¹ For evidence that terrorism is generally ineffective see Abrahms, Max, ‘Why Terrorism Does Not Work’, International Security 31 (2006), pp. 42–78CrossRefGoogle Scholar. (But for evidence that it can under certain circumstances be effective see Gould, Eric and Klor, Esteban, ‘Does Terrorism Work?’, The Quarterly Journal of Economics 125 (2010), pp. 1459–510CrossRefGoogle Scholar.)

² For instance, Gregory McNeal reports that the US military's ‘collateral damage methodology’ for pre-planned operations involving air-to-surface weapons and artillery seeks to ensure a probability of less than 1/10 of serious or lethal wounds to non-combatants, and in practice has assured that less than 1 per cent of such operations resulted in collateral damage. See McNeal, ‘The U.S. Practice of Collateral Damage Estimation and Mitigation’ (Pepperdine Working Paper, 2011), and ‘Targeted Killing and Accountability’, Georgetown Law Journal 102 (2014), pp. 681–794, at 730–54.

³ John Mikhail reports that, of a diverse group of subjects, 92 per cent say it's wrong to perform the transplant and 90 per cent say it's permissible to divert the trolley. See Mikhail, ‘Universal Moral Grammar: Theory, Evidence, and the Future’, Trends in Cognitive Science 11 (2007), pp. 143–52CrossRefGoogle ScholarPubMed. The trolley problem originates with Judith Thomson, building on a discussion of Philippa Foot. See Thomson, ‘Killing, Letting Die, and the Trolley Problem’, The Monist 59 (1976), pp. 204–17CrossRefGoogle ScholarPubMed; Thomson, ‘The Trolley Problem’, Yale Law Journal 94 (1985), pp. 1395–415CrossRefGoogle Scholar; and Foot, ‘The Problem of Abortion and the Doctrine of Double Effect’, Oxford Review 5 (1967)Google Scholar, reprinted in her Virtues and Vices (Oxford, 2002), pp. 19–31, at 23 and 27–8.

⁴ This is Foot's ‘Rescue II’ case. See Foot, ‘Killing and Letting Die’, Abortion: Moral and Legal Perspectives, ed. Garfield, J. L. and Hennessey, P. (Amherst, 1984), pp. 177–85Google Scholar.

⁵ That is, we might think (roughly) that moral reasons against inflicting harm are intrinsically stronger than moral reasons to prevent harm, where reasons against inflicting harm are of sufficiently greater strength that our reasons not to kill one decisively outweigh our reasons to save five. There are ways of believing in something like this without believing in deontological constraints. One might be a consequentialist who believes that it's intrinsically worse for the world to contain inflictions of harm than failures to prevent harm, or a virtue ethicist who believes that any fully morally admirable person would be more intrinsically averse to inflicting harm than allowing harm. But the dialectic of the debate over the trolley problem would be identical for these views: it would simply be a debate about which outcomes to value intrinsically or which people to morally admire.

⁶ See Unger, Living High and Letting Die (New York, 1996), ch. 4, esp. pp. 96–101Google Scholar (for a similar suggestion see Thomson, ‘Turning the Trolley’, Philosophy and Public Affairs 36 (2008), pp. 359–74, at 374)CrossRefGoogle Scholar. Unger himself thinks that projective grouping helps us reach more accurate moral judgements, and that ‘projective separating’, or seeing some people as ‘not having the same problem’, causes moral distortions. But Unger offers no argument for this; he never even considers questioning the intuition that it's permissible to divert the trolley.

⁷ See Kagan, ‘Does Consequentialism Demand too Much? Recent Work on the Limits of Obligation’, Philosophy and Public Affairs 13 (1984), pp. 239–54Google Scholar.

⁸ See Kagan, The Limits of Morality (Oxford, 1989)Google ScholarPubMed; Fischer, J. M., ‘Tooley and the Trolley’, Philosophical Studies 62 (1991), pp. 93–100CrossRefGoogle Scholar; Fischer, ‘Thoughts on the Trolley Problem’, Ethics: Problems and Principles, ed. Fischer, J. M. and Ravizza, M. (Fort Worth, 1992), pp. 308–17Google Scholar; Alexander Friedman, ‘Minimizing Harm: Three Problems in Moral Theory’ (PhD Dissertation, MIT, 2002).

⁹ Thomson, ‘Turning the Trolley’, pp. 364–7. On p. 366 Thomson presents the third step by saying that unless we would be willing to divert the trolley into ourselves in the three-option case, we cannot decently regard ourselves as entitled to divert it into Bugsy in the two-option case. But on p. 367 she makes it clear that the argument is not supposed to depend on what we are willing to do. While Thomson never explicitly explains how the third step is supposed to go given this clarification, it seems that she is relying on what Unger calls a ‘second-order intuition’ that it's just obvious that it shouldn't matter to the permissibility of choosing option 2 whether option 1 is present.

¹⁰ Cf. Fitzpatrick, William, ‘Thomson's Turnabout on the Trolley’, Analysis 69 (2009), pp. 636–43, at 641CrossRefGoogle Scholar.

¹¹ Thomson, ‘Turning the Trolley’, pp. 364–5. Cf. Unger's ‘Principle of Ethical Integrity’ in his Living High, pp. 139–43. Unger, however, seeks to argue from the intuitive permissibility of imposing losses on others by diverting trolleys to the claim that giving to Oxfam is no mere good deed, in that failing to do so is seriously morally wrong. Since earlier in his book Unger (following Singer, Peter, ‘Famine, Affluence, and Morality’, Philosophy and Public Affairs 1 (1972), pp. 229–43Google Scholar) makes an extremely strong independent case that we are morally required to donate most of our resources to organizations like Oxfam, Thomson's analogy is less than ideal.

