doing what redirects the trolley with two people on it is not necessary as a means to saving the three people in Unger’s case; the means to saving them in Unger’s case is just not to send the trolley toward them. By contrast, in my Cases 1 and 2, a trolley is already headed toward the three, and the necessary means to save them is to redirect the trolley, leading to the further consequences described.
Of course, if we simply do not redirect the trolley toward the three people and do nothing else in Unger’s case, the six people will be killed. Hence, performing the alternative act of redirecting toward the three is a means necessary to save the six, but this is a different issue entirely from what is necessary to save the three. And it has been said that (roughly) a means to saving the six people that also directly causes harm to innocent bystanders should not productively cause that greater good but rather have a noncausal relation to that greater good (or be caused by something that has such a relation). Since the alternative acts in Unger’s case (i.e., push in the second trolley or push in one person) are not means necessary to achieve GG2 (save the three) but to achieve GG1, these alternative acts should have GG1 as their noncausal flip side. But they do not. Hence, they are impermissible.
D. Harms with an Intervening Agent
In the Switches and Skates Case, I would be the only agent involved in causing harm to others unless I merely let the trolley head toward the six. In another type of case that Unger discusses, I do not alone harm others, but a decision I make to help reduce harm overall can lead to someone else harming others. These are the UNICEF Card and Lesser-Loss-Card cases.33 In the UNICEF Card Case, if I pick up a UNICEF card, aid will go to many people in Africa, but the villain Strangemind will then send his henchmen to chop off someone’s foot in Asia. In the Lesser-Loss-Card Case, if I pick up a UNICEF card, I will save the lives of fifty people in Africa from Strangemind’s henchmen, but he will then send other henchmen to Asia to chop off someone’s foot. Unger thinks that our intuitions are that it is impermissible for me to pick up the UNICEF card in the UNICEF Card Case but permissible for me to do so in the Lesser-Loss-Card Case. He distinguishes the cases on the basis of Unger-grouping and Unger-separating.
In the UNICEF Card Case, the suffering people in Africa are not suffering from any problem that Strangemind gave them, so the cause of the threat to the Asian whom Strangemind’s henchmen will hurt is different from what threatens the Africans. This makes the Africans and Asians seem separate from each other, according to Unger. By contrast, in the Lesser-Loss-Card Case, the same threat—Strangemind—threatens both the Asians and the Africans, and his henchman’s movement to Asia is the flip side of the removal of the henchmen from Africa. This, according to Unger, creates the sense that the one person in Asia and the fifty people in Africa are part of a group. Hence, it seems permissible to help the (p.208) Africans by doing something that leads to harming the Asian. Once we see that this is all that grounds our sense of permissibility, we should be free to act on our general values and conclude that it is permissible to pick up the UNICEF card in the UNICEF Card Case as well, he claims.
It should be clear (though he does not make it clear) that Unger has modeled the Lesser-Loss-Card Case on the redirection cases we have been discussing, with the following variation: There are, strictly speaking, two different threats in the Lesser-Loss-Card Case. That is, either one set of henchmen is sent to attack Africans if I do not pick up the card or another set of henchmen is sent to Asia if I pick up the card stopping the first set of henchmen. This is by contrast with the villain sending the same henchmen elsewhere. (It might seem that there is another variation introduced into this case. It is my picking up the UNICEF card that is both a means to stop one threat and what starts another. This is by contrast with the stopped threat itself causing another threat. But on reflection it seems that this variation is not really present, for presumably it is Strangemind’s recognition of the fact that his henchmen will not go to Africa that is necessary in order for him to send the other henchmen to Asia.)34
Despite the variation, I think that, as in the Switches and Skates Case, Unger’s aim is to move from the permissibility of redirecting a threat in order to save a greater number to the permissibility of saving a greater number by any means. The first problem with Unger’s strategy, I believe, is that picking up the card in the UNICEF Card Case is not, intuitively, impermissible. I may save many people despite the fact that (I know that) someone else will, voluntarily and uncoerced by me, make a decision to harm someone else based either on how I act or on what the consequences of my acting are. The same can be said to account for the permissibility of picking up the card in the Lesser-Loss-Card Case.
But suppose we assume, for the sake of argument, that intuitively only action in the Lesser-Loss-Card Case is judged to be permissible. My strategy in dealing with Unger’s attempt to assimilate the Lesser-Loss-Card and UNICEF Card cases would be to show that Unger-grouping does not account for permissibility, and Unger-separation does not account for impermissibility.
First, consider those who are Unger-grouped and yet are not fair game to be harmed. In my Lesser-Loss-Card Case 2, if I pick up the UNICEF card (a) Strangemind sends fewer henchmen to kill a group of Africans than he otherwise would have, killing only twenty instead of fifty, but he also sends the rest of his henchmen to chop off the foot of an Asian. So, the one in Asia and the many in Africa are Unger-grouped. But if I pick up the UNICEF card (b) in my Lesser-Loss-Card Case 2, Strangemind will send a henchman to kill the Asian and, because this has such a calming effect on him, he will not send any of the other henchmen he would otherwise send to Africa. I think that it is intuitively impermissible to pick up the UNICEF card (b) in the Lesser-Loss-Card Case 2, even if the only alternative is to pick up the UNICEF card (a) instead. Hence, Unger-grouped people are not fair game. My explanation of the impermissibility of killing the Unger-grouped Asian is as follows: If I pick up UNICEF card (b), my means to saving many lives (the greater good) would be a lesser evil* (having Strangemind be (p.209) calmed by killing one person). By contrast (and abstracting from the significance of another agent’s intervening voluntary act), in the original Lesser-Loss-Card Case and in the Lesser-Loss-Card Case 2 where we pick up UNICEF card (a), the following is true: The lesser evil* seems to be either an effect of means that have a greater good as a noncausal flip side (because threat  not facing group  is the noncausal flip side of doing what leads to threat  facing group ) or an effect of the greater good itself (if Strangemind acts to harm the Asian because the Africans will not be threatened).
