The New Problem of Numbers in Morality

Abstract

Discussion of the “problem of numbers” in morality has focused almost exclusively on the moral significance of numbers in whom-to-rescue cases: when you can save either of two groups of people, but not both, does the number of people in each group matter morally? I suggest that insufficient attention has been paid to the moral significance of numbers in other types of case. According to common-sense morality, numbers make a difference in cases, like the famous Trolley Case, where we must choose whether to kill a person (or persons) as a side effect of saving a greater number. I argue that recognition of the role of numbers in killing cases forces us to reassess purported solutions to the problem of numbers.

There is a well-established literature on the moral significance of numbers. This literature has focused almost exclusively on the significance of numbers in whom-to-rescue cases in which you can save one group of people or save another group, but cannot save both groups. I will argue that this literature has not paid adequate attention to a different kind of case in which numbers seem to make a difference: cases like the famous Trolley Case in which we must decide whether to kill a person (or persons) as a side-effect of saving a greater number. In this paper, I will argue that consideration of such cases gives us reason to reassess accounts of the moral significance of numbers. First, it brings out some additional implausible implications of the number sceptic’s position. Second, it raises problems for those who want to argue that numbers make a moral difference while denying that our moral obligations can depend on overall benefits aggregated over several distinct individuals.

1 The Old Problem

Consider the following classic whom-to-rescue case:

Rocks: Five people are stranded on one rock. One person is stranded on another. In your boat, you are able to reach either rock before the tide comes in, but not both. Anyone left standing on a rock at high tide will drown (Anscombe 1967, p. 17).

Most literature on the problem of numbers in morality focuses on whether the number of people in each group matters morally in cases like Rocks. What effect do the numbers have on how you should behave?

We can pick out five distinct positions. The most obvious response is that, other things being equal, you must save the greater number. If all the people involved are strangers to you, you have no reason to think that any is more deserving of extended life, and the further effects of each death will be broadly similar, then it would be wrong to save one when you could save five. At the other end of the spectrum, Taurek (1977) has famously argued that when all else is equal, we must toss a coin, giving each person an equal chance to be saved. In between, we find several alternative positions. Anscombe (1967) argues that in cases like Rocks we are not required to save the greatest number, but to do so is both permissible and intelligible. This argument is developed by Munoz-Dardé (2005). Another option is to agree with Taurek that we do need to give each individual a chance to be saved, but claim that unequal numbers require some adjustment to the procedure: when there are five on one side and one on the other, a 50:50 chance is not right.¹ Finally, some argue that we should save the greatest number when the asymmetry in numbers is very large but give equal chances when the asymmetry in numbers is small (Lawlor 2006; Sanders 1988).

All but one of these positions gives numbers some kind of moral significance in whom-to-rescue cases. For all but Taurek-style number sceptics, the number of people in each group makes a moral difference, either determining how the agent must decide how to act or, at least, providing a reason on which the agent may act. There are very few Taurek-style number sceptics. The claim that you are morally required to toss a coin, giving each group a 50:50 chance, no matter how many people are on each side, is deeply counter-intuitive. Most people think that numbers count in some way in cases like Rocks.

Nonconsequentialists face a difficult decision about how to explain the role of numbers in morality. Some versions of Nonconsequentialism do not permit appeal to the aggregated good of distinct individuals. For example, according to Scanlonian Contractualism, an action is wrong if and only if it would be forbidden by any set of rules for the regulation of behaviour that no one could reasonably reject. When considering whether a rule can reasonably be rejected we must adhere to what is known as the Individualistic Restriction: “the justifiability of a moral principle depends only on various individuals’ reasons for objecting to that principle or alternatives to it” (Scanlon 1998, p. 229). Appeals to aggregate losses suffered by a group of distinct individuals are not admissible. Scanlon states: “This feature is central to the guiding idea of contractualism, and is also what enables it to provide a clear alternative to utilitarianism and other forms of consequentialism.” (Ibid.) I will refer to those who refuse to let our moral obligations depend on costs or benefits aggregated over distinct individuals as Strong Nonconsequentialists.² How can the Strong Nonconsequentialist explain the moral significance of numbers in cases like Rocks? The Individualistic Restriction rules out any appeal to the idea that more is lost if five people die than if one person dies. Each individual stands to lose his or her life. The loss to the one if we save the greatest number is therefore just as great as the individual loss suffered by each of the five if we save the smallest number. Thus, Strong Nonconsequentialists must either provide an alternative account of the moral significance of numbers or face a stark choice between abandoning their position and accepting highly counter-intuitive claims about our obligations in cases like Rocks.