¹² See Fitzpatrick, ‘Thomson's Turnabout’, pp. 638–9 and Thomson, ‘Killing, Letting Die, and the Trolley Problem’.

¹³ For a review see Brewer, Marilynn, ‘In-Group Bias in the Minimal Intergroup Situation’, Psychological Bulletin 86 (1979), pp. 307–24CrossRefGoogle Scholar. Psychologists have documented in-group bias in response to such trivial similarities as a shared tendency to over- or under-estimate the number of dots flashed on a screen.

¹⁴ See Singer, ‘Sidgwick and Reflective Equilibrium’, Monist 58 (1974), pp. 490–517, at 515–17CrossRefGoogle Scholar.

¹⁵ I believe that once we fully appreciate what the differences between the cases amount to and how irrelevant they look, it is clear not only that they cannot be of underivative moral significance, but that there is just as little hope of deriving their non-instrumental significance from distinct factors as there is of doing so for skin color and salience. (There are certainly differences which are of derivative instrumental significance – such as differences which make the benefits of harvesting in Transplant less certain in the actual world than those of diverting in Trolley – but these are not at issue.) This is what I mean when I claim that these factors are obviously devoid of intrinsic moral significance. I am grateful to a referee from Utilitas for raising the issue of derivative non-instrumental significance, and for encouraging me to clarify the sense in which I take my argument to be ‘conclusive’ (or at least more conclusive than its predecessors).

¹⁶ Cf. Unger, Living High, p. 93. It is important that the individual who dies if you take option 2 is distinct from those who die if you take option 1, since otherwise we might plausibly reason that ‘we are already allowed to leave him badly off, so it can't be wrong to do something to him that leaves him no worse off’ (see Kamm, Frances, Intricate Ethics (New York, 2007), pp. 169–73CrossRefGoogle Scholar, which makes trouble for Friedman's arguments from multiple option cases in his ‘Minimizing Harm’).

¹⁷ I am indebted to a referee from Utilitas for raising the objection that multiple-option cases might obscure the relevance of certain factors, which I believe is a real problem with Unger's reliance on unsupported intuitions about permissibility in such cases. My focus on the plausibility or implausibility of citing factors as justifications (once we understand what they amount to) is a respect in which I take my argument to be more conclusive than Unger's.

¹⁸ This strategy is doubly conservative. First, since the whole moral difference would have to be distributed among the eleven steps, if less than 1/11 of the difference were present at any step, more would have to be present at others. Second, if we subscribe to the Prioritarian idea that we should be more concerned with decreases in well-being the worse off the individual suffering them will be (see Parfit, Derek, ‘Equality and Priority’, Ratio 10 (1997), pp. 202–21CrossRefGoogle Scholar), then decreasing someone's future years from seventy-seven to seventy may have less than 1/11 of the moral importance of decreasing someone's future years from seventy-seven to zero.

¹⁹ One might worry that my argument has the form of a sorites (a bazillion grains of sand make a heap, the subtraction of a single grain from a heap can't make a non-heap, so 1 grain of sand is a heap), and is unsound for whatever puzzling reasons sorites arguments are unsound. But my argument involves no continuous dimension of variation, and eleven steps seem far too few to generate a sorites. To the extent that there is a sorites-esque worry, it seems to be that several small moral differences between Trolley and Transplant are diffusely distributed among the steps I consider, which I address by examining whether 1/11 of the difference is present at any step.

While my broader aim is to establish the stronger conclusion that there are no intrinsic moral differences of any kind between diverting in Trolley and harvesting in Transplant, my tests for 1/11 of the difference between permissibility and wrongness most directly support the weaker conclusion that there are no intrinsic moral differences which make diverting permissible but harvesting wrong. One might be inclined to accept only the weaker conclusion if one thought some differences between the acts were intrinsically important, but insufficient to make the difference between permissibility and impermissibility. To test for this, I invite the reader to consider whether any of the differences I discuss could justify any additional risk or harm less than a 1/11 chance of death or deprivation of seven years. One might also be inclined to accept only the weaker conclusion if one thought that the salvation of five lives was insufficient to justify either diverting in Trolley or harvesting in Transplant, but that more good would be needed to justify harvesting Bugsy's organs than diverting a trolley into him. To test for this, I invite the reader to increase the number of children who could be saved by inflicting harm in the cases under consideration from five to ten, one hundred, or whatever large number one wishes (see for instance my discussion of using this test to argue against certain of Warren Quinn's views in n. 51). I submit that my remarks below, reinterpreted to apply to these tests, support the stronger conclusion. I am grateful to a referee for Utilitas for drawing my attention to the importance of inviting readers sceptical of the stronger conclusion to consider these alternative tests.

²⁰ If you think it matters that the parked trolleys be not merely overdetermining causes of the deaths of the victims (cf. Kamm, Intricate Ethics, pp. 143–4, 159–60), suppose that the oncoming trolleys lack sufficient momentum to crush the victims, but kill by creeping onto and depressing platforms located directly in front of the parked trolleys, which send the parked trolleys forward over the victims. Obviously this too makes no moral difference.