Now consider those who are Unger-separate. In my Angry Strangemind Case, if I pick up a UNICEF card, money gets sent to save many people in Africa who are suffering independently of any threat of Strangemind’s. As a consequence of seeing them so well off, Strangemind gets angry and sends his henchmen to kill an Asian. I believe that it is intuitively permissible to pick up a UNICEF card in this case, even though the Asian and the Africans are Unger-separate. It is permissible (even if we abstract from the significance of another agent’s intervening voluntary act) because the greater good results in the lesser evil*, even if the lesser evil* is caused by a different threat than that faced by the Africans. (This would contrast with a case in which it is the means—our choice—that has a greater good as a mere causal effect and that also triggers, independent of Strangemind’s voluntary act, a lesser evil*.) Hence, it is not always impermissible to do what harms those who are Unger-separate.
E. Tolerating and Imposing Nonphysical Harm on Ourselves
Recall that Unger wishes to move from the permissibility of harming others to our having a duty to harm ourselves in order to prevent greater mortal loss. I have examined and criticized a significant part of his attempt to prove the first premise, that it is often permissible to harm others in ways nonconsequentialists would rule out. One route from the first premise, had it been proven, to the conclusion that we have a duty to harm ourselves would involve what Unger calls the Principle of Ethical Integrity (PEI). Roughly, this principle says that what you would permissibly do to others, you cannot fail to do to yourself for the sake of equal or greater reduction in mortal losses.35 In dealing with cases, as we have seen, Unger’s goal is to assimilate our treatment of people in one sort of case (e.g., pushing them in front of a trolley) to other cases (e.g., redirecting a threat toward them). Similarly, his attempt via the PEI is to assimilate our treatment of ourselves to our treatment of others.
A less direct route to the conclusion is to show, by using cases, that we think we should tolerate serious harm to ourselves rather than have others suffer even greater losses. Then we would have to move from showing what we should tolerate to showing what we should impose on ourselves. I shall first deal with this less direct route, putting off discussion of the PEI. I do this because I have already argued that Unger has not shown that we may always harm others to lessen mortal loss. If we may not harm others, then we cannot show that we must harm ourselves on the ground that we must do to ourselves what we would permissibly do to others. (p.210) Perhaps showing that we must tolerate serious harm to ourselves will be a more successful route to the requirement that we seriously harm ourselves. Let us start by examining tolerating less serious nonphysical harm.
In Unger’s Bankcard Case, Strangemind will impose a $1,000 loss on your bank account—we start with a property loss—unless you pick up a bankcard, and if you do pick it up, he will send a henchman to chop off a distant stranger’s foot. Unger agrees that intuitively we think that it is impermissible to pick up the bankcard. He contrasts this with our intuition, in what he calls the Envelope Case, that it is permissible to refuse to send $1,000 to a charity that would save a distant stranger from losing his life, let alone his foot. Unger thinks that he can account for the intuitive distinction we draw between these two cases on the basis of Unger-grouping and Unger-separating, respectively. Only in the Bankcard Case is the individual who would threaten your bank account the same person who would threaten the distant stranger, and his threatening the stranger is the strict alternative to your not picking up the bankcard. Hence, according to Unger, you think of yourself as grouped with the Asian. This liberates you to what Unger thinks is a true value, that a lesser loss to you is not as important as preventing a bigger loss to another person quite generally. Given the actual moral irrelevance (in Unger’s view) of Unger-grouping, from this you should deduce that your intuition in the Envelope Case was wrong.
My analysis of the Bankcard Case differs from Unger’s, and I do not think the case supports his conclusion. The structure of the Bankcard Case is, once again, like that of a redirection-of-threat case, except that it involves a different agent (Strangemind) voluntarily deciding to do one bad thing if he is prevented from doing another bad thing. (To reflect this, I will call it a “redirection” case.) Picking up the bankcard would involve redirecting Strangemind from causing a small loss to you to his causing a greater loss to others. Hence, it seems wrong to redirect Strangemind. But the impermissibility of doing what leads to harm in the Bankcard Case is not dependent on Unger-grouping or even on the presence of a redirection structure. To see this, suppose that a villain will cause a loss to my bank account and the only means I can use to stop him has as a direct effect that a distant stranger is killed. Here the distant stranger is separate, according to Unger’s criteria, because what threatens her is my means, and this is different from what threatens me. Yet it is still impermissible to do what helps me and harms her. So Unger-grouping is not necessary to account for why we must tolerate cases of lesser loss to ourselves rather than do what leads to harm to others.
Notice, in addition, that Unger-grouping may make it intuitively morally easier (rather than harder) to harm another on our own behalf than does Unger-separation. Suppose that a villain threatens to chop off my legs unless I pick up the bankcard, and if I pick it up, he sends a henchman to chop off a distant stranger’s foot. Because this is “redirection” of the villain from greater harm to lesser harm, it may be permissible. (Even if it were strict redirection, it might be permissible.) Such redirection also seems to be endorsed by the values that Unger thinks are correct, that is, it reduces overall severe harm. But if a villain will chop off my legs unless I shoot him, it is not clear that I may shoot him if I know that the bullet will kill the villain but also shoot off the leg of a distant stranger. (This is on the model (p.211) of its being impermissible to set off a bomb in order to stop a trolley from killing five people when we foresee that it will also kill an innocent bystander. However, it might be permissible for me to shoot, if what we permit in defense of self were broader than what we permit in defense of others. This would, of course, create problems for Unger’s PEI.)
More evidence can be given for the view that there is a particular duty to sometimes suffer losses instead of doing what leads to harm (even lesser harm) to other people, rather than any general duty to bear lesser losses in order to prevent greater mortal loss. Such evidence comes from the fact that Unger-grouping is not sufficient to support intuitions that we must suffer losses in order to aid those with whom we are Unger-grouped. For example, consider another revision of Unger’s case, my Bankcard Plus $1,000 Case: If I pick up a bankcard, I prevent Strangemind from taking $1,000 from my account, but he goes off to chop off a distant person’s leg. If I do not pick up a bankcard, he chops off only that same distant person’s foot. I could prevent his even chopping off the foot by giving another $1,000 to him. Suppose that I judge intuitively that I should not pick up the bankcard, so as not to be involved in harm to the distant person. I do not also intuitively judge that the duty to give the extra $1,000 to aid the person with whom I am Unger-grouped is any stronger (or weaker) than the duty to aid a person from whom I am Unger-separate who would lose a foot in an Envelope Case. Hence, it is not Unger-grouping that is responsible for our different intuitions in the Bankcard and Envelope cases, nor is it grouping that overcomes intuitions that support the Envelope Case.