2 The New Problem

The Old Problem of Numbers in Morality is the task of explaining how and why the number of people on each rock makes a moral difference in cases like Rocks. I will argue that there is another problem of numbers in morality. The new problem arises because cases like Rocks are not the only cases where we face questions about the moral significance of numbers. Consider Judith Jarvis Thomson’s famous Trolley Case:

Trolley: Five people are trapped on the track ahead of a runaway trolley. You can turn the trolley onto a side-track where a single person is trapped. If you do nothing, the one will live and the five will die. If you turn the trolley, the five will live and the one will die (Thomson 1976, p. 207).³

In cases like Trolley, as in the Rock case, we face the question of whether the numbers count. Does it matter how many people are on each side? The commonsense view is that numbers do make a difference in such cases. Most people think that it is permissible to turn the trolley in the classic version of the case. We can redirect it towards the one to save five lives. This no longer holds when we change the numbers. It seems obviously impermissible to turn the trolley towards the one simply to save a single innocent stranger on the main track.

Such cases have not been completely neglected in the literature on numbers. Kamm (2005) gives a brief, but illuminating, discussion. Kamm’s stated aim is to “make clearer the contrast between the numbers and the anti-numbers camps”, pointing out “some implications of [the number sceptic’s] position that I think are striking” (Kamm 2005, p. 2). Although Kamm explicitly states that she does not see her observations as arguments against the number sceptic, they may be seen as a reductio ad absurdum. Kamm points out that the numbers sceptic must see the following cases as equivalent:

• (1) You turn the trolley away from five towards one, killing one to save five;

• (2) You turn the trolley away from one towards one, killing one to save one;

• (3) You turn the trolley away from one towards five, killing five to save one.

The combination of the view that numbers do not count with denial of the moral significance of the distinction between killing and letting die has particularly counter-intuitive consequences. As Kamm points out, on this view we are required to toss a coin to see whether to turn the trolley away from one towards five (2005, p. 2).

A numbers sceptic may seek to avoid this implausible conclusion by appealing to the distinction between killing and letting die, arguing that we should not turn the trolley in any of the cases because it is not permissible to kill. They could thus avoid having to say that we should toss a coin to decide whether to turn the trolley away from one towards five. Yet, as Kamm notes, this appeal does not save the numbers sceptic. Consider the following case:

Crossroads: The trolley is at a crossroads and if it remains where it is it will set off a nuclear weapon, destroying civilisation. You must turn it onto a track where five people will be killed or onto another track where one person will be killed (Kamm 2005, p. 2).

The numbers sceptic must either hold that you should allow civilisation to be destroyed or that you should toss a coin to decide whether to kill five or kill one. We might add to Kamm’s observations by noting that the number sceptic who appeals to a strong constraint against killing must hold that it is impermissible to turn the trolley towards one even when there are 500, 5,000 or 5 million people on the other track.⁴

Kamm’s discussion focuses on what cases involving killing tell us about the number sceptic’s position. However, I suggest that these cases have a wider significance. Recognition that the problem of numbers also occurs in cases involving killing has important ramifications for those who argue that numbers do count. Most importantly, it places new constraints on the adequacy of proposed accounts of how numbers count. Indeed, I will suggest that some accounts of the moral significance of numbers that seem appealing when we focus simply on whom-to-rescue are less plausible when we consider cases involving killing.