²¹ Thomson, ‘The Trolley Problem’, p. 1408.

²² Mack, Eric, ‘Three Ways to Kill Innocent Bystanders’, Social Philosophy and Policy 3 (1985), pp. 1–26, at 17CrossRefGoogle Scholar; and Fitzpatrick, ‘Thomson's Turnabout’. See also Montmarquet, James, ‘On Doing Good: The Right and the Wrong Way’, The Journal of Philosophy 79 (1982), pp. 439–55, at 446–55Google Scholar; and Hanna, Robert, ‘Morality De Re’, Ethics: Problems and Principles, ed. Fischer, J. M. and Ravizza, M. (Fort Worth, 1992), pp. 318–36, at 328–32Google Scholar.

²³ Thomson, ‘Killing, Letting Die, and the Trolley Problem’, p. 215.

²⁴ Michael Costa claims that ‘redirecting an existing threat . . . is not . . . something that is evil in and of itself’ but ‘bring[ing] some new threat or source of foreseeable harm into the world . . . is itself a prima facie evil’ (Costa, ‘Another Trip on the Trolley’, Southern Journal of Philosophy 25 (1987), pp. 461–6, at 464CrossRefGoogle Scholar). But surely neither pulling a switch nor pushing a button is ‘evil in and of itself’ in the standard sense of ‘independent of its consequences’, and it is entirely unclear in what sense or why one of these acts but not the other would be ‘prima facie evil’ when they both have the same consequences of disconnecting the first track and connecting the second.

²⁵ Kamm, Intricate Ethics, pp. 140–4. On Kamm's view this difference might not forbid diverting in Different Sections, because the harm this causes might count as ‘indirect’ (see discussion of Dual-Action Bomber below).

²⁶ The events are totally separated from each other in time: one begins and ends at t, while the other begins later at t + 1 and lasts all the way to t + n. The events are also completely separated in space: one takes place on a bit of trolley track some way away from the five, while the second takes place in and around the five and everywhere they go (home, work, the park, etc.).

²⁷ Kamm claims that ‘when one person's not being threatened is the non-causal flip side of sending the threat elsewhere’ anyone threatened ‘occupies the very same position that another person would have occupied relative to the threat’, so ‘this is substitution’ (165). But this applies only to a distinction between diverting threats and creating new ones. Kamm's distinction is supposed to apply more broadly; it is, for instance, supposed to entail that it's permissible to move five people away from a trolley, even though this causes them to tumble down a hill and crush one (Kamm's ‘Tumble Case’, 139). Kamm also claims that when we harm someone ‘as a causal means to save someone else from a threat . . . the position he occupies . . . makes essential reference to his usefulness to achieving a good for that other person’ (165). But it is simply untrue in many cases of inflicting harm that Kamm's distinction is supposed to condemn that the victim is herself a useful means of achieving goods. Take Foot's Rescue II, where the only way to save five drowning swimmers is to drive over one individual trapped in the road. Here the effect you have on the one does nothing to help you save the five – she is in no way useful to achieving goods for them.

²⁸ John Taurek might complain that it is unfair to Bugsy that he be abandoned simply because he is in the less numerous group (Taurek, ‘Should the Numbers Count?’, Philosophy and Public Affairs 6 (1977), pp. 293–316Google ScholarPubMed). If you are moved by this thought then any time I speak of the permissibility of saving the five assume that I mean the permissibility of saving them after having flipped a fair coin to determine whom to save. There is still a debate to be had over the trolley problem, because intuitively it is not permissible to flip a coin to determine whether to harvest Bugsy's organs in Transplant (cf. Friedman, ‘Minimizing Harm’, pp. 162–3). For good reasons not to be moved by Taurek's thought, see e.g. Parfit, ‘Innumerate Ethics’, Philosophy and Public Affairs 7 (1978), pp. 285–301Google Scholar.

²⁹ Of course, if it is for independent reasons permissible to save the tool rather than Bugsy but wrong to divert the trolley from the tool into Bugsy (say because it's harder to justify inflicting harm than failing to prevent harm), we might truly say that in the first case but not the second you save the five by wronging Bugsy, and in this sense ‘subordinate’ him to the five. But whether you ‘subordinate’ someone in this sense is explained by, and cannot itself explain, whether you wrong her.

³⁰ See Moore, Causation and Responsibility (Oxford, 2009), pp. 75–6Google ScholarPubMed; Unger, Living High, p. 101.

³¹ Perhaps you can ‘blame the trolley’ for its being true that someone will die in Trolley, but you can just as easily ‘blame the trolleys and the effects of the button’ for the fact that someone will die in Different Sections, or even ‘blame the organ failures and the inability of Bugsy and the five to live without organs’ for the fact that someone will die in Transplant.

³² This might be suggested by Hanser, Matthew, ‘Intention and Accident’, Philosophical Studies 98 (2000), pp. 17–36CrossRefGoogle Scholar. See discussion of The Big Bomb below.

³³ This might be suggested by Kamm, Intricate Ethics, p. 149. But Kamm seems to place relevance on the distinction between causes of harm that are ‘introduced’ as opposed to ‘already present’ only when they cause harm ‘directly’ (see discussion of Dual-Action Bomber 2 below).

