By In Announcements, JMP Discussions Comments (32)

JMP Discussion at PEA Soup: Kerah Gordon-Solmon’s “Self-Defence Against Multiple Threats” with a critical précis by Renee Jorgensen Bolinger

Welcome to the third Journal of Moral Philosophy discussion here at PEA Soup. It will take place on July 21-23. This is sure to be another insightful and productive discussion, this time on Kerah Gordon-Solmon “Self-Defence Against Multiple Threats.” This paper is currently available in the “Advance Articles” section online at the Journal of Moral Philosophy. They have kindly provided free access to the paper, which can be viewed or downloaded here. Renee Jorgensen Bolinger wrote a critical précis and commentary which is posted below. Please join the fun!

Renee Jorgensen Bolinger’s critical précis

Kerah Gordon-Solmon’s paper gives an impressively clear and concise statement of the problem of self-defense against multiple threats and offers a clever reply on behalf of the responsibility theorist. The central puzzle is how to restrict liability-based justifications for defense from permitting a defender to kill arbitrarily many minimally responsible threats in self-defense.

How the Problem Gets Going

To be liable to some harm H is to have forfeited your right against suffering H certain circumstances; in those circumstances, a defender who imposes H on you does not wrong you, and you lack counter-defensive rights. Kerah outlines three conditions which she treats as jointly sufficient individually necessary for liability:

  1. Instrumentality (also called a ‘success’ condition): the harm must achieve a justifying good (e.g. saving the victim’s life), or be an unavoidable side-effect of an act that does.
  2. Necessity: the harm-producing act must be necessary for bringing about this good.
  3. Narrow proportionality: the harm to the liable party must be proportionate to the good being brought about, given each party’s comparative responsibility. All else equal, if the defensive act would save a non-responsible victim’s life, then killing a responsible threat will be narrowly proportionate.

The fact that a threatener is liable to be killed gives the defender pro tanto justification for defensively killing her, but this justification can be defeated if the necessary defensive action would violate wide proportionality. A defensive action violates the wide proportionality constraint if the badness of the act (in terms of the harms it imposes on parties not liable to them) outweighs the good it achieves.

A Minimally Responsible Threat is someone who non-culpably exercises their agency in a way that foreseeably but improbably poses a threat to Victim. To illustrate, Kerah (following McMahan) offers the conscientious driver:

Conscientious Driver: A person who keeps her car well maintained and always drives cautiously and alertly decides to drive to the movies. … [F]reak circumstances cause the car to go out of control. It has veered in the direction of a pedestrian whom it will kill unless she blows it up [with an] explosive device.” (127)

Assume that other things are equal: there aren’t relevant variances in the two parties’ life expectancies, associated harms to third parties, etc. Since either Driver or Pedestrian must die, and this fact holds because of a risk that Driver took, it seems fair that Driver be the one to die. Using the explosive (and thus killing the driver) satisfies the three conditions, and so Driver is liable and Pedestrian is permitted to kill her in self-defense.

This seems to unavoidably commit the theorist to the troublesome verdict that is the focus of the paper. Suppose rather than being threatened by one driver, Pedestrian’s life is in fact threatened by many drivers, each minimally responsible but liable to be defensively killed, in precisely the same way as the conscientious driver case.

1,000 Drivers: A thousand people, each of whom keeps their car well maintained and always drives cautiously, decide to drive to run various errands. Freak circumstances cause all their cars to go out of control, veering in the direction of a single pedestrian. Each car is sufficient to kill the pedestrian, and unless stopped will kill the pedestrian at precisely the same moment. The pedestrian can save her life only by blowing up each of the 1,000 vehicles before impact; luckily, she is prepared and can do exactly that.

Intuitively, given the minimal responsibility of the threats, it is not morally permissible—because disproportionate—for the pedestrian to kill all thousand drivers to preserve her own life, regardless of whether doing so would impose harms on third parties.

The responsibility account doesn’t seem able to get this verdict. By stipulation, each killing is necessary for saving the pedestrian, and, so long as she in fact kills all the drivers, each is instrumental in saving her. As Kerah writes, “Each killing will save the pedestrian’s life. Because in each case the driver is minimally responsible and the pedestrian is non-responsible, each killing is narrowly proportionate.” (128) So, screening off complications (e.g. third parties), Pedestrian’s liability-based justification for defensively killing each of the thousand drivers will be defeated only if the badness of the harms to the drivers violates wide proportionality. But only harms to which the sufferers are not liable count against wide proportionality, and each driver is liable to be defensively killed, so their deaths don’t count against wide proportionality. Absent other harms to tally, the responsibility theorist seems committed to the conclusion that the liability justification is undefeated, and Pedestrian is morally permitted to kill all thousand.

The Solution: Side-Effect Harms

Kerah’s general strategy is to solve the puzzle by demonstrating that each killing also imposes a side-effect harm on each of the other drivers, to which they are not liable, which when taken together violate the wide proportionality constraint. In brief, “the pedestrian’s killing all the drivers also makes each of them substantially worse off as a side-effect, by bringing it about (1) that each is liable to be killed, and (2) that, consequently, she kills each.” (130)

The arguments for this conclusion proceeds roughly as follows. It’s a necessary condition on each driver’s liability that all the other drivers are killed. If they are not, then Pedestrian will still die, so the good achieved by killing a given driver is only protecting Pedestrian from this driver, which is not narrowly proportionate to the harm of killing the driver. So unless all the drivers will be killed (and thus Pedestrian saved), none of them is liable. Each killing thus has the side-effect of fulfilling one of the necessary conditions on each other driver’s liability to be defensively killed. The aggregate effects of killing the other 999 drivers therefore renders a given driver substantially worse off (because it makes her liable to be defensively killed) than she would otherwise be.