3 Implications for Attempted Solutions

There have been several ingenious attempts to solve the old problem of numbers in morality while retaining Strong Nonconsequentialism. These arguments attempt to account for the significance of numbers in whom-to-rescue cases without appealing to aggregative good. I will focus on two particularly prominent approaches: the Balancing/Tiebreaker Arguments put forward by Kamm (1985, 1993, 2005) and Scanlon (1998) and the Non-Comprehensive Reasons Argument put forward by Munoz-Dardé (2005). I will consider the implications of the new problem of numbers for these proposed solutions to the old problem.

Balancing/Tiebreaker Arguments aim to show that we are morally required to save the greatest number in whom-to-rescue cases. They are based on the idea that individuals can complain if their presence is not given the appropriate recognition in the decision about what to do. Consider the simple case in which we can save either A alone or both B and C. If A were the only person present we would save A. If A and B were the only persons present, then we would toss a coin. If we toss a coin to decide between A alone and both B and C, then we are behaving just as we would if C were not present. The fact that C is there and that he might die is ignored. In Kamm’s words: “If the presence of each individual person would make no difference, this seems to deny the equal significance of each person” (2005, 6). Failure to save the greatest number wrongs C because it fails to recognise the significance of his presence.⁵

If Balancing/Tiebreaker Arguments work, they provide an attractive solution to the problem of numbers in whom-to-rescue cases. Accepting this kind of argument allows the Strong Non-consequentialist to endorse the common sense response that we must save the greatest number for “the distinctively non-consequentialist reason that what we owe each individual is to weigh him against an opposing equal and allow the remaining person’s weight to count as well” (Kamm 2005, p. 6).

However, the Balancing/Tiebreaker approach does not seem hopeful when it comes to cases involving killing. It does not seem to be able to explain the commonsense position according to which it is permissible to turn the trolley towards one to save five, but impermissible to turn the trolley towards one to save one. If it is impermissible to turn the trolley towards one to save one, the claim not to be killed and the claim to be saved should not be seen as “opposing equals”. If the person on the side track were the only person present, we would not turn the trolley. If there was one person on the side track and one on the main track, we still would not turn the trolley. It is hard to see how a second person on the main track can argue that failure to turn the trolley does not appropriately recognise her equal claim to be saved. It does not seem as if any individual’s claim to be saved should make a difference to how we behave in this context.⁶

Kamm supplements her Balancing Argument with a Virtual Divisibility Argument, designed to accommodate whom-to-rescue cases in which the benefits are not equal. For example, suppose we can help A to live an extra 20 years or help B and C to live 12 and 8 years respectively. Kamm argues that a version of the Balancing Argument can show that we should save B and C, rather than A, in this case. We imagine creating a tie between A and B by marking off the point at which they would be tied: 12 years each. We then ask whether the additional 8 years are better allocated by saving C or by giving A additional time. Even though it is not actually possible to divide the benefits in this way, Kamm claims that it is useful to imagine what would be done in the counterfactual case in which divisibility is possible (2005, p.15-18).⁷

It might be thought that this technique could be used in the Trolley case. Suppose that A is on the side track and B, C, D, E, and F are on the main track. B’s claim to be saved does not create a tie with A’s claim not to be killed. However, most people would agree that there is some lesser harm such that we could do that harm to A to save B’s life. For example, I could turn the trolley, breaking A’s leg, to save B’s life. Perhaps we could reason as follows: A’s claim not to have his leg broken ties with B’s claim to have his life saved. Once this tie has been created, we are in a position to compare A’s remaining claim with the claims of the other persons on the side track. Unfortunately, it does not seem very likely that this technique will work. It is hard to make sense of the notion of the “remaining claim” here. What is left over from A’s claim not to be killed when we “subtract” his claim not to have his leg broken?