³⁴ Kamm, Intricate Ethics, p. 149. Note that because the reinforced concrete would otherwise protect Bugsy in Dual-Action Bomber, the basement chemicals are a necessary (not just overdetermining) ‘direct’ cause of Bugsy's death, and because the chemicals are absent in Dual-Action Bomber 2, the rear bombs are a necessary (not just overdetermining) ‘direct’ cause of his death (cf. n. 20).

³⁵ Kamm, Intricate Ethics, p. 166. It should be obvious, however, that this criterion of the ‘transmission’ of the importance of the five to our means fails to distinguish between any cases in which what we do will result in the survival of the five and the death of the one. Since ‘we should not allow ourselves’ to fail to save one individual at relatively trivial cost to ourselves ‘if a greater good were not at stake’, we can conclude that the weight of the greater good of saving the five ‘is being transmitted to elevate’ our means of saving five ‘relative to the importance of saving a person’, thus ‘subordinating’ the one we fail to save to the five we do save.

³⁶ Hanser, ‘Intention and Accident’, esp. pp. 22–4, 26–7, and 31–2 (including n. 19).

³⁷ On the absence of a salience-independent metaphysical criterion of what (short of a full causal history) is ‘sufficient to explain’ an event – just as there is no salience-independent metaphysical criterion of what is ‘the cause’ of an event – see the various contributions to Causation and Counterfactuals, ed. J. Collins, N. Hall and L. A. Paul (Cambridge, MA, 2004).

³⁸ Fitzpatrick, ‘The Intend/Foresee Distinction and the Problem of “Closeness”’, Philosophical Studies 128 (2006), pp. 585–617CrossRefGoogle Scholar, esp. 594–7 and 599–600.

³⁹ Fitzpatrick probably had in mind cases in which you destroy an entire area, not because your munitions cannot destroy one part without destroying another, but because you cannot tell where in the area your target is, so to be sure of destroying it you set out to destroy the entire area; cf. Thomas Nagel's description of area attacks during the Vietnam War (Nagel, ‘War and Massacre’, Philosophy and Public Affairs 1 (1972), pp. 123–44, at 130–1Google Scholar). In this sort of case, unlike The Big Bomb, destroying the entire area does serve your purpose of destroying your target.

⁴⁰ Frey, ‘The Doctrine of Double Effect’, A Companion to Applied Ethics, ed. Frey, R. G. and Wellman, C. (Oxford, 2003), pp. 464–74, at 467Google Scholar. I suspect that driving your car and wearing your clothes causes and does not constitute the damage in virtue of which they wear out, and that flattening your skull causes and certainly does not constitute the distinct states of your neurons dying or losing their capacity to support mental life. Still, I think you can intend one state of affairs and merely foresee another even when the first constitutes the second. Suppose I buy a six-pack of pop, and intend to drink one pop each day for the next six days. While my drinking a pop on each of the next six days appears to constitute my running out of pop, I think that I foresee running out, but do not intend to run out, of pop.

⁴¹ Fitzpatrick's views on when a state of affairs constitutes someone's being ‘seriously injured’ are not spectacularly clear. He suggests that someone's being blown up or having her skull crushed constitutes her ‘being seriously injured or killed’, while he claims that in cutting an attacker in self-defence, her being sufficiently injured so as to be distracted does not constitute her being killed – or apparently even ‘seriously injured’ – although her ‘death . . . may foreseeably follow later (perhaps because there is no medical aid available to stop the bleeding)’ (604–7). Fitzpatrick suggests that his categorization is guided by whether the victim's dying is identical to ‘or at least . . . nearly coincident’ with some event that the agent believes causally necessary to her end. But whether what you need is ‘nearly coincident with’ the victim's death seems to amount to no more than whether her death occurs at roughly the same time as what you need, which is clearly of no intrinsic moral significance. Fitzpatrick acknowledges that there could be a spectrum of cases between crushing someone's skull and cutting her, and claims that there will be grey areas and matters of degree, but it is difficult to understand how he can suggest that these correspond to grey areas and differences of degree in our intrinsic moral reasons against inflicting harm. Permissive intuitions about killing in self-defence are misleading, since they may reflect the plausible idea that deontological constraints against inflicting harm do not apply to harming attackers. Apart from this, inflicting small cuts seems easier to justify because in real life they are much less likely to lead to death than crushing someone's skull or blowing someone up – there is always some (usually excellent) chance that blood loss or infection will not prove fatal. Absent these distractions, it seems obvious that there is no intrinsic moral difference between intending those states of affairs Fitzpatrick categorizes as constituting ‘seriously injuries’ and intending those he categorizes as constituting lesser injuries that you foresee with equal certainty will lead to death.

⁴² Dropping the bombs will cause the crew to believe that you have killed a child they care about in a way that resembles Edmund Gettier's examples of justified true beliefs that do not constitute knowledge (Gettier, ‘Is Justified True Belief Knowledge?’, Analysis 23 (1963), pp. 121–3CrossRefGoogle Scholar). Cases like this are discussed by Mack, ‘Three Ways to Kill’, pp. 7–9; Fischer, J. M., Ravizza, M. and Copp, D., ‘Quinn on Double Effect: The Problem of “Closeness”’, Ethics 103 (1993), pp. 707–25, at 720–2CrossRefGoogle Scholar; Friedman, ‘Minimizing Harm’, p.169 n. 9; Fitzpatrick, ‘The Intend/Foresee Distinction’, pp. 612–13. Remarkably, Fitzpatrick suggests that on his account the Gettier terror bomber counts as intending Bugsy's death. But clearly he does not. All the Gettier terror bomber needs to save the five is the flash from the explosion some tens or hundreds of metres above Bugsy's head; the movements of bomb fragments towards Bugsy's location, let alone their wreaking havoc at that location, is nothing to his purpose. So all the Gettier terror bomber need intend is the exploding of the bombs tens or hundreds of metres above Bugsy's head, but this state of affairs – just like that of the trolley being diverted in Trolley – is obviously a mere cause and in no way ‘constitutive’ of Bugsy's dying or being injured.