Suppose that we give a redirection structure to the aid itself. Consider my Keeping Busy Case, in which Strangemind will send a henchman to chop off a distant stranger’s foot, but I can prevent this by keeping Strangemind busy removing $1,000 from my bank account. By making Strangemind a threat to me, I redirect him to where he does the least damage. In this case, I and the stranger are Unger-grouped, but intuitively there is no stronger (or weaker) obligation to aid the stranger in this way than in the Envelope Case. Now consider my Generous Strangemind Case, in which Strangemind offers to either give me $1,000 or—the flip side of not doing this—give a distant needy stranger $2,000. Again, there is Unger-grouping of the people, because we are redirecting a benefit, but, intuitively, there is not as strong a duty to refuse the money as there was not to resist Strangemind’s taking my money in Unger’s original Bankcard Case. Presumably, this is because I would be involved in Strangemind’s harming the distant stranger in Unger’s Bankcard Case, but only involved in Strangemind’s not-aiding the stranger here. Now consider my Joint Venture Case, in which Strangemind offers to give some of his own money to a distant needy stranger, if I allow him to take some of my money. If he cannot take my money, he will not give to the distant stranger. In this case, if I resist his offer, a benefit does not go to the distant stranger with whom I am Unger-grouped. Yet, intuitively, it is no more or less permissible to resist the offer than to refuse aid in the Envelope Case.36
In the Keeping Busy Case, my suffering a loss in order to aid another would take the form of my redirecting a threat from a distant stranger to myself. In such cases, I do not merely fail to resist a nonmortal loss that someone else will impose (p.212) on me; I impose it on myself. I concluded that I need no more (or less) impose the loss on myself in that case than in the Envelope Case. But Unger believes that in his Bob’s Bugatti Case, our intuitions tell us that we must impose at least a very big financial loss on ourselves. In this case, Bob’s Bugatti (in which he has invested his entire retirement fund) is parked on one arm of a branching trolley track. On the other arm is a child and a trolley is headed toward killing her. Unger thinks that we will agree that Bob has a duty to turn the trolley away from the child, even though he foresees that it will certainly cost him his retirement fund. Suppose that we agree. This raises the question: Why is Bob intuitively required to redirect a threat toward his Bugatti, but the person in my Keeping Busy Case is not required to redirect Strangemind toward harming his bank account? Four hypotheses suggest themselves.
First, in the Bob’s Bugatti Case, someone will lose his life, whereas in the Keeping Busy Case, only a foot is at stake. (But the monetary loss is also lower in the Keeping Busy Case.)
Second, the Bugatti will be destroyed as a consequence of turning the trolley away from the child. Hence, the case’s destruction is a side effect of means that have a greater good as a noncausal flip side. In the Keeping Busy Case, we intend that Strangemind pay harmful attention to our bank account and, I believe, we conceive of this as a means that has the greater good (of the other person not threatened by Strangemind) as its noncausal flip side. So here our involvement that leads to our financial harm is itself a means to helping another person.37
We can make this second point as well by considering another variation on the Bob’s Bugatti Case. Suppose that turning the trolley away from killing the child cannot be done quickly enough to prevent its running over the child and partially paralyzing its legs. The Bugatti would also only be partially destroyed by the redirection, leaving Bob with a still-significant retirement fund. However, if instead of redirecting the trolley, Bob sends the Bugatti across the tracks into the trolley, he will completely prevent harm to the child and the Bugatti will be completely destroyed. (Call this Push-the-Bugatti Case.) Do we, intuitively, believe that he must do this for the child with whom he is Unger-grouped? I do not believe so. This means that we get no stronger (or weaker) conclusion from an Unger-grouped case for a duty to aid by imposing losses on ourselves as a mere means to preventing mortal harm (as in Push-the-Bugatti Case) than we would get from a case that involved people who are physically near to us (as they are in Bob’s Bugatti) but who are nevertheless Unger-separate.
The third hypothesis to distinguish the Bob’s Bugatti Case from the Keeping Busy Case is that physical distance makes a moral difference. The person to be harmed in the Keeping Busy Case is a distant stranger, as are those in the Envelope Case. The one in the Bob’s Bugatti Case is a near stranger. Unger does not believe that physical distance can affect obligations. Perhaps he is wrong about this.38
The fourth hypothesis to distinguish the Bob’s Bugatti Case from the Keeping Busy Case is that when an intervening agent who acts voluntarily, like Strangemind, is involved, we need not take as much responsibility for preventing harm, because we should not allow ourselves to be manipulated by evil people who are (p.213) primarily responsible for the harm. On these grounds, it might be argued not only that we should not impose harm on ourselves but that we should not be involved in redirecting more generally. After all, there is a more appropriate way to diminish the threats of evil people: They should control themselves. They are persons, not physical objects like threatening trolleys that have no self-control.
This hypothesis, however, implies that we need not rescue a drowning baby right near us, simply because she was pushed in by an evil person whose aim was to affect our behavior. But this is not true.39 Note also that in the Envelope Case, as in the Bob’s Bugatti Case, evil perpetrators are not the (usual) cause of harm to distant strangers who need the money in the envelope.