In Kamm’s example, she looks not at injuries, but at length of life lost. Does this approach help? It does not look like it will help to explain the permissibility of turning the trolley towards one (A) to save five (B, C, D, E and F). This approach must be based on claiming that it would be permissible to shorten A’s life by a certain amount in order to save B. To defend the permissibility of turning the trolley on one to save five, the claim must be something like the claim that it would be permissible to turn the trolley towards A in order to save B alone if A had 5 years to live and B had 25. Any such claim seems deeply implausible.

A Strong Nonconsequentialist might respond to this by simply denying that it is permissible to turn the trolley in the original Trolley Case. Some Nonconsequentialists do hold that there is a very strong constraint against killing, such that it is impermissible to kill one person even to save five others. We should then consider the Extreme Trolley Case, in which an extremely large number, say 5 million, will die if we do not turn the trolley. Even those who hold that it is impermissible to turn the trolley towards one to save five may baulk at the claim that it is never permissible to turn the trolley towards a single individual, no matter how many people are on the main track. The Virtual Divisibility Argument may look more hopeful when we focus on the Extreme Trolley Case. It does seem as if there is some length of time such that we can turn the trolley shortening A’s life by that length of time in order to save B’s life. Suppose that it is absolutely certain that A will die in 2 s. It seems that, given this, it is permissible to turn the trolley towards A to save B, even though we would thereby kill A. The length of time that A will lose must be very short, but in the Extreme Trolley Case we have very large numbers to play with.⁸

I do not think the Virtual Divisibility Argument can explain the permissibility of turning the trolley in the Extreme Trolley Case. Setting up the argument is still problematic. Let us assume that A is on the side track and B1, B2,……B50000000 are on the main track. We need to set up a series of virtual ties. To set up the first tie we would need to show that there is some length of time X_B1 such that we could take A’s last remaining X_B1 seconds to save B1. To set up the second tie, with B2, we then need to look at A’s “remaining claim”. What is this? One might well think that, if this question makes sense at all, what remains is simply the claim not to be killed. Claims not to be harmed function differently from claims to be given benefits. If you have a claim to be given 20 lbs and I give you 10 lbs, I have partially fulfilled your claim in a way that reduces what I owe to you. You no longer have a claim to be given 20 lbs. In contrast, if you have a claim that I should not steal 20 lbs from you and I do not steal 10 lbs from you, this does not partially fulfil your claim in a way that reduces what I owe to you. You still have a claim that I should not steal 20 lbs.

Even if we can set up an appropriate series of ties, doing so ignores something important about A’s loss. A does not simply lose a series of unrelated 2-second intervals. A loses a significant period of time. Two seconds is a negligible amount of time. It is very unlikely that A would be able to do anything worthwhile in those 2 s. I suggest that our intuitions about the permissibility of turning the trolley reflect this. Indeed, note that our intuitions would change if we knew that those 2 s were vitally important to A, suppose she would use them to complete a project on which she worked for her entire life. I suggest we would then be less sure that it is permissible to turn the trolley on A, depriving her of those last few seconds, to save B.⁹ Using virtual divisibility to divide the rest of A’s life into 2 s chunks is inappropriate because there is a difference in kind between the divided and the undivided loss. Another way of pursuing this thought is by arguing that the claims of B1, B2, B50000000 to be saved, when considered individually, are not appropriate virtual tie-formers. Here we are envisaging that B’s individual claim to be saved is just slightly more weighty than A’s claim not to have the last few seconds of her life taken away. We may turn the trolley on A to save B if A is about to die but not otherwise. A’s claim not to have decades taken away and A’s claim not to have a few seconds taken away are very different. It is inappropriate to use claims that are roughly equivalent to the latter to form virtual ties when the former claim is in play. In the same way, we should not use virtual divisibility to form a series of ties between a claim to be saved from lifetime chronic pain that prevents all worthwhile activity and temporary headaches of similar intensity but very short duration.¹⁰ I thus conclude that, if we assume that it is impermissible to turn the trolley towards one to save one, the Balancing/Tiebreaker Argument can explain neither the permissibility of turning the trolley towards one to save five, nor the permissibility of turning the trolley towards one to save five million.