⁴³ One might argue that saving the five by demoralizing the crew is intrinsically harder to justify than saving them by activating the photo-sensor, because the first act involves ‘terrorism’ while the second does not. But I do not see how the mere fact of accomplishing one's goals by lowering the crew's morale could make a significant moral difference. The emotional harm they would suffer would be trivial in comparison to the lives of the five, and should presumably be discounted by the fact that they are culpable attackers. This said, it is not essential to my argument to consider cases that involve saving the five by demoralizing their attackers. I do so primarily to compare the causation of ‘collateral damage’ with terrorism. But my main points can be made by comparing the causation of ‘collateral damage’ with organ harvesting. So if you feel that there is a significant intrinsic difference between demoralizing the crew and activating the photo-sensor, feel free to reinterpret all my ‘Terror Bombing’ cases as involving mechanical devices (like photo-sensors) as opposed to demoralization.

⁴⁴ Warren Quinn reports the attribution of this example to David Lewis. See Quinn, ‘Actions, Intentions, and Consequences: The Doctrine of Double Effect’, Philosophy and Public Affairs 18 (1989), pp. 334–51, at 343 n. 16Google ScholarPubMed.

⁴⁵ Quinn, ‘Double Effect’, pp. 341–4. Mack also seems to propose essentially this version of the DDE (‘Three Ways to Kill’, pp. 7–9).

⁴⁶ Bennett, The Act Itself (Oxford, 1995), pp. 218–21.

⁴⁷ It could very easily have been that this didn't cause Bugsy's death – the Lewissian terror bomber might have had evidence that, unbeknownst to the crew, Bugsy would be protected by a layer of reinforced concrete, but received last-minute intelligence that the concrete had actually been compromised.

⁴⁸ Quinn was aware of this issue. After giving his formulation in terms of intended property instantiations (or ‘harmful involvement’), he explicitly distinguished between actions that take advantage of the presence of a victim (‘direct opportunistic agency’) like Lewissian terror bombing, and actions that aim ‘to remove an obstacle or difficulty that the victim presents (direct eliminative agency)’ like blasting the fat man out of the cave, and held that the DDE ‘strongly discriminates against’ the first kind of actions and ‘more weakly discriminates against’ the second (344). (Incidentally, the only intuition Quinn used to motivate this was that it doesn't seem much harder to justify saving a mother's life by crushing her foetus's skull than fatally removing her foetus along with her uterus, which is complicated by issues of foetal moral status and the mother's right to control her body). Quinn conceded that ‘it is less plausible . . . to think of the victims of direct eliminative agency as used’, but insisted that ‘Someone who gets in your way presents a strategic problem’, and that this rationalized its being somewhat harder to justify direct eliminative agency than harming victims without intending them to instantiate any properties (‘indirect agency’).

But once we dismiss Quinn's obscure claims that to intend that someone instantiate a harmful property you must ‘see her as material to be strategically shaped’ or ‘treat her as existing for your purposes’, all we are left with are his remarks about victims being used. And these, it seems, do not support even a weaker form of discrimination against direct eliminative agency, since, as Bennett observes, it is not just less plausible, but entirely implausible to say that in eliminating victims who pose obstacles we use them as means (220). All these victims do is ‘ennoble’ or raise the status of your eliminating them to that of a cause of the survival of those you save, and as Stephen Yablo observes, things that simply make it necessary for other things to happen if an event is to happen are certainly not causes of that event or means of bringing it about (Yablo, ‘Advertisement for a Sketch of an Outline of a Prototheory of Causation’, Causation and Counterfactuals, ed. Collins, J., Hall, N. and Paul, L. A. (Cambridge, MA, 2004), pp. 119–37Google Scholar, at 120).

⁴⁹ You can implicitly undertake to tell the truth or provide information simply in virtue of conventional norms governing assertions, expressions of interest, or acceptances of offers. On the reduction of moral reasons not to lie to moral reasons to honour implicit undertakings, see Ross, W. D., The Right and the Good (Oxford, 1930), p. 21Google Scholar; McNaughton, David, ‘An Unconnected Heap of Duties?’, The Philosophical Quarterly 46 (1996), pp. 433–47, at 437CrossRefGoogle Scholar. For a similar account of the complaint that ‘you were just using me!’ and the wrongness of such using in terms of violations of obligations to provide information, see Scanlon, T. M., Moral Dimensions: Permissibility, Meaning, Blame (Cambridge, 2008), pp. 106–21CrossRefGoogle Scholar.

⁵⁰ See Shoemaker, Sydney, Self-Knowledge and Self-Identity (Ithaca, 1963), p., 22Google Scholar; Olson, Eric, The Human Animal: Personal Identity without Psychology (Oxford, 1997), pp. 42–6Google Scholar. Even Olson, who defends the view that we are organisms, does so by divorcing numerical identity from what matters ethically, and suggests that we might use something like the notion of ‘the same person as Bugsy in the practical sense’ to describe what I am describing as ‘Bugsy as a subject of ethical concern’ (ch. 3, esp. pp. 65–71). On Olson's view, we might have to admit that properties of Bugsy's body are properties of Bugsy, but since our subject of ethical concern with Bugsy is not really Bugsy himself (an organism) but Bugsy*, the welfare subject who inhabits Bugsy, they would not be properties of our subject of ethical concern.