In sum, Unger-grouping does not account for our intuitions that we should tolerate losses to ourselves rather than be involved in harming someone, because we also should not harm those from whom we are Unger-separate. Unger-grouping also does not lead us to change our intuitions about how much we should do to minimize loss to others by imposing losses on ourselves in order to aid. The harming/not-aiding distinction and the distinction between suffering losses as a means to aid or as a consequence of aid, rather than the Unger-grouped versus Unger-separate distinction, seem to be doing work in accounting for the intuitions involved in the Bob’s Bugatti and Keeping Busy cases. Therefore, we cannot take the next step in Unger’s argument, which is to reflect on the per se moral insignificance of Unger-grouping and then deduce the duty to impose losses of property on ourselves in order to prevent greater loss quite generally, including to those from whom we are Unger-separate.
In concluding this section, consider Unger’s claim that our duty to our dependents (such as our children) might override the duty to impose property losses on ourselves in order to lessen mortal loss to distant strangers. First, I do not understand how, on Unger’s view, one has a right to come to have dependents at all, if one foresees that this will prevent one from engaging in projects that minimize mortal loss (as is probably true in our world). For if I have a duty to suffer mortal loss in order to minimize mortal loss—a claim of Unger’s I will examine in the next section—then it would seem that I have a duty to suffer the loss of having offspring, in the current world as it is, if this would help me to save distant strangers instead.
Suppose, however, that I may permissibly have dependents (perhaps because I am more productive in aiding if I have them than if I do not, due to a limitation of my human psychology). Unger says that I may be justified in not minimizing mortal loss in order to pay for my dependents’ education instead. But if Unger wants to take our intuitive judgments about what losses we should undergo in redirection cases, such as the Bob’s Bugatti Case, as an indication of what we should be committed to sacrificing for others in general, then his view that we need not give away money that is necessary to provide for our dependents’ needs would be called into question. This is because if a trolley is headed toward killing a stranger, we should redirect it to a track where it will destroy irreplaceable money needed to educate our dependent (Education Fund Case). If Unger’s strategy of assimilating cases were correct, this would further imply that we should give our dependent’s education fund away to save a stranger’s life, contrary to what he claims.
(p.214) I, by contrast, suggest that we (or others) may be obligated to redirect the trolley, though it will foreseeably destroy the money, but not to give the money away. On Unger’s views, however, to say that I need not give the money to famine relief implies that I need not turn the trolley. Also, Unger’s claim (examined earlier) that we should quite generally kill someone else (even as a means) in order to save yet others from mortal loss implies that not saving life is at least as wrong as killing. But this claim, in conjunction with his view that we need not sacrifice funds set aside for our dependents in order to save the lives of strangers, implies that we may do what kills others if this is necessary to save the education money for our dependents. Unger’s general strategy of equating harming with not-aiding, and notharming with aiding, combined with the exceptional, nonoverrideable duty to dependents, yields a duty to kill strangers in order to protect funds for one’s children’s education. This is an unacceptable conclusion.40
F. Imposing Mortal Loss on Ourselves
Unger would ultimately like to show that we have a duty quite generally to impose mortal loss on ourselves for the sake of reducing overall mortal losses.41 So far, when it comes to imposing mortal (rather than property) loss, I have argued that, despite Unger’s arguments, direct intuitions or deductions from intuitions show only that it is permissible to redirect threats (or people from threats) in order to minimize mortal loss. Of course, Unger also thinks that reflections on general theses, such as the overriding importance of reducing mortal losses, support imposing mortal loss on ourselves. But his aim was to quell disagreement about the correctness of such general theses by showing that intuitions in selected cases—plus reflection on the moral unimportance of particular factors in those cases (missing in others) that lead to those intuitions—liberate us to follow the general theses. I claim that he has not shown this.
What do we get when we combine what we already believe intuitively about cases with a principle that we should treat ourselves as we treat others, such as his PEI, a general principle that Unger thinks that we accept on reflection? Unger seems to believe that we can, at least, derive the duty to do mortal damage to ourselves in redirection cases. And given his assimilation of different types of cases to redirection cases, he believes we can also derive a duty to do mortal damage to ourselves in order to minimize mortal losses quite generally. I emphasize that he derives this result. He does not present a case in which one has the direct intuition that one must sacrifice one’s life in order to save two strangers, though he thinks that we have case intuitions (as in the Bob’s Bugatti Case) that support imposing property damage on ourselves. To illustrate what the PEI implies, Unger presents the Trilemma Case: If a trolley is headed toward six people, I should redirect it toward three, but if I can redirect it toward one instead, I should do that. If the one happens to be me, the conclusion still holds.
The only duty that Unger allows may take precedence over the duty to minimize mortal loss is a duty to our closest dependents, for example, our children. But since they, too, have duties to sacrifice themselves to prevent others’ losses, (p.215) what we do for them presumably should be tempered by the thought of what they are duty-bound to suffer for others.42
I am not convinced by these arguments. In the Trolley cases, I argued that intuitively it is permissible to redirect the trolley and cause mortal loss in order to minimize overall mortal loss, not that it is a duty to do so. But we already know that it is permissible for us to suffer mortal loss for the sake of minimizing overall mortal loss; we do not need the PEI to show that we are permitted to sacrifice ourselves for others. If we do not have a duty to kill others, the PEI cannot show that we have a duty to kill ourselves.
However, the Trilemma Case does, I believe, provide a scenario in which we have some duty to turn a trolley in a certain way, though I do not think that Unger brings this point out. It may be merely permissible, not required, to turn the trolley from six toward three, but if I am actually about to do this, and I also have the option to redirect from six toward one other person, then I think, intuitively, I would have a duty to redirect the trolley toward the one other person, other things being equal. In other words, I may not have a duty to reduce harm by redirecting the trolley, but if I decide to do so, I have to choose the route that causes the least harm, other things being equal. We have now derived a conditional duty to kill one person.
However, other things would not be equal if I would have to suffer a large cost, even downstream, to redirect from three to one other person. Intuitively, I do not think that I would have to suffer a large personal cost to reroute from the three toward the one, any more than to redirect anywhere. (For example, suppose I could easily enough redirect the trolley from six to one except that this would set off a bomb that will kill me. I do not think I then am required to turn the trolley from six to one.)