This obviously implies that in killing the drivers, Pedestrian harms each twice over: once in securing their liability, and once in killing them. So, counterintuitively, Pedestrian inflicts significantly more harm on each driver in the 1,000 drivers case than she does in the single driver case, despite performing the same action on each. Kerah embraces this implication, defending it by offering two ways in which each of the thousand drivers, but not the single driver, are made worse off by the pedestrian’s defensive killing.

  • In the 1,000 driver case, it is initially indeterminate whether each driver is liable to be defensively killed: it depends on whether the pedestrian will succeed in killing all 1,000. By contrast, the single driver starts out liable. So, in making the 1,000 drivers liable, the pedestrian worsens their position compared to where each started, but the same cannot be said of the single driver.
  • Non-comparatively, each killing deprives the 1,000 drivers of a good they previously enjoyed: it removes a shield each driver had enjoyed between her and liability, namely, the presence of the other threatening drivers. The single driver had no similar good, and so does not suffer similar deprivation.

For the imposition of these side-effect harms to render Pedestrian’s defensive killings widely disproportionate, the drivers cannot be liable to suffer these harms. As defined, to be liable to some harm requires having forfeited a right against suffering that harm. Kerah notes (plausibly enough) that it’s doubtful the drivers ever had a right against “having the liability justification on the basis of which she’s killed secured by the killing of other people” (132), and of course one cannot forfeit rights that one never had. So, at least on this definition of liability, it follows that the drivers are not liable to these side-effect harms.

Kerah contends that for minimally responsible threats, the aggregate impersonal badness of these harms outweighs the badness of the death of the pedestrian. Importantly, she holds that this impersonal badness decreases with increasing culpability, such that the harm to fully culpable threats is either not impersonally bad or below the threshold of additivity. If it holds, this argument should give the responsibility theorist the resources to secure the intuitive verdicts, allowing defenders to kill arbitrarily many culpable threats, but setting some limit to the number of minimally responsible threats that can be permissibly killed.

There’s a lot to say about the overall argument, but for brevity I’ll focus on just one concern.  It’s a little unclear whether we are supposed to understand the side-effect harms as consisting in (i) making the drivers more likely to be defensively killed, since the pedestrian is more confident that she is morally justified in defensively killing, or (ii)  directly making the drivers morally liable to such killing. The former makes sense of the epistemic language used at various points in the paper (especially p. 130), but doesn’t seem to address what was troubling about the puzzle. Whether the pedestrian is likely to kill all thousand drivers is a bit beside the point, really; the worry is that the theory says she is morally permitted to, and that seems false.

The latter would address the worry, but I’m not sure the arguments in the paper secure it. I’m concerned that (merely) epistemic uncertainty is doing a lot of the work in making the case for the existence of the side-effect harms. To screen that out, consider a variant of the thousand drivers case:

Simultaneous: Like the thousand drivers case, but Pedestrian has been armed with a button that will, either immediately or in a sequence that will definitely complete in time to save her life, vaporize all 1,000 cars and drivers.

The important difference between this and the original thousand drivers case is that in this variant, there’s no question about whether the Pedestrian will succeed in securing the justifying good (saving her own life).  It’s hard to see a case for considering it indeterminate whether each driver is liable here, and equally hard to construe the presence of the other drivers as a shield from liability. But then it doesn’t look like the argument from side-effect harms can get purchase: none of these drivers are made worse off by the other drivers’ deaths.

Intuitively, though, if liabilities are fact-relative (as they are on the Responsibility Account), the verdicts in 1,000 Drivers and Simultaneous should match. Holding fixed that she will in fact succeed in saving her life, it isn’t especially plausible that whether Pedestrian violates wide proportionality in defensively killing a thousand minimally responsible threats depends on whether she must kill them one by one or whether she has a button that will accomplish all the killings at once.

32 Responses to JMP Discussion at PEA Soup: Kerah Gordon-Solmon’s “Self-Defence Against Multiple Threats” with a critical précis by Renee Jorgensen Bolinger