As Olson (following authors like Shoemaker and Derek Parfit) observes, cases of fission give us good reason to think that the question of what matters in Bugsy's survival is not the same as the question of what happens to entities to which Bugsy is numerically identical. (See Shoemaker, ‘Persons and their Pasts’, American Philosophical Quarterly 7 (1970), pp. 269–85, at 284Google Scholar; Parfit, ‘Personal Identity’, Philosophical Review 80 (1971), pp. 3–27CrossRefGoogle Scholar. It seems as good as survival for Bugsy for half of his brain to be transplanted into two bodies, resulting in two individuals with whom Bugsy is fully psychologically continuous. But we cannot say that Bugsy is identical to one but not the other of these individuals without being arbitrary, and we cannot say that he is identical to both without the absurd consequence that they are identical to each other.) But I think that Olson's total divorce of identity from what matters in survival goes too far. Following Jeff McMahan, I think it best to tie the account of our numerical identity as closely as possible to what matters in our survival – roughly, to a non-branching, minimal degree of what matters in survival (McMahan, The Ethics of Killing (New York, 2002), pp. 51–5Google ScholarPubMed).

⁵¹ Quinn himself probably would have thought it wrong to save five by Gettier or Lewissian terror bombing Bugsy, since he defended a version of the Doctrine of Doing and Allowing (DDA) on which the greater difficulty of justifying doing harm in relation to allowing harm makes it wrong to kill one to save five (Quinn, ‘Actions, Intentions, and Consequences: The Doctrine of Doing and Allowing’, Philosophical Review 98 (1989), pp. 287–312CrossRefGoogle Scholar). Presumably, he thought strategic bombings that kill innocents were permissible, while terror bombings with the same (expected) ‘consequential profile’ were wrong, only when the harms they could be expected to prevent were more than five times as great as the harms they could be expected to inflict. While my immediate aim is to argue against using Quinn's DDE to solve the trolley problem, my broader aim is to argue that there is no intrinsic difference between ‘collateral damage’ and directly inflicted harm. I thus consider making the five children into ‘a bazillion’ – a number such that it is morally permissible to kill one individual to save that number of individuals.

(Incidentally, Quinn believed – falsely – that his version of the DDA treated the choice of whether to divert in Trolley as a choice between killing Bugsy and killing five, and thus permitted diverting. On Quinn's version of the DDA, you count as doing harm if harm is caused by your actions or actions of objects or forces under your control that you deliberately fail to prevent. Quinn claimed that if someone fails to divert the trolley, ‘it is because he intends the [trolley] to continue in a way that will save [Bugsy]. But then he intends that the [trolley] continue forward past the switch, and this leads to the death of the five’ (‘Doing and Allowing’, p. 305). But in failing to divert the trolley to save Bugsy, you need only intend that it not continue in a way that threatens him. That it continues at all, and certainly that it continues past the switch, is nothing to your purpose of saving Bugsy – he would be saved equally well if the trolley simply stopped or fell off its track before reaching the switch. So in failing to divert the trolley to save Bugsy you need not intend that it continue, so Quinn's DDA counts you as merely allowing the five to die.)

⁵² Only part of the absurdity stems from the fact that, as Thomson observed, the DDE as it is often understood implausibly tells us to ‘look inward’ and determine what our intention in doing something would be to decide whether to do it (Thomson, ‘Self Defense’, Philosophy and Public Affairs 20 (1991), pp. 283–310, at 293Google ScholarPubMed). Absurdity would remain if Quinn claimed that the explanatory relationship between harmful property instantiations and benefits itself matters intrinsically. Suppose you were about to engage in what you thought was Gettier terror bombing, when you received allegedly urgent news: ‘Wait! Stop! The crew found out that Timmy is gone and Bugsy is in the hospital! If you explode the bombs, the explosion's being over Bugsy's body will play a role in explaining the salvation of the five (or bazillion)!’ This reasoning also seems ridiculous.

⁵³ Mack proposed a ‘Doctrine of Antecedent Causation’, according to which ‘When. . .inevitably injurious forces confront a person such that, no matter how that person acts, some nonaggressor(s) will be injured, the antecedent perilous forces bear the predominant causal responsibility for the subsequent injuries – unless those injuries are intended by a person whose action is an intermediate cause’ (‘Three Ways of Killing’, p. 17). Although Mack speaks of ‘injuries’ being intended, his discussion of a Quinn-like way of getting the DDE to condemn acts like Lewissian terror bombing makes one suspect that he really means ‘unless properties of nonaggressors in virtue of which they will be injured are intended by a person whose action is an intermediate cause’. Mack's suggestion would then be that if you kill Freddy and Suzy, the situation was primarily responsible for their deaths, but if you kill Bugsy, it is you rather than the situation that was primarily responsible for his death. But reading Mack's Doctrine this way makes it straightforwardly preposterous. There is no reason at all why (and it sounds self-evidently ludicrous to assert that), in killing someone by exploding bombs over him, you become primarily responsible for his death if you intend that they explode over his body, but your situation bears primary responsibility if you intend that they explode over the building in which he is trapped knowing full well that they are in fact exploding over his body.