Munoz-Dardé (2005) has put forward an alternative approach to the old problem of numbers. Developing a brief set of remarks from Anscombe (1967), Munoz-Dardé argues that in cases like Rocks we are not required to save the greatest number. Nonetheless to do so is both permissible and intelligible.

For Munoz-Dardé, the heart of the issue lies in the challenge expressed by the title of Anscombe’s paper: “Who is wronged?” Is any individual wronged if we save the smaller group? Anscombe and Munoz-Dardé suggest not. The only claim that any individual involved has is that what is needed go to him rather than being wasted. If we save the smaller group, then we use what is needed to respond to human need. We do not waste it. So we do not wrong any individual.¹¹ “So who was wronged? And if no one was wronged, what injury did I do?” (Anscombe 1967, p. 17)¹²

To sharpen this challenge, Munoz-Dardé appeals to a distinction between conclusive and merely intelligible reasons. “A conclusive reason in a given context for ϕ-ing is one which, when introduced into our deliberation, leads directly to the conclusion to ϕ rather than not” (2005, p. 195). An intelligible reason is a reason “on which one may act but need not” (p. 203): it is “not a ‘bad’ reason” (such as saving one group because they are rich and the others are poor); it is the type of reason such that others may comment on your deliberative process: “In the same situation, I would not have taken this course of action for that reason, but I understand why you did it” (p. 202). As Munoz-Dardé points out, it does seem as if our account of rationality (and thus morality) should make room for merely intelligible reasons. We do not always have to do whatever we have most reason to do: there is no obvious irrationality in feeling the attraction of eating a peach but failing to do so, even in the absence of countervailing considerations (p. 203).

Munoz-Dardé argues that, given that no individual is wronged, there are no conclusive reasons “at least within the perspective of what we owe to each of the people involved” to save one group instead of the other: “we have one and the same conclusive reason on each side, namely the human need to be saved” (p. 201). Given the lack of conclusive reasons favouring one side, we are left with merely intelligible reasons to act in one way rather than another. The fact that we might save more people is such an intelligible reason. We may save the greatest number for this reason, but we need not do so.¹³

As Munoz-Dardé points out, Anscombe’s challenge left open the possibility that we might commit a wrong without any person being wronged. However, Munoz-Dardé argues, this would be the case only if some impersonal value grounded a duty to save the greater number. “And we have seen no reason to suppose any such impersonal value grounds such a duty” (p. 203). To show that we do wrong by failing to save the greatest number, it is not enough merely to show that we have a reason to save more people. We would need to show that we have conclusive reason to do so. Munoz-Dardé does not believe that considerations of saving more people provide conclusive reasons. Thus, as before, she concludes that it is permissible and intelligible, but not required, to save the greatest number.

A slightly modified version of Munoz-Dardé’s approach will seem attractive to Nonconsequentialists who wish to admit that it can make sense to save the greatest number without admitting aggregative considerations into their core moral theory. Munoz-Dardé’s account remains open to the possibility of impersonal moral reasons that are conclusive and thus ground duties (p. 203). The Strong Nonconsequentialist version holds that conclusive moral reasons are determined by obligations which must be based on wrongs to individuals. Numbers, and other impersonal values, play a role by providing intelligible reasons to act in a certain way given that there are no such conclusive reasons to act one way rather than another. This modified version of Munoz- Dardé’s approach allows Strong Nonconsequentialists to maintain the individualistic restriction under which our moral obligations are based only on wrongs to individuals without barring numbers altogether from their moral theory.