  1. Thanks to Renee for her précis. Here is my take on Kerah’s paper:
    Kerah thinks that McMahan’s account of liability and the justification of self-defense has embarrassing implications in the case of multiple conscientious drivers. McMahan’s account (setting his recent amendments in the volume The Ethics of War aside, which I find ad hoc) would imply that a large number of conscientious drivers may be killed to save just one pedestrian. That, says Kerah, is counter-intuitive. Her solution is to argue that killing any one of the conscientious drivers without also killing all the others would be “narrowly disproportionate” and hence a rights-violation and impermissible (without a lesser evil justification) since the pedestrian would still get killed, and to then further argue that “by killing all the drivers, the pedestrian would also cause a non-trivial side-effect harm to each one” (129), namely “by securing the liability justification on the basis of which she kills” them (130). This side-effect harm concerns “wide proportionality” and thus not liability (but still permissibility). Since Kerah finds the claim that the single conscientious driver can be permissibly killed but not the claim that 1000 conscientious drivers can, she then offers the following explanation for this distinction: “Since the thousand drivers start off in better-off positions than the single driver (of indeter¬minate liability; each with a greater chance of survival), by making the thousand drivers and the single driver equally badly off, the pedestrian worsens the position of each of the thousand drivers more than she worsens the single driver’s.” (131)
    But why is the single driver supposed to be liable, while the pedestrian is not? Because, Kerah says, “‘conscientious walking’ does not impose on any¬one a risk comparable to the risk imposed by driving.” (128) Actually, on the basis of her very own account it does. If the conscientious driver can foresee that he might collide with a pedestrian, a pedestrian can foresee that he might collide with a conscientious driver, and at least if the pedestrian accepts Kerah’s account (and Kerah seems to require that she does, see p. 130), she can foresee that thereby she imposes the risk on drivers of securing their liability by walking into their path– after all, if she weren’t there on the street to collide with the car, the driver would not become liable in the first place. So the pedestrian imposes on the driver the risk of securing his liability by being in his path; and on Kerah’s assumptions (130) the pedestrian will then proceed to kill the driver.
    Now there are different possibilities here. If the pedestrian (as Kerah claims) does not become liable in this situation, then this means that the risk the driver imposes on her is only the risk of killing (or injuring) her, but not the risk of making her liable. Yet, the pedestrian imposes both risks on the driver: the risk of securing the driver’s liability and the risk of killing him. (Kerah expressly accepts and in fact relies on this way of harming “twice over,” see pp. 130-131). Thus, the pedestrian actually imposes a greater risk on the driver than the driver imposes on the pedestrian. But then it should be the pedestrian (contrary to what Kerah claims, but her account implies) who is liable, not the driver. In others words: on Kerah’s account, if the driver is liable and the pedestrian is not, then the driver cannot be liable but the pedestrian is – a veritable contradiction. The other possibility is that one claims that the pedestrian also becomes liable. However, Kerah’s very definition of liability is not compatible with symmetrical liability, at least if she means by a right to defend oneself also a liberty-right, not only a claim-right (126). Thus, however you tweak and twist it, Kerah’s account seems to be incoherent.
    Let us go back to Kerah’s claim, already quoted: “Since the thousand drivers start off in better-off positions than the single driver (of indeter¬minate liability; each with a greater chance of survival), by making the thousand drivers and the single driver equally badly off, the pedestrian worsens the position of each of the thousand drivers more than she worsens the single driver’s.” In reply: there is no such thing as “indeterminate liability” here. Kerah defines: “[I]f a person is liable [to be defensively killed], it means that she has forfeited her right against being killed in certain circumstances, for a certain reason. Killing her does not wrong her, and she has no right to defend herself against it.” (126) Yet on her own account, none of the drivers has forfeited her right [not to be killed without the other drivers also being killed] (I use the brackets to indicate the scope of the right). If a driver would be killed without also killing the other drivers, he would be wronged (since his death was non-instrumental for achieving a proportionate goal, like saving the life of the pedestrian). However, again on her very own account, all of the drivers have forfeited their different right [not to be killed together with the others]. Killing them together with the others is instrumental. Kerah doesn’t prohibit this killing because the killed are not liable to it (she insists they are), but because it would, allegedly, impose harms on them to which they are not liable, namely the harm of “having the liability justification on the basis of which [they are] killed secured by the killing of other people” (132). Thus, they are all liable to be killed together, while they are not liable to be killed without the others also being killed. There is nothing “indeterminate” about this. Accordingly, the “liability justification on the basis of which [they are] killed” is not “secured” by the killing of other people; rather, there simply IS a “liability justification” for killing any of them together with the others – it need not be “secured” or “created,” it’s just there –, while there is no “liability justification” for killing any of them without the others. In other words: by killing them all, the pedestrian has not made any of the drivers “worse off” by “making” any of them liable to something they have not been liable to all along, and thus Kerah’s argument from “wide proportionality” seems not to get off the ground.
    Let us now step back and have a look at the bigger picture, which raises the question: why would Kerah – or anyone, for that matter – want to defend McMahan’s responsibility account in the first place? After all, the “multiple threats” challenge is the least of its problems. To wit, from the very beginning on McMahan has admitted that his account has counter-intuitive implications. It would imply that the mother of a culpable attacker would be liable to be killed (if the threat cannot be averted otherwise) because she “chose to engage in an activity (having a child) that had a tiny probability of resulting in an unjust threat” and this threat now materialized. McMahan then rightly points out that the mother is not liable, but he claims that “it is less obvious what the right explanation is of why the mother is not liable” (“The Basis of Moral Liability to Defensive Killing,” p. 396.). Actually, the explanation is obvious: McMahan’s account is wrong. I gave a few other counter-examples to his account (“Rights, Liability, and the Moral Equality of Combatants,” pp. 334-335). Here is a different one: Bill produces and sells many hammers, foreseeing (he actually googled “killed with a hammer”) that there is a chance that a criminally insane person will use one of his hammers to kill an innocent person. On McMahan’s account the hammer manufacturer would become liable to defensive force if the threat cannot be otherwise averted. But again, that is absurd.
    It seems that some people think that the account can be tamed by appealing to “minimal responsibility” or to differences in causal contributions. Unfortunately, while the term “minimal responsible threat,” which is also used by Kerah, and the talk about “degrees of responsibility” have recently become fashionable, they do not make much sense within the McMahanian framework. To wit, the moral responsibility account of liability to defensive killing is meant as an alternative to the culpability account. Someone can be morally responsible yet entirely innocent. But unless somebody explains how and by which criteria (and one would also like to hear concrete examples) a minimally responsible morally entirely innocent threat is to be distinguished from a maximally responsible morally entirely innocent threat, “minimal responsibility” simply means responsibility and will thus be of no help to either the mother or the hammer manufacturer. Do not take my word for it, take McMahan’s word for it (he is not talking about the hammer producer although what he says applies): “Although he is responsible for the threat he poses, he is only minimally responsible, by which I mean ‘merely responsible,’ or responsible though not culpable.” (“Proportionality in War”, in The Ethics of War, p. 5, italics added.) I think it would help to avoid a lot of meanwhile rather widespread confusion to then just say “merely responsible” instead of “minimally responsible.”
    The other strategy might be to distinguish different degrees of causal contribution. Of course, it would have to be morally responsible causal contribution. But in the conscientious driver case such contribution did not only come from the driver, but also from the people at the gas station, the car manufacturers, the sales man, and the adult daughter who gave daddy the key he just couldn’t find. Yet, we would hardly deem any of these other people liable although, at least in terms of “but-for-causation” (and, incidentally, also in some other terms), the daughter, the lone guy at the gas station, and the salesman who just sold the car this morning, contributed no less to the threat than the driver himself. At this point, then, one might want to appeal to morally responsible proximate causation. Unfortunately, while thereby we get the daughter of the hook, we also get the hammer producer back on (he is the proximate morally responsible cause since the attacker is insane). In short, we have all reason to believe that the moral responsibility account is beyond saving (even if the multiple threats challenge could be met), and therefore it should simply be abandoned.
    There are a number of other things I disagree with, and which I would like to mention just for the record. Kerah believes, like McMahan, in “liability justifications.” However, liability is one thing, justification is another. Contrary to Kerah, the fact that a threat is “liable to be defensively killed” does not provide a “pro tanto (or defeasible) justification for killing her.” (125) In that context, I also do not accept the whole instrumentalist and internalist background. Nor do I deem it useful to define “liability,” as Kerah does, in such a way that questions like the following are excluded by definitional fiat: can there be symmetrical liability, is one’s liability to a harm dependent on the reason the harm is inflicted for, can one only become liable by forfeiting rights through one’s own responsible action? However, these concerns are peripheral to Kerah’s main argument, so I will not go further into them (nor do I expect Kerah to respond to them here).