⁵⁴ Bennett, ‘Morality and Consequences’, The Tanner Lectures on Human Values, vol. 2, ed. McMurrin, S. (Salt Lake City, 1981), pp. 110–11Google Scholar; and The Act Itself, pp. 210–13). In his more recent discussion, Bennett suggests that, to avoid counter-intuitive results, we should count agents as intending as means not only those states of affairs that they believe are causally necessary to achieving their ends, but other states that ‘we have not the faintest idea what it would be like to have the means’ of separating from those needed to achieve their ends. Thus, while Bennett concedes the possibility of Bugsy's flying into small pieces without dying, he suggests that because we cannot easily imagine a technology for reassembling Bugsy, a standard terror bomber who must demoralize the crew by causing them to see Bugsy fly into small pieces should count as intending Bugsy's death as a means.

Of course, Bennett would be the first to note that how easy it is for us to imagine something lacks intrinsic moral significance, and there are variants on which Bennett's criterion allows the standard terror bomber to foresee rather than intend Bugsy's death (if he knows that the bombs will render Bugsy such that he could be resuscitated by the best contemporary medical technology, but that the crew lacks this technology). But I think it is a mistake to abandon the view that agents intend as means only those states of affairs they believe causally necessary to achieving their ends. If an agent cannot easily imagine two states coming apart, it is likely that he conceives of them as the same thing, believes them both necessary, and thus intends them both as means. But if an agent knows the states are distinct, I see nothing wrong with saying that she intends one but not the other. A standard terror bomber cannot truly deny that he intended a destructive or fatal effect on Bugsy's body, but he can truly say that he foresaw but did not intend his permanently eliminating Bugsy or causing Bugsy to never again experience goods (in which the harm of death consists). Of course it would be preposterous to use this as a justification or excuse for killing Bugsy, but that, I suspect, is because inflicting foreseen harm is just as difficult to justify as inflicting intended harm.

⁵⁵ See Shaw, Joseph, ‘Intentions and Trolleys’, Philosophical Quarterly 56 (2006), pp. 63–83, at 69–71CrossRefGoogle Scholar; Mikhail, ‘Universal Moral Grammar’, p. 145; Mikhail, Elements of Moral Cognition (New York, 2011), pp. 133–6, 148–52Google Scholar. In places Shaw and Mikhail seem to speak of an absolute prohibition of all intended assaults or batteries, or at least all ‘serious’ (as opposed to ‘trivial’) such assaults or batteries. But what makes an injury or battery morally ‘serious’ is not the dramaticality of the bodily damage it involves, but its probability of proving lethal or otherwise depriving the victim of intrinsic goods or subjecting her to intrinsic bads like pain. It is just as difficult to justify giving someone a small cut as blowing her body to bits if (as is almost never actually true) it is just as sure to kill her and she is just as sure to survive if you don't do it (cf. n. 41). Similarly, if you could anaesthetize Bugsy before blowing his neck-down body to bits and, as soon as this causes the missile crew to surrender, quickly reassemble his body so that he awakens moments later without noticing a thing, there could be no serious moral objection to doing so.

⁵⁶ Again, the absurdity is not simply that Mikhail's DDE tell us to ‘look inward’ to determine what to do. Suppose you thought that the button would cause a hover-vest to be put on Bugsy, and you were about to push it, when Mikhail burst in with allegedly urgent information: ‘Wait! Stop! The vest isn't a hover-vest but an ordinary one! If you press the button, you’ll be saving the five in virtue of a vest touching Bugsy's body instead of hovering around it!’ This too is at least as preposterous as treating race, gender or religious affiliation as morally relevant.

⁵⁷ Tooley, Michael, Abortion and Infanticide (New York, 1983), p. 215Google Scholar, following Thomson, ‘Killing, Letting Die, and the Trolley Problem’, p. 213.

⁵⁸ If we could destroy and replace Bugsy's organs with equally good ones without killing him, causing him any pain, inconveniencing him or going against his wishes, it seems clear to me that there would be no moral objection to doing so, just as I see no objection to taking a few hairs or skin cells from someone's body without his consent if it does not harm him or contradict his wishes. Surely any objection to doing this (even if it does contradict his wishes) would be nowhere near as weighty as the objections to seriously harming him. Even if those with strong views about ‘self-ownership’ disagree, they should agree that self-ownership gives one just as strong a claim against others’ interfering with one's vital processes as against their destroying one's vital organs.

⁵⁹ Note that my argument for the irrelevance of factors (1)–(11) considers cases where the later factors in the list are present in addition to the earlier factors. As such, I have not been assuming that the moral relevance of the factors would have to be ‘additive’ in the sense that each factor's relevance would not be affected by the presence or absence of the others. I have given the later factors on the list ample opportunity to make non-additive moral differences that are present only when the earlier factors on the list are present (for the idea – and plausible examples – of non-additive moral factors, see Kagan, ‘The Additive Fallacy’, Ethics 99 (1988), pp. 5–31CrossRefGoogle Scholar; Dancy, Jonathan, Ethics without Principles (Oxford, 2004), ch. 3CrossRefGoogle Scholar).