From these intuitive judgments, in combination with the PEI, we can conclude that I have a conditional duty to turn the trolley toward myself away from the six and the three, if I am the one other person, so long as this causes me no large personal sacrifice. But I would have to suffer a large cost if I rerouted the trolley toward myself for I would die. Hence, intuitions in cases, combined with the PEI, do not show that I have a duty to do so.43
Could we revise the argument for a duty to harm one versus three, if we redirect from six, so that it implies a duty to harm oneself? Suppose it is permissible for me to redirect a trolley toward three people and I have a duty, given that I would do this, to redirect toward one other person instead. Suppose also that I must abide by the PEI. Then, if I have a duty to impose a loss as large as death on one person rather than kill the three, it is not true, it may be said, that I need not redirect from three people toward one other person just because the cost to me of doing so is great. For example, contrary to intuitions, I would have a duty to make a big effort (e.g., break my back) to shift the trolley from three to one. What is odd in this revised argument is that we could find ourselves with a duty to suffer a big loss in order to fulfill a duty to kill one rather than three even if, intuitively, we did not have a duty to suffer such a loss to save the original group of six, as saving them was merely permissible. (For example, I would not have had to break my back to redirect the trolley from the six toward the three if that were my only option.) I (p.216) conclude that the duty to kill one instead of three to save the six does not give rise to a duty to impose large losses on oneself.
Unger’s claim that I morally must redirect the trolley toward myself also conflicts with our ordinary intuitions about what those others toward whom we redirect threats may do, whether we think that we are merely permitted or duty-bound to redirect toward them. That is, we do not think that they have a duty not to resist the assault on them. If the single other person toward whom I redirect can physically push the trolley away from him, we intuitively think, I believe, that he may do so, even if he foresees that it will kill me, who redirected toward him, or the six toward whom the trolley was originally headed. Intuitively, he may, at least, permissibly make things as they would have been without redirection by re-redirecting, even if this does not minimize overall mortal loss. If he may also permissibly harm the agent who redirected toward him, as a side effect of sending the trolley away, he would sometimes be permitted to even make the situation worse than it would have been without the original redirection.44 Finally, if he has a button that would permit him to turn the trolley away from the six and toward himself, he would not have a duty to use it.
If another person would not have a duty to direct the trolley toward himself, I would not, contrary to Unger, have a duty to do it to myself either even if one accepted the PEI. We impartially universalize the permission not to sacrifice oneself and in this way support the PEI. But in this universalized permission, there is still embedded a rejection of the complete PEI, since all potential victims would be permitted to treat themselves differently from how they treat other people, as they redirect a trolley toward others that they would not redirect to themselves. Our intuitions, at least, support a self-other asymmetry at some level. So, whereas previously I argued, contra Unger, that we cannot move from redirection cases to ones where we directly give up or cause someone to give up something to help reduce mortal losses, here I suggest that one cannot move from what one even has a duty to do to others to what one must do to oneself.
I conclude that Unger’s arguments for the conclusion that we have a duty to impose mortal loss on ourselves in order to reduce overall mortal losses fail. More generally, I believe that his inventive attempts to prove his other substantive claims do not succeed. This failure is connected with the fact that his methodological claims are also not sustained; he does not succeed in showing that intuitions in the relevant cases are manipulatable by morally insignificant factors nor that intuitive differences in cases are due to the factors that he identifies as, and believes we agree are, morally insignificant.45
G. Sensitive Reconciliationism
I have argued that Unger’s arguments fail to support his radical conclusions, but of course he believes that he succeeds. So, in addition to presenting substantive normative claims and a theory about how to develop a normative theory, Unger tries to develop what he calls “context-sensitive semantics” in order to make it possible for him to live with the enlightened but speak with the vulgar. That is, he tries to explain (p.217) how someone who accepts his normative claim that it is wrong not to send a lot of money to Oxfam (the Envelope Case) can still agree with the common folk that not sending the money is not morally so bad. He also tries to explain how someone can believe that it is permissible to steal in order to save distant people from mortal loss and yet still agree with the common folk who morally condemn such theft.
Unger’s first explanation relies on distinguishing two contexts. The first context is one in which one’s judgment about the correctness of someone’s act flows from one’s primary normative values (e.g., reduce suffering). The second context is one in which one’s judgment about the correctness of someone’s act flows from one’s secondary values. These values tell us that it is important to know, and be motivated to carry out, the acts that are correct according to one’s primary values. If it is hard for someone to know how she should act or to be motivated to act as she should, we may, from the point of view of secondary values, think that she is acting morally well, even if she does not do the morally right act.
However, notice that it is an implication of Unger’s context-sensitive semantics that when someone with Unger’s normative views speaks with the “commoners,” in saying that a person who does not send much money to Oxfam is not doing something morally very bad, he is not really agreeing with the commoners. This is because they are claiming that from the point of view of primary normative values, the person is doing okay morally if she does not give money in the Envelope Case, whereas in agreeing, the Ungerite is really just saying that the person is not to be blamed for not doing the right act. If the commoner asks Unger whether the person who does not give to Oxfam is doing something wrong, he is not asking whether it is hard for her to know the correct values or to be motivated by the correct values or whether she should be blamed. But that is the question Unger would be answering.
Unger notes that agreement between the Ungerite and the commoners in condemning stealing while the Ungerite believes that stealing is right cannot be explained by Unger’s first (supposed) explanation. This is because the thief who acts on Unger’s theory, and whom the commoners condemn, is not someone who can be condemned on the basis of secondary values, that is, he cannot be condemned because he does not know a moral truth that it is easy to know or is not motivated to act when it is easy to be motivated by moral truth. Rather, the thief is someone who is imagined to know what Unger believes are true primary values and who is motivated to act on them. The commoners just think that his values do not truly justify his conduct.
Unger’s suggestion for addressing this second problem is to distinguish a context in which our judgment is based on primary normative values about which act is morally right from a context in which our judgment is based on social norms. However, on this account, Unger is not really agreeing with the commoners, for they do not only claim that stealing is wrong according to social norms (as the Ungerite does), nor do they ask Unger whether stealing is wrong according to the social norms. They are claiming that stealing is wrong according to the true moral values. They are thinking about moral truth, but Unger is thinking about sociological truth when he says that the thief acts wrongly.