  2. Victor Tadros says:

    Hello Kerah,

    thanks for the excellent paper. I have only had time to read this a bit quickly, but I had a question about a premise that appears to give rise to the problem under consideration.

    As I understand it the premise is something like this:

    If X would alone be liable to be killed to achieve some goal, g (saving Victim’s life) in virtue of some property (such as responsibly posing a threat to another person, Victim) and n people have similar properties to X (each poses a threat to Victim in succession), then all n are liable to be killed to achieve g.

    I think that this premise is clearly mistaken, and although many people writing about multiple attackers seem to accept it, I haven’t seen an argument for it.

    Think about a similar idea in other cases, that do not involve responsibility for threats, but where we lack rights. It is often the case that I have no right not to bear some cost for the sake of some goal, but yet if the only way of achieving the goal is to inflict the same cost on many people, we all have a right not to have the cost inflicted on us. For example, suppose that I can build a bridge over a river that will significantly help my community, but it will cost me £100 to build it. I may be required spend the £100, and I thus lack a right not to spend the £100. Furthermore, it might then not infringe my rights to tax me to take the £100 to build the bridge. But that does not imply that if 1 billion people would each have to spend £100 to build the bridge, they would be required to do so. If they were taxed £100 each, this would plausibly violate all their rights.

    I think that we should say the same thing about multiple attackers – the fact that one attacker is liable to be killed to save Victim’s life does not imply that 1billion similarly placed people are each liable to be killed to save Victim’s life.

    Here is a final reason to think that this is true. The solution that you present is ingenious, and I really like it. But that is also its problem. The intuition that it is wrong to kill some number of minimally culpable threats seems grounded in the simple idea that death is very important, and the number of people killed is morally significant. But if that is right, we should be suspicious of ingenious solutions to the problem – the solution should reflect our intuitions about the significance of deaths, and the significance of an aggregation of deaths, more directly.

  3. Sandy Steel says:

    Hello Kerah.

    Thanks for your very interesting paper.

    A possible objection: for every harm the victim the inflicts on each attacker by killing the others – namely, satisfying one of the necessary conditions of the other attackers’ liability – doesn’t each attacker also inflict a similar harm on the victim? Each attacker makes it harder for the victim to establish liability against any particular attacker, by adding a necessary condition on each attacker’s liability. Might the harms not then ‘cancel out’ in the wide proportionality question?

  4. Saba Bazargan-Forward says:

    Hi Kerah. This is an ingeniously argued paper. But I have a worry.

    Replace 999 of the drivers with rolling boulders. That is, suppose that 999 boulders are rolling sequentially toward the pedestrian, prior to the threat that the single driver imposes. The pedestrian, with a bazooka, blows up boulder 1, then boulder 2, and so on. Call this the “driver & 999 boulders case”. (Call the original version, in which there is just one driver and no other threats to the pedestrian, the “original case”. And call the version in which there are 1000 drivers each of whom is a minimally responsible threat, the “1000 drivers case”). In the driver & 999 boulders case, destroying all the boulders makes the driver worse off, since doing so makes her liable to be killed — on your view. But presumably you think the driver has no right that we refrain from making her worse off in this way. Since there’s no general right against this harm, the driver is not liable to it.

    But it seems strange to say that, at the end of the day, when the pedestrian has destroyed all the boulders and killed the driver, that the driver suffered a greater wrong than she would have if she were just a single driver threatening the pedestrian, absent the boulders. Put differently, it seems that on your view, killing the driver is worse when she is preceded by boulders, that it is when she isn’t preceded by boulders. This is a strange consequence.

    You anticipate this kind of criticism. She argue that that killing the driver in the original case is indeed worse than killing the driver in the 1000 drivers case. This is because, in the original case, that driver is in a worse position than any single driver in the 1000 drivers case. This is because in the 1000 drivers case, any given driver is liable to be killed only if the others are killed. But in the original case, the driver is liable to be killed ‘sans phrase’. So by killing all the threats in the 1000 drivers case, the pedestrian worsens the position of each of the thousand drivers more than she worsens the single driver’s position in the original case.

    You can, then, apply this reasoning to the driver & 999 boulders case. With each boulder we destroy, we remove from the driver a benefit – the shield from liability that the presence of each boulder provides. With each boulder we destroy, we worsen her position. So by destroying all the boulders the pedestrian worsens the position of the driver more than she worsens the single driver’s position in the original case. So it actually is worse, you might conclude, to kill the driver when she is preceded by the boulders than when she isn’t.