Unfortunately, again, it does not seem likely that the Strong Nonconsequentialist version of the Non-Comprehensive Reasons Argument can explain the intuitions that it is permissible to turn the trolley towards one to save five in the Trolley Case but not permissible to turn the trolley towards one to save one. Again, let us suppose that A is on the side track and B, C, D, E, and F are on the main track. We have a conclusive reason not to turn the trolley: our reason not to kill A. If only A and B were present, it would be impermissible to turn the trolley. Thus we must assume that B’s need to be saved does not undermine the conclusive status of our reason not to turn the trolley. The fact that turning the trolley would kill A is still a conclusive reason not to turn, even given that turning the trolley would save B. The same will be true of the individual need of each of C, D, E, and F. On the Strong Nonconsequentialist version of Munoz-Dardé’s account, conclusive reasons can only be based on individual complaints. The need to minimise deaths is an intelligible but non-conclusive reason (at least when each of our actions results in some people dying who would have lived if we had acted differently). Thus there is nothing to stop our reason not to kill A from being conclusive. It does not look as if the Strong Nonconsequentialist version of Munoz-Dardé’s argument can account for the permissibility of turning the trolley when we can save five by killing one. Moreover, it does not seem as if greatly increasing the numbers on the main track would make any difference to the argument. We face just as much difficulty explaining the permissibility of turning the trolley towards one to save five million.

My argument applies to any attempt to relegate the role of numbers to a separate realm, subordinate to the “core” area of morality which is based on individual considerations.¹⁴ If we are to account for the role of numbers in cases like Trolley, we must explain how the fact that a large number of people could be saved can defeat constraints against harming. In such example, numbers seem to interact with, and indeed overcome, the core of the Strong Nonconsequentialist’s individualistic morality. The segregationalist approach cannot explain this.

Throughout this discussion I have looked at Munoz-Dardé’s argument through a narrow focus. Her argument has been heralded as a way for Strong Nonconsequentialists to account for the moral significance of numbers by allowing numbers to count but only when it comes to intelligible reasons, permitting only individual complaints into the core area of morality which generates conclusive reasons. I have suggested that Munoz-Dardé’s argument does not provide what the Strong Nonconsequentialist needs. To explain the permissibility of turning the trolley, we need to explain why our reason not to kill A is conclusive when only B could be saved but not when B, C, D, E and F could be saved. Facts about numbers must be allowed into the realm of conclusive reasons. This claim seems to me to be compatible with much of what Munoz-Dardé says. The distinction she draws between intelligible and conclusive reasons remains important, both in moral philosophy in general and in explaining the moral significance of numbers. For example, this distinction will be crucial for explaining why even if it is permissible to turn the trolley towards one to save five, the agent is not required to do so. I see consideration of killing cases as providing interesting new conclusions about the relationship between conclusive and intelligible reasons, individual complaints and overall goodness. I only criticise Munoz-Dardé’s approach insofar as I argue that it will not save the Strong Nonconsequentialist.

Could the Strong Nonconsequentialist deal with Trolley cases simply by denying that it is impermissible to turn the trolley towards one to save one?¹⁵ One might deny that there is a strong constraint against killing, holding that the claim not to be killed and the claim to be saved are morally equivalent. If this were so, the Trolley case would not differ significantly from the classic Rocks case and success in explaining the latter would imply success in explaining the former. I suggest that most Nonconsequentialists (and a fortiori most Strong Nonconsequentialists) would endorse a constraint against killing that implies it is impermissible to turn the trolley towards one to save one. I claim that a Nonconsequentialist must endorse some distinction in the doing/allowing family. Without such a distinction, we are unable to distinguish between respecting and promoting values and unable to endorse any constraints at all. A constraint against lying (even in order to reduce the number of deceptions believed) only makes sense if we can distinguish the lies you tell and the deceptions you fail to correct. A Nonconsequentialist may choose to build on an alternative distinction, such as the distinction between intended and foreseen consequences. Unless she holds that this distinction grounds an absolute constraint, she will face some analogue of the Extreme Trolley case. She will need to either embrace absolutism or explain, for example, why it is impermissible to intend harm to one to save one, but permissible to intend harm to one to save five million.¹⁶ Of course, some Strong Nonconsequentialists will be happy to endorse absolutism. They might hold that there is an absolute constraint against killing. They might deny that there is any constraint against killing and build their deontology on some alternative absolute constraint. Each of these are coherent positions. Nonetheless, if the Strong Non-Consequentialist response to the problem of numbers in cases like Trolley implies absolutism, this should be made explicit. Many will be see this as too great a price to bear.