(p.218) What shall we say in a case in which it is easy for someone both to know and to be motivated to do what Unger thinks is correct according to primary values, yet the person rebels and does what Unger thinks is wrong but what the commoners think is right, all according to primary values? For example, suppose that someone grows up in a home of act utilitarians who educate him in accord with act utilitarianism. Nevertheless, the person rebels and represses his tendency to give much of his money away (Rebellious Utilitarian Case). We think that he is doing something permissible, but according to Unger’s theory, he does the wrong act, though it was easy for him to do the right act. From the point of view of secondary values, as well as from the point of view of primary values, he seems to be doing very badly on Unger’s view. The rebellious utilitarian is unlike the thief who steals for Oxfam and does well on both primary and secondary values according to Unger, yet is condemned according to the social norms. From the commoners’ point of view, the rebellious utilitarian is acceptable on primary values and does very well from the point of view of secondary values, because against all odds, he has found out the truth (as commoners believe it to be). This is a case in which Unger can assent to the common judgment that the rebellious utilitarian’s behavior is permissible only by using the social-norm standard, that is, the rebellious utilitarian is doing what is all right according to social norms. But Unger cannot make what he will consider a true moral judgment about the rebellious utilitarian from the point of view of either Unger’s primary values or his secondary values and still agree with the commoners.
Furthermore, the introduction of two different standards—secondary values and social norms—for reconciling Unger’s theory with common judgment makes for a problem that Unger does not notice. If his reconciliation semantics is correct, he should also be able to agree when the commoners say, “Stealing money to send to Oxfam is worse than not giving your own money in the Envelope Case,” which is an implication of their views that “stealing to give to Oxfam is wrong,” and “not giving money in the Envelope Case is not so bad.” But how is Unger to understand that sentence so that he can agree with it? He himself warns us about using two standards (or referring to two standard-determined contexts) simultaneously.46
Unger uses two different standards to understand each of the component claims that lie behind the sentence (i.e., he uses the secondary-values standard to understand that not sending money to Oxfam is not so bad and the social-norms standard to understand that stealing is wrong). Will he use both standards in agreeing with the comparative claim? But then is he committed to the claim that it is worse to violate society’s norms (using the social-norms standard for the first part of the comparative) than to fail to know or be motivated by the truth when it is hard to do so (using the secondary-values standard for the second part of the comparative)? In excusing someone from doing the wrong act because it is hard to know or be motivated by the truth, Unger gives the impression that he should judge harshly someone who knows the truth and is motivated to act on it, yet instead chooses to abide by society’s norm. Why, then, should he condemn so strongly someone who does not abide by society’s norm and acts on the truth (as Unger sees it)?
(p.219) There is a single standard that Unger could use to translate both components that lie behind the comparative sentence and the comparative judgment of the sentence: the standard of social norms. He could agree with the comparative in the sense that he agrees that society’s norms condemn the stealing more than they condemn not giving money in the Envelope Case. But then, when he “agreed” with the commoners, he would never be making a genuine moral judgment—not even one from the point of view of secondary rather than primary values; he would just be engaging in sociological reflection.
The commoners may also make another comparative judgment: They may say that the rebellious utilitarian who does not send his money to Oxfam is in some respects morally superior to the ordinary person in the Envelope Case who does not send his money. He is superior from the point of view of secondary values, for he had to work hard to see the truth that he is permitted to act in this way. By what standard can Unger agree? From the point of view of social norms about primary values, the two agents are equally good. From the point of view of secondary values, as Unger would see it, the rebellious utilitarian is worse.47 So there seem to be some cases where Unger cannot hold his moral views, and his two translation standards, and also agree with what the commoners say. I conclude that Unger’s context-sensitive semantics has significant problems.
There may, it might be suggested, be a different sort of reconciliation possible between Unger’s normative views and ordinary nonconsequentialist moral views. In conclusion, let me sketch this possibility and its implication. In ordinary morality, consent by the person to be harmed (when he is competent to give consent) can often justify acts that would otherwise be impermissible. Unger endorses a theory which says one ought to harm people to minimize mortal losses (a limited version of negative utilitarianism). Can such endorsement be taken to be a form of consent to his (Unger’s) being harmed to minimize mortal losses? At the very least, it implies that he (and others who accept his theory) would not, if they were rational (consistent), raise a moral objection to being harmed to minimize mortal losses. If someone would not, insofar as he is rational, raise a moral objection to our harming him, could ordinary morality permit harming him in ways that would ordinarily be impermissible? For, in many cases, even if people would object to our harming them on grounds of personal preference, if they cannot raise a moral objection to this, we may still harm them.
Of course, Unger may object that others whom his theory says are also eligible to be harmed will not be eligible to be harmed according to ordinary morality, as they do not accept his theory or actually consent to being treated as his theory recommends. This implies that, according to his theory, he and others who believe as he does are unfairly burdened relative to other people, if we treat only them according to his theory. But his own theory, presumably, yields the conclusion that fairness should be sacrificed if it stands in the way of redressing mortal losses. Hence, those who endorse his theory could not rationally raise a moral objection based on unfairness to being harmed to prevent mortal losses to others.
The problem with this attempt at a partial reconciliation between Unger’s theory (and also between other theories, such as act utilitarianism) and ordinary (p.220) morality is that the inability of a rational Ungerite to raise a moral objection does not mean there is no moral objection to be raised. From the point of view of ordinary morality, the absence of consent, even due to a personal preference inconsistent with an incorrect moral theory one holds, will leave in place the moral objection to a harmful act that ordinary morality raises. If Unger’s theory is wrong, reconciliation is possible only if the theory’s endorsement by its supporters might appropriately be taken as a proxy for their consent to some of the acts it prescribes when done to them. Only then would ordinary morality accept that those who endorse the view that people may be killed in order to reduce mortal losses are likely candidates to be permissibly be killed to reduce mortal losses.48
(1) . Peter Unger, Living High and Letting Die (New York: Oxford University Press, 1996).