⁶⁰ While my case against the intrinsic relevance of factors (1)–(11) does not depend upon any empirical story of how we might be tricked into thinking that they are intrinsically relevant when they are not, I cannot refrain from observing that (a) all of these factors seem to increase the salience of the fact that you are harming Bugsy in relation to the fact that you are saving the five, and (b) most of them would in real-world cases decrease the probability of saving the five or increase the probability of killing Bugsy. It thus seems to me quite likely that the greater salience or real-world probability of killing in relation to saving is responsible for our greater tendencies to judge it wrong to kill in cases where more of the factors are present.

⁶¹ Foot, ‘Killing and Letting Die’, p. 181; and Williams, ‘A Critique of Utilitarianism’, Utilitarianism: For and Against, ed. Williams, Bernard and Smart, J. J. C. (New York, 1973), pp. 77–150, at 99Google Scholar.

⁶² For a serious attempt to show that, on closer inspection, the distinction between doing and allowing harm amounts to something that, like skin colour, seems obviously morally irrelevant (viz. a difference between there being few as opposed to many ways we could act that would result in the harm), see Bennett, ‘Whatever the Consequences’, Analysis 26 (1966), pp. 83–102CrossRefGoogle Scholar; ‘Morality and Consequences’. For a persuasive response to these arguments, see Dinello, Daniel, ‘On Killing and Letting Die’, Analysis 31 (1971), pp. 83–6CrossRefGoogle Scholar; Quinn, ‘Doing and Allowing’. Whether the DDA can in fact survive critical scrutiny is a serious question on which I think the jury is still out. Below I address what I take to be one of the strongest objections to the DDA, but a full defence of the principle is beyond the scope of this article.

⁶³ Norcross, ‘Killing and Letting Die’, A Companion to Applied Ethics, ed. Frey, R. G. and Wellman, C. (Oxford, 2003), pp. 451–63, at 455–7Google Scholar; Thomson, ‘Killing, Letting Die, and the Trolley Problem’, p. 209.

⁶⁴ Bennett, ‘Negation and Abstention: Two Theories of Allowing’, Ethics 104 (1993), pp. 75–96, at 95–6CrossRefGoogle Scholar.

⁶⁵ There are instances of allowing harm that benefit some at the expense of (i.e. in virtue of ensuring the occurrence of events that harm) others, such as refraining from treating someone's disease so that you can learn how to treat others by observing her disease's fatal progression. I think that these acts are particularly difficult to justify for the same reason that inflicting harm is difficult to justify, and the DDA is best understood as asserting that doings of harm are harder to justify than allowings of harm that do not have this feature.

⁶⁶ McMahan, ‘Intention, Permissibility, Terrorism, and War’, Philosophical Perspectives 23 (2009), pp. 345–72, at 361CrossRefGoogle Scholar. McMahan also objects to this view on the grounds that it is unstable because ‘some of the objections to the relevance of intention seem also to challenge the asymmetry between doing and allowing’. But my argument here does not rely on the general objections to the relevance of intention that McMahan discusses; it relies only on the fact that on reflection it seems manifestly absurd to hold that the differences between paradigm cases of inflicting foreseen but unintended harm (diverting in Trolley) and paradigm cases of inflicting intended harm (harvesting in Transplant) could justify a moral difference between permissibly and impermissibly killing one to save five.

⁶⁷ This is, I think, a general way of putting the insights of Frank Ramsey, John von Neumann and Oskar Morgenstern about the relationship between the preferences we (should) have for probability distributions over the various properties our acts might have and the preferences we (should) have for our acts having those properties (Ramsey, ‘Truth and Probability’ (1926), reprinted in Decision, Probability, and Utility, ed. Gardenfors, P. and Sahlin, N. (New York, 1988), pp. 19–47CrossRefGoogle Scholar; and von Neumann and Morgenstern, Theory of Games and Economic Behavior (Princeton, 1944)Google Scholar). Decision theorists are usually interested in the probabilities of our acts resulting in various outcomes, but the idea easily generalizes to the probabilities of our acts having other properties, like inflicting harm.

⁶⁸ Indeed, some have argued that given the horrors that would have resulted from a Nazi victory, even the terror bombing of German cities by the British before the tide turned against the Nazis was justified on account of its having mind-bogglingly greater expected benefits than expected harms (see Walzer, Michael, Just and Unjust Wars (New York, 1977), pp. 251–68Google Scholar; and Rawls, John, ‘50 Years after Hiroshima’, Dissent 42 (1995), pp. 323–7Google Scholar). Walzer rightly makes this conclusion contingent on the assumption, which he does not endorse, that ‘unless the bombers were used in this way, the probability that Germany would eventually be defeated would be radically reduced’. Although I am no historian, I cannot see how it would have been reasonable for the British to think that terror bombing German cities would radically increase the probability of victory. Since this is probably the best candidate for an actual-world example of terror bombing that was likely to do mind-bogglingly more good than harm, I do not believe any actual instances of terror bombing have had it. But Walzer argues persuasively that acts which harmed civilians and did radically increase the probability of winning the Second World War, which some acts that risked collateral damage did do, were likely to do mind-bogglingly more good than harm.

⁶⁹ See McNeal, ‘The U.S. Practice’, esp. p. 18 n. 67.

⁷⁰ This article has benefited from conversations with too many people to remember or name. I am deeply grateful to John Ku for many extremely helpful discussions and for feedback on multiple drafts. Special thanks are also due to Stephen Darwall, Boris Kment, Adam Morton, Alastair Norcross, Peter Railton, and an anonymous referee at Utilitas.