To sum up, neither the Balancing/Tiebreaker arguments nor the Comprehensive Reasons Argument allow the Strong Nonconsequentialist to explain why it is permissible to turn the trolley towards one to save five, but impermissible to turn the trolley towards one to save one. A Strong Nonconsequentialist might deny that we can turn the trolley towards one to save five. She still faces, and is unable to use these arguments to explain, whether and why it is permissible to turn the trolley towards one to save some extremely large number, say 5 million. Holding that it is permissible to kill one to save one does not help: it leads either to the collapse of Nonconsequentialism, the same problem popping up elsewhere, or to absolutism.¹⁷

It might be suggested that I am expecting too much. The proposals I have discussed were put forward as solutions to the old problem of numbers: they are intended to explain whom-to-rescue cases not cases involving killing. I suggest that a solution to the problem of numbers that only applies to rescue cases is not fully satisfactory. In general, we prefer unified accounts which give the same kind of explanation of a given phenomenon wherever it appears. A solution to the problem of numbers that only explained their significance in rescue cases would lack the virtue of simplicity. More than this, the phenomenology of the choices in Trolley and Rocks suggests that we should seek the same kind of account for both cases. For it feels as if we use the same kind of reasoning in each case. Finally, if we have a solution that applies only to rescue cases, we will still need an account of cases like Trolley. If the Strong Nonconsequentialist cannot explain why we may turn the trolley without appeal to the fact that saving more is better she faces a difficult choice. She must either abandon her Strong Nonconsequentialism, admitting that what we are required to do can depend on costs and benefits aggregated over distinct individuals, or claim that numbers do not matter in the Trolley Case. As argued above, refusing to let numbers count in Trolley Cases has highly counterintuitive results. Thus many Strong Nonconsequentialists will respond to this dilemma by admitting aggregated benefits. Once that is done, a complex Strong Nonconsequentialist solution to the whom-to-rescue cases becomes otiose.

4 Conclusion

I have argued that discussion of the role of numbers in morality has focussed too much on whom-to-rescue cases. Numbers also seem to play a role in cases like Trolley, in which we must decide whether to kill a person (or persons) as a side-effect of saving a greater number. I argue that recognition of the role of numbers in Trolley type cases has important ramifications for the debate. These cases bring out some addition implausible implications of the number sceptic’s position. They also have significant ramifications for Strong Nonconsequentialists who want to argue that numbers do count without letting our moral obligations depend on costs or benefits aggregated over several distinct individuals. Accounts of the role of numbers in morality that seem to offer attractive Strong Nonconsequentialist solutions to whom-to-rescue cases face problems in explaining the role of numbers in killing cases.

I have not argued that no Strong Nonconsequentialist solution to the problem of numbers is possible. Instead, I have shown that cases like the Trolley Case make the task of producing such a solution more difficult. I have also shown that the position of the number sceptic has additional implausible implications. In my view, these considerations provide some support for an alternative solution to the problem of numbers in morality. I suggest they support the obvious answer: we should/may save the five because it is so much worse if five people die than if one dies; overall value is one of the things that matter morally and in certain circumstances considerations of overall value may over-ride individual complaints. However, I shall not argue for that view here.