    Ok. But now suppose we have a choice. The driver & 999 boulders case is occurring to my left, where one pedestrian is threatened. To my right, the original case is occurring, where another pedestrian is threatened. I can only save one pedestrian – the one to my left or the one to my right. What should I do? If you are correct, I have a definitive moral reason to choose the one on my left, and not the one on my right. Why? Because on your view, killing the driver in the driver & 999 boulders case is worse than killing the driver in the original case. But this strikes me as a strange consequence.

    Similarly, suppose that to save either pedestrian, the harm we have to inflict on the drivers is non-lethal. But, suppose that the harm necessary to save the pedestrian on my left – i.e., in the driver & 999 boulders case – is slightly less than the harm necessary to save the pedestrian on my right – i.e., in the original case. (Maybe the difference is: a broken leg versus a broken leg and a broken wrist). If you are correct in your view that it is comparatively worse to harm the driver in the driver & 999 boulder case than it is to harm the driver in the original case, then there is presumably some degree of greater harm we must be willing to accept to avoid what she takes to be the worse outcome. As a result, if you are correct, then when deciding whether to inflict a broken leg on the driver to the left or a broken leg plus a broken wrist to the driver on the right, we might be required to inflict the greater harm. But again, it is difficult to accept this…

  5. Hi all,

    Thank you for joining this discussion of paper my paper! Renee, thanks especially for your really thoughtful précis.

    So, Renee, here’s my response to your “Simultaneous” case:

    I would maintain that whether each of the 1,000 drivers is liable to be killed in this case is indeterminate: it’s indeterminate because whether each is liable depends on what the pedestrian in fact does. If she evaporates the lot, then she secures her liability justification for killing each. If she doesn’t evaporate the lot, then she lacks a liability justification for killing any – since the deaths of all the others are necessary conditions for the liability of each one.

    The same verdict holds, then, both in the original, and in the Simultaneous case. In both, the drivers’ liability is indeterminate because it is contingent on the pedestrian’s not-yet-performed action(s). If she acts (and succeeds, which we’re assuming she would in both cases) the drivers are liable. Otherwise, they’re not.

    Uwe, Victor, Sandy, and Saba: I’ll be back…. (I might not reply to all of you today, but you’ll from me tomorrow, at the latest.)

  6. Hi Saba,

    Oh, I like this case.

    First: you’re right that, on my view, the driver who’s preceded by boulders is harmed (in the comparative sense of being made worse off than she would otherwise be) by the pedestrian in a way that the original, single driver is not. Her (multi-boulder) shield against liability is removed; and the liability justification on the basis of which she’s subsequently killed is secured.

    The boulder-preceded driver has no right against this harm. But her being harmed in this way is an impersonally bad thing – because her death is an impersonally bad thing. (The basis of her liability – agent-responsibility for the pedestrian’s otherwise being accidentally killed – does not make her death any less of a tragedy.) The badness of her death is already accounted for in narrow-proportionality calculations; it doesn’t get counted again in assessing wide proportionality. But the badness of securing the conditions of her liability (and so, the justification on the basis of which she’ll be killed) hasn’t yet been counted, and so does matter for wide proportionality. Multiplied by a thousand, you get a lesser evil prohibition.

    Which is, in the first instance, a sidebar to the challenge to put to me, so….

    Alright: Here’s a wrinkle in the case of the 3rd-party-defender facing two potential victims, one whose killing is over-determined a thousand times over (by 999 threatening boulders, plus a driver); one who’s killing is not. I wonder – does the unpreceded driver in this version possess a shield roughly akin to that of the boulder-preceded driver? Granted, there’s a liability justification for killing her regardless of what the defended does. But whether she’s killed on the basis of that justification is rendered up-for-grabs by what the defender does. So while the unpreceded driver doesn’t stand to be comparatively harmed, but she does stands to reap a symmetrical comparative benefit. I’d really like to say that it’s wash on that basis….

    Re: the broken-leg-plus-broken-wrist variant, I take your point: if the harms of having-the-liability-justification-on-the-basis-of-which-one-is-harmed-secured aggregate in the thousand driver case, then, disaggregated, any one such harm must be enough to outweigh a substantial lesser harm, i.e., that of a broker wrist. On the other hand: there’s a difference between these types of harm, which is that a person has a presumptive (and in this case, by stipulation, unforfeited) right against having her wrist broken. There’s no such right against having the liability justification on the basis of which one is harmed secured. It’s not obvious that the badness of latter harm defeats the wrong of contravening someone’s right against the former harm.

  7. Hi Kerah – Thanks for the interesting and well-constructed paper – and thanks to Renee for the thoughtful and clear précis. (Thanks, too, to the editors of peasoup for the invitation to join the conversation.)

    I praise the paper as “well-constructed” b.c it’s clearly written, carefully argued – and it builds well on the foundation of McMahan’s liability account, offering an inventive strategy to ameliorate one of the account’s more counterintuitive implications. Still, I must agree with Uwe’s critique above re: the paper’s foundations: “why would anyone… want to defend McMahan’s responsibility account in the first place?” – “liability is one thing, justification is another.” (I would add that pro tanto wrongdoing is one thing and all-things considered wrongdoing is another – and justifiability is one thing and justification is another – and excused wrongdoing is one thing and blameworthy wrongdoing is another…)

    A larger set of conceptual tools (what I call a “reasons-based account”) can help us think through cases in terms of pro tanto wrongdoing (which all killings of human beings are, no matter the blame of the human being who was killed), justifiable killings, justified killings (which are not all-things-considered wrong), excused killings, and blameworthy killings.