(2) . Unger contrasts his position on intuitions with what he calls “negativism.” He describes the latter as the complete denial that intuitions about cases reveal any moral truth. See ibid. p. 13n. Presumably, he says this because positive distortion can lead to intuitions that reveal moral truth. In the same paragraph, however, he also describes negativism not as a methodological stance dismissing intuitions, but as the attempt to bring about consistency between seemingly conflicting intuitions about aiding by denying the duty to aid quite generally, rather than extending the duty to aid quite generally. This makes negativism a substantive position about aiding: We need not suffer any loss to reduce mortal loss. Such substantive negativism also liberates us from what, according to its claim, is negative distorted (i.e., nonvalue-tracking) intuitions about cases, only it implies that the negative distorted intuitions are the ones that tell us to aid. These are the very ones that Unger thinks salience distorts positively(i.e., in a value tracking way).
(3) . As suggested by Gopal Sreenivasan.
(4) . Consequentialists have often accused nonconsequentialists of being inconsistent. For example, they ask, why do nonconsequentialists say that we may redirect a trolley to save five people from being killed by it when we foresee that this action will kill one, if we may not save five people from the trolley by pushing someone in front of it? As we shall see, Unger repeats this criticism, but he also goes beyond it. He tries to show that sometimes we would say that it is permissible to push someone in front of a trolley to save five and sometimes we would say that it is impermissible to do this, and this (he claims) is a clear inconsistency.
(5) . Insofar as he tries to change our responses to acts on the basis of altering what he thinks are morally irrelevant features of the context, his technique is like the framing technique used by cognitive psychologists Daniel Kahneman and Amos Tversky. See chapter 14 for a discussion of Kahneman and Tversky.
(6) . Unger, Living High and Letting Die, p. 37, n. 7.
(10) . Unger here follows, I believe, my focus on this flip-side aspect of the Trolley Case, as presented first in my “Harming Some to Save Others,” Philosophical Studies 57 (November 1989): 227–60, and then in my Morality, Mortality, vol. 2. I continue to focus on it in chapter 5 in this volume.
(11) . Unger, Living High and Letting Die, pp. 101–8.
(13) . The method of aptly combining cases, which Unger also describes, is just a way, I believe, of creating cases with several options. See Unger, Living High and Letting Die, pp. 106–14.
(14) . Unger, Living High and Letting Die, p. 90.
(15) . It is possible that this intuition can (and should) change when we conceive of the second trolley as a device meant to be used to stop other trolleys on which the people have inappropriately been placed. A person on skates, however, is not someone using a device meant to be sent in to stop trolleys, and so there would be no shift in the intuitive judgment that we may not send that person in front of the trolley.
(16) . However, I have also said in my response to Unger’s (4) that there is, in general, no inconsistency in thinking that an act of a certain type is impermissible in one context and that an act of the same type is permissible in another context. For example, I believe that pushing a person in front of the trolley is impermissible even if it would merely paralyze him in a twooptions case. But suppose that it is permissible for me to redirect a trolley from five people toward one, thereby killing a person on the other track. Suppose that I am about to do this, when I find out that if I push that one person on the track in front of the trolley instead, it will only paralyze him. Since it is better for him if I push him in front of the trolley than if I redirect the trolley toward him, it is permissible in this context to do what it would not have been permissible to do in another context, where I had no option of redirecting a trolley that would kill him. This is an example of the Principle of Secondary Permissibility (PSP) that I discussed in chapters 1 and 5. Unlike what is true in the Switches and Skates Case, here there is a morally important difference between doing the act in one context rather than in another because, only in one context will someone suffer less harm than he would otherwise have suffered.
(17) . This is a form of argument that Unger is famous for using in his early skeptical work in epistemology.
(18) . This is one of John Rawls’s objections. See Rawls, A Theory of Justice (Cambridge, MA: Harvard University Press, 1971), pp. 29ff.
(19) . In chapters 4 and 5, I considered a possible exception to this. In the Tree Trolley cases, it seems that the determination of who is uninvolved, in the sense of physically nonsusceptible to a threat, or involved, in the sense of physically susceptible to a threat, precedes and determines permissibility.
(21) . I owe this case to Michael Otsuka.
(22) . I should point out that my intuitions do not agree with Unger’s in some of the cases he uses to support his theory about protophysical distinctions. For example, he says that if there is a resting bomb that will explode and kill five people, we intuitively think that it is not permissible to start it rolling away from them when it will foreseeably kill an innocent bystander. By contrast, he says, we think that it is permissible to redirect a bomb that is already in motion away from five it will kill, though we foresee that it will kill an innocent bystander. But I think that it is permissible to start the first bomb moving; it is a threat to the five and may be moved away though it harms the one. It is possible, however, that if a threat (e.g., germs) were already (p.222) incorporated into a person, then ejecting it from him when this results in its being directed toward others is wrong. For example, should a doctor tell five patients to cough, when he knows this will cure them but fatally infect a bystander? This is a Threat Incorporated type of case that I briefly discussed in chapter 5.
(23) . See Unger, Living High and Letting Die, p. 98.
(27) . In his review of Unger’s book, David Lewis accepted that varying grouping and protophysics accounts for our intuitions, about what seems to be morally correct. However, contrary to Unger, he denied that these factors are not really morally relevant. See David Lewis, Eureka Street 6 (1996). I have denied that these factors account for our intuitions and also denied that they are really morally relevant.
(29) . See Unger, Living High and Letting Die, pp. 121–23.
(30) . Unger notes that intuitively we might not redirect when on balance we only maximize utility by a smaller amount than a foot. I discussed the question of what is an “irrelevant utility” in such cases (and in others as well) in detail first in F. M. Kamm, “The Choice between People, ‘Common Sense’ Morality, and Doctors,” Bioethics 1 (1987): 255–71; and then in F. M. Kamm, Morality, Mortality, vol. 1 (New York: Oxford University Press, 1993), chaps. 5–10. See also chapters 1 and 2 in this volume.