    For fear of turning this brief comment on Kerah’s excellent paper into a “Hey, here’s MY paper’s idea,” I’ll stop there. (But if anyone is interested, it’s explained in the chapter, “Against Liability,” in the recent Coons/Weber OUP volume on Ethics of Self-Defense.)

    Again, Kerah, many thanks for the paper. I enjoyed reading it.

  8. Hi Victor,

    Thank you so much for reading and for joining in!

    First, just in case, premise you state is one I’ll provisionally accept, albeit with small amendments:

    If X would alone be liable to be DEFENSIVELY killed to achieve some goal, g (saving Victim’s life) in virtue of some property (such as responsibly posing a threat to another person, Victim) and n people have similar properties to X (each poses a threat to Victim in succession), then EACH n is liable to be DEFENSIVELY killed to achieve g.

    “Defensively” because I’ve not thought enough about liability in other contexts (i.e., punishment) to have a view about the truth of the premise, there; “each” rather than “all” because “all” suggests that they’re liable as a group, which is misleading.

    Your bridge example:

    (1) A bridge might be worth the cost of $100; and it might be worth the cost of $100 being fully charged to a single individual.

    (2) But let’s say the cost of that bridge went up to $100,000. (Maybe after Brexit, the tariffs on construction materials skyrocket.) We’ve no grounds to infer from (1) that the bridge is worth the cost of $100,000, nor that it’s worth the cost of $100,000, distributed equally among 1,000 people.

    I of course completely agree. But what we’re doing, in this example, is weighing the good (the bridge) against some function of (i) its total, or aggregate cost, and (ii) its per-capita cost. (Sidebar: in this case we might think the total cost high enough, compared to the original $100 estimate, that it wouldn’t be worth $100,000 distributed over 100,000 people, at a penny a head. That is, we might judge the total cost prohibitive, regardless of the per capita cost.)

    But I don’t think the aggregative considerations that are doing the work in your example have a foothold in liability justifications (at least as the Responsibility Account understands them). Liability justifications proceed through pairwise comparisons: how does the magnitude of harm that the victim would otherwise suffer compare with the magnitude of harm that her defense would inflict on her threatener; what are the victim’s and the threatener’s respective degrees of responsibility for the fact that one of them will be harmed, etc. Whether the threatener has forfeited her right against being defensively harmed depends on the answers to these questions. (Assume that the circumstantial conditions of liability are satisfied.) The aggregate cost, as measured in lives successive pairwise comparisons deem forfeit – this, again, just seems to have no foothold. (The first part of my reply to Saba might be relevant on this point.)

    My worry applies to Jeff’s recent solution to the problem of the 1,000 minimally responsible threateners (MRTs) (in “Liability, Proportionality, and the Number of Aggressors,” in Saba Bazargan and Samuel Rickless, eds., The Ethics of War (New York: Oxford University press, 2017) – his solution aligns, I think, with what you suggest here. (Uwe – thanks for bringing this piece up, first!) I find the basic idea – that 1,000 MRTs are not liable, in the aggregate, to being killed – attractive; as you say, it speaks far more directly to our intuitions than my solution to the problem does. But I’m skeptical about its compatibility with the structure of liability on the Responsibility Account. This strikes me as the bigger bullet to bite.

    Of course, your view of liability to defensive harm has a different structure altogether (that one is liable to harms that one has an enforceable duty to bear). So ‘liability in the aggregate’ looks to be of a piece with your view.

  9. Hi Kerah,

    thanks for the reply. But why think that liability justifications work through pairwise comparisons in the way that you suggest? I know that some people (Frances Kamm; Jeff) have said this, but they haven’t defended it, and I can’t see how to do so. Once we see that one person’s interests can be insufficient to ground a duty to not to achieve some goal, but comparable interests of many people together can ground a duty to achieve that goal, and thus that there are group rights of a kind (rights that exist because a person is a member of a group of people whose interests together ground a duty) we have no reason to accept the pairwise comparison view of liability.

    Think about it this way: the moral importance of each MRT’s interests are weakened because of their responsibility for the threat, and so her interests do not ground a duty not to kill her in the single case. But that says nothing about whether these weakened interests in the aggregate ground a duty not to kill, and thus say nothing about whether killing violates their rights.

    This does not rely on my own view that liability justifications are often grounded in duties (though, of course, the fact that this view can provide a straightforward explanation of the intuition that it is wrong to kill a certain number of MRTs counts in its favour…)

  10. Saba Bazargan-Forward says:

    Hi again Kerah,

    I admire your fortitude in keeping up with this round-robin of comments! I just want to make sure I understand your response to my broken-leg-plus-broken-wrist variant. You say that in the example there is a moral difference between breaking the driver’s leg and breaking his wrist. A broken leg is a harm a right against which he has forfeited. But a broken wrist is *not* a harm a right against which he has forfeited. There is, though, a pro tanto reason to inflict the broken wrist: doing so is the lesser evil. This is because the alternative — inflicting a broken leg on the other driver — treats that driver even worse, given that he was preceded by the boulders. But this pro tanto lesser evil reason might not be substantial enough to override the driver’s right against having his wrist broken. Is this all correct?

  11. Hi Uwe,

    We’ve not met before, have we? It’s nice to meet you on the Internet. Thank you for reading my paper so carefully, and thinking so much about it.

    Below is a response to what I take to be the key internal objection to my paper (which gets unspooled in your subsequent challenges).

    You say (quoted from your comment, lightly edited):

    “[W]hy is the single driver supposed to be liable, while the pedestrian is not? Because…. “‘conscientious walking’ does not impose on anyone a risk comparable to the risk imposed by driving.” (128) Actually, on the basis of her very own account it does. If the conscientious driver can foresee that he might collide with a pedestrian, a pedestrian can foresee that he might collide with a conscientious driver, and… can foresee that thereby she imposes the risk on drivers of securing their liability by walking into their path – after all, if she weren’t there on the street to collide with the car, the driver would not become liable in the first place. So the pedestrian imposes on the driver the risk of securing his liability by being in his path; and on Kerah’s assumptions (130) the pedestrian will then proceed to kill the driver.”

    Here’s one difference: let’s grant (for the moment) your characterization, of the pedestrian as imposing a risk on the driver with her presence. The risks mutually imposed by the driver and the pedestrian are hardly symmetrical in magnitude. The driver imposes a tiny risk of her car going out of control and threatening a bystander’s life. The risk the pedestrian “imposes” is that, IF a driver’s car goes out of control, threatening the her life, and IF, should that occur, she can defend herself by, and only by, killing the driver, and IF in the circumstance, no disproportionate harms would be imposed on third parties (etc.), THEN, she has a liability justification for killing the driver, on which she’s entitled, but presumptively not morally required, to act. A threatened party’s presence is a prerequisite for a threatener’s liability (since it’s a prerequisite for a threatener being a threatener), but it’s pretty far upstream to “securing” a liability justification.

    (To bring out the contrast, think of Renee’s Simultaneous case, where the act of securing the liability justification on the basis of which each driver is killed, and the act of killing each driver, are one and the same.)

  12. Hi Sandy,

    It’s nice to meet you, and thank you for joining in!

    I’m curious — is your question inspired, in part, by Gerald Overland’s paper, “Moral Obstacles” (published in “Ethics,” 2014)? His view, if I remember, is that bystanders who are positioned such that they’ll be significantly harmed if I defend myself against an aggressor, and who therefore make impermissible otherwise justified defensive action, acquire duties to bear extra harms on my behalf. This is grounded in the fact that their presences makes my situation worse (by foreclosing what would otherwise by my best option); this is bad brute luck, the burdens of which should not fall entirely on me but be shared between us.

    (I hope I’m not mangling Overland’s view; it’s been a while since I’ve read the paper.)

    I have to admit — while I liked the paper, the whole line of thinking of bystanders as harming or disadvantaging me by *being there* leaves me cold.

    This isn’t a response, I realize, but — well, Jon Quong claims the opposite (in his “Ethics” paper of 2009): that people have presumptive property rights over whatever space they occupy at a given time; which give them special rights to defend that space. (The context…. actually, doesn’t matter for the purposes.)

    I’m bringing it up because — the views are near opposites. One says that your occupying a given space grounds special duties; the other that it grounds special rights. Why should we believe one over the other? (Or: why should we believe that the mere fact of being in a particular place has moral salience, of either valence?)

  13. Hi Michelle,

    Thanks for joining the discussion; it’s nice to meet you, and I’m so pleased you liked the paper!

    I’ve not read yours, yet, and so can’t really speak to your alternative – but I’m intrigued, and glad you drew my attention to it.

    I agree with both you and Uwe that this isn’t the place for me to offer a general defense of the Responsibility Account. But: I advance one elsewhere, namely, in the paper “What Makes a Person Liable to Defensive Harm?” (Forthcoming in “Philosophy and Phenomenological Research,” http://onlinelibrary.wiley.com/doi/10.1111/phpr.12369/abstract). The reason I “want to defend McMahan’s responsibility account in the first place” is that I think it’s largely right; the forthcoming paper makes the case.

  14. Hi Kerah,

    Thanks for your reply (and, like Saba, I admire your ability to keep up with all of these comments).

    So it sounds like you consider it indeterminate whether the drivers in Simultaneous are liable, *because* the pedestrian has not yet acted; once she does hit the button, it is no longer contingent whether she will act (and by stipulation, succeed in self-defense), and therefore no longer indeterminate whether the drivers are liable.

    That does secure symmetry between the 1,000 Drivers and Simultaneous, but I’m worried about this move for two reasons:

    1. It seems to imply that if the pedestrian decides not to defend herself, she *thereby* renders the threatener not liable, making self-defense impermissible. But liability-based defensive permissions seem like the sort of thing that can endure despite deciding not to defend oneself, with the success condition understood counterfactually (if she were to defend, she would succeed) rather than as a question about the defender’s actions in the actual world. At least it doesn’t seem like a conceptual confusion to say that a pedestrian may choose not to defend herself against a genuinely liable (but minimally responsible) threatener.

    2. More centrally, giving this answer to the Simultaneous case forces us (I think) to say that whether the single driver in the original case is liable is indeterminate until the Pedestrian acts defensively. The single driver’s liability is, after all, also contingent on the success of the Pedestrian’s not-yet performed actions.

    But if that’s so, then the thousand and the single driver are made worse off in the same type of way by the Pedestrian’s defensive action: she secures a liability that had previous been indeterminate. If we still need to say that the thousand are harmed more, then we will need to be able to make sense of how the numbers matter to the harm involved in securing liability. A naive way to do this is to count the harm as worse the more ‘shields’ one loses, such that a driver in the thousand drivers case is 10x more harmed than a driver in an otherwise parallel 100 driver case. But this strategy doesn’t look promising: it’s hard to see how a button that vaporizes each of the thousand threats at once harms each *by rendering them determinately liable* more than a button that vaporizes each of a hundred threats at once.

    Of course if we can’t make these comparisons compelling, then the claim that the thousand are each harmed comparatively more than the single driver looks tenuous.