(31) . Suppose one had a case with several options but with different agents in control of each option. For example, A could redirect the trolley toward the three; B could send in the trolley with two in it; C could send in the man on skates. C knows that if he does not send in the man on skates, A will turn the trolley toward the three, unless B sends in the trolley with two. Will those who think (mistakenly, I would say) that a single agent with several options may send in the man on skates, also think that C may do so in order to prevent A and B doing worse? I doubt it, for in this case no agent is redirecting himself to do less harm.
(32) . This is consistent with the Principle of Secondary Permissibility (PSP), as I argued in chapter 5. The PSP allows that sometimes it is permissible to do something to someone in the second instance that it is not permissible to do in the first instance to the very same person. But Unger is concerned with harming other people, not the same person who would otherwise have been harmed.
(33) . Unger, Living High and Letting Die, pp. 126–28.
(34) . Hence the Lesser-Loss-Card Case is not like the following two versions of the Trolley Case: (1) I press a button that causes the trolley headed toward five people to be diverted onto a track with no one on it. Pressing the button also sends out a death ray that kills a bystander. (2)I press a button that sends out an electric current that redirects the trolley headed toward five people to a track with no one on it. The electric current also moves down another track and electrocutes the person on the track. In these cases (discussed in chapter 5), pressing the button itself causes two different things and they are cases in which a mere means to greater good causes lesser evil*.
(35) . Unger, Living High and Letting Die, pp. 139–40.
(36) . If I see my refusal as a way of stopping a plan that Strangemind already had to give aid to the stranger—and I see this as a form of making the stranger worse off than he would have been without my refusal—would I, intuitively, think that I should allow Strangemind to take my money? I do not think so. Someone’s making his continuing aid contingent on my aid cannot, by itself, obligate me. If Strangemind puts the offer as a case in which he will follow my lead (i.e., if I (p.223) am willing to lose money, he will lose money too), I also do not think that our intuitions suggest a stronger obligation to aid than we think exists in the Envelope Case. However, it is true that I can do more good when I help in these other cases, than when I aid in the Envelope Case. This is because I have the power to make someone else also aid, and so more is at stake in my refusal. This can create psychological pressure to help. (This mechanism is at work in a small way when one is told that one’s employer will match one’s charitable donations.)
(37) . In addition, our helping has a causal relation to providing someone else with a benefit in the Envelope Case; the benefit is not a noncausal flip side of our involvement. However, I doubt that the latter difference has any moral significance, for, I argue, I have no greater obligation to keep Strangemind busy (where the benefit is a noncausal flip side of his involvement with me) than to send money in the envelope.
(39) . I discuss this issue further in “Harming Some to Save Others from the Nazis” in its revised (unpublished) version. The published version is in Moral Philosophy and the Holocaust ed. E. Garrard and G. Scarre, pp. 155–168 (Hants, UK: Ashgate, 2003).
(40) . In this connection, it is interesting for me to remember an early discussion in which Unger argued for the position that one might harm others to benefit one’s family. At that time, his ground for this claim was that one had no duties to anyone but one’s family. His more recent views try to combine the claim that we have very strenuous duties to others trumped only by duties to one’s family.
(41) . He does not expect people to do what they are morally required to, but this does not show that they do not have the duty.
(42) . Unger, Living High and Letting Die, p. 155.
(43) . Suppose that I realize that if I were the one to be killed, I would decide not to redirect at all, but if someone else were the one to be killed, I would decide to redirect. Have I then violated the PEI because I respond differently to a situation as a function of whether I am in it or not? It seems so. But then I might as well violate the PEI and also reduce the total number harmed by turning only from six toward three, and not toward myself.
(44) . We noted this in chapter 5. We also noted, possibly, that the one may not use means that will stop the trolley that is redirected toward him, if as a side effect it harms the original six, unless this is a substitute for his capacity to permissibly turn back the trolley that would have done the same or more damage to the six. (This makes it secondarily permissible.) Nor may he turn the trolley away in a direction where it harms six people who were not originally threatened by the trolley. He would also not be permitted to redirect a trolley that was originally coming at him, foreseeing that it will then kill six people. Furthermore, suppose that Strangemind had redirected a threat from killing people in Africa to imposing a loss on someone’s bank account—a property damage. I do not believe that this person is permitted to redirect back Strangemind to killing people in Africa. Property loss versus mortal loss makes a difference here. This is consistent with the Bankcard Case.
(45) . In discussing Unger’s substantive and methodological views, I have tried to examine the details of his cases and arguments. This approach differs from, and yields somewhat different results from, discussions by others. For example, in her review of Living High and Letting Die, Martha Nussbaum (“Philanthropic Twaddle,” London Review of Books, September 4, 1997) concedes that Unger shows that there are irrationalities in our ordinary thoughts about aiding. (Philip Kitcher seems to agree. See his “Global Health and the Scientific Research Agenda,” Philosophy & Public Affairs  Winter 2004, pp. 36–65.) I have argued that Unger does not show this. In his review of Living High and Letting Die, Colin McGinn (“Saint Elsewhere,” New Republic, October 14, 1996) presents Unger as drawing his conclusion about a strong duty to aid distant people from a general commitment to producing the best state of affairs. But, in (p.224) fact, Unger attempts to derive his conclusions about aiding independent of that particular general commitment.
(46) . Unger, Living High and Letting Die, p. 166.
(47) . Possibly Unger could commend someone for taking the difficult step of thinking for himself, even if he gets the wrong answer when he could easily have had the right answer.
(48) . This connects with Kant’s view that a murderer wills his own death, for example, by capital punishment. For if murder is wrong, how can the fact that a murderer wills the principle that permits it, make it right to treat him according to an incorrect principle? But if consent of the victim can make an act that would otherwise be impermissible, permissible, and if willing a principle is a form of consent to acts implied by it, the consent can justify what the incorrect principle cannot justify. In Kant’s case, however, willing implies acting on rather than merely endorsing.