  15. Hi Kerah,
    no, I think we did not meet before.
    Thanks for your reply. Here is my response.

    First, your redescription of the harm the pedestrian “imposes” (your scare quotes) seems to be, well, tendentious. That can be seen by the fact that I can provide the same kind of redescription for the driver. Here we go: The risk the driver “imposes” is that, IF his car goes out of control threatening a pedestrian’s life, and IF, should that occur, the pedestrian can save himself by, and only by, killing the driver, and IF in the circumstances, no proportionate benefit would be imposed on third parties so that the driver has a lesser evil justification for imposing the threat, THEN she would unjustifiably kill the pedestrian unless the pedestrian kills her. Thus, the pedestrian does not pose a lesser risk to the driver than the driver to the pedestrian. Moreover, as you explicitly describe the case, the pedestrian *will* indeed defend herself with lethal force. So he obviously does impose the risk of death on the driver. Look at this form from another point: objective justification: if the pedestrian actually knew all the relevant facts (that if she goes walking, she will end up killing a conscientious driver), is she then permitted to go walking? It doesn’t seem so. McMahan says that it would be objectively impermissible for the driver to drive on the occasion where he will actually threaten a pedestrian’s life (instead of only imposing the tiny risk). Unfortunately, this is no different for the pedestrian (do you disagree?). Second, whether “upstream” or not: the pedestrian’s walking does secure the liability under the circumstances, so my argument stands. (You did not make any distinction between “upstream securing” and “downstream securing” in your paper, so if you add this now, this seems rather ad hoc, and you would certainly have to provide some arguments for the moral relevance of this distinction.)

    As to your remark that in Renee’s Simultaneous case “the act of securing the liability justification on the basis of which each driver is killed, and the act of killing each driver, are one and the same” – I of course argued that nothing was “secured” there: the drivers were liable to be killed all together; this liability did not need to be secured, it was just there all along.

    Finally, you say that your paper in PPR makes the case the McMahan’s responsibility account is correct. I think it doesn’t, not least because you do not deal at all with the kind of counter-examples that I have mentioned above. Again you are mainly concerned with the conscientious driver. But as I pointed out, that is the least of the problems of McMahan’s account.

  16. Hi Kerah,

    Thanks for writing this interesting paper and thanks to Rachel for the helpful précis. I am, perhaps naïvely, wondering why the purported side-effect harms you identify are, in fact, harms. My worry is that these purported side-effect harms are not something that we have reason to care about and, if we accept them as harms, we can get counterintuitive verdicts in certain cases. I will raise three separate, but related questions, about this. However, I should note at the outset that I am not an expert on the literature. I apologize in advance if I have misunderstood something important.

    You nicely explain why we should think the purported side-effect harms are harms on pp. 130-131, at one point writing “the pedestrian inflicts the same harm on each of the thousand drivers and the single driver, in the sense of making each of the thousand drivers and the single driver equally badly off. But the pedestrian also harms each of the thousand drivers more than she harms the single driver, in the sense of making each of the thousand drivers worse off relative to where she started than she makes the single driver.” This is because “whether each of the thousand drivers is liable to be killed is indeterminate. (The latter is contingent on whether the pedestrian succeeds in killing all one thousand; if the pedestrian does not, none of the thousand drivers will be liable, and [on the assumption that the pedestrian will not kill any drivers who are not liable] all will survive). Since the thousand drivers start off in better-off positions than the single driver (of indeterminate liability; each with a greater chance of survival), by making the thousand drivers and the single driver equally badly off, the pedestrian worsens the position of each of the thousand drivers more than she worsens the single driver’s.”

    (1) Here is what I do not understand. Why think that decreasing one’s chances of survival is itself a harm rather than an event that will increase the chances that one will be harmed? Why is losing the shield of liability itself a harm, rather than an event that may increase the chances that a person is harmed? Provided that I have the same life with the same net pleasure, desire satisfaction, and so on, I would be indifferent between the following two worlds: one in which my chances of survival are 10% with a weak shield of liability and one where my chances of survival are 90% with a strong shield of liability. This consideration makes me worry that these are not really harms at all.

    (2) I worry about the following sort of case. Suppose that you are in a position to prevent either ten drivers from being killed by the pedestrian in a version of Conscientious Driver or prevent a car crash that would kill eleven different people. The pedestrian killing the drivers subjects them to the harm of death, as well as the side-effect harms you identify, while the people who would die in the car crash only suffer the harm of death. All else is equal.

    Now, suppose that the sum of the side-effect harms is greater than the harm of any individual’s death. If this seems implausible, just play with the numbers to make it the case (e.g. make it 1,000 or 10,000 or 1,000,000 drivers who suffer the side-effect harm).
    Here’s my worry. Given a plausible assumption about one being obligated to generally prevent the greatest harm one can, positing these side-effect harms entails that I am obligated to prevent the pedestrian from killing the ten drivers because they collectively suffer a greater harm than the eleven who die in a car crash. But that doesn’t seem right to me. I think I should clearly save the eleven innocent lives over the ten. Or, if one is a numbers skeptic, then I should at least choose randomly. Either way, it doesn’t seem wrong to prevent the car crash.

    Of course, maybe you want to deny that we should prevent the greater harm in such a case or maybe think that the harm of death is lexically prior to the side-effect harms you posit or maybe you want to bite the bullet. I would be interested to hear what you think about such a case.

    (3) Here’s one more issue that I worry about. Imagine the following two versions of Conscientious Driver.

    Conscientious Driver 1: The set-up is the same as your version in the paper. There is one pedestrian and one driver. Suppose that narrow proportionality is barely met. The driver and the pedestrian are the same age, equally good people, have the same number of loved ones and so on. But the pedestrian has one more day of good life ahead of her than the driver and this just tips the scales in favor of narrow proportionality. Suppose it’s the driver’s 40th birthday and he has 40 years and a day of good life ahead of him, while the pedestrian had her 40th birthday yesterday and has exactly 40 years of good life ahead of her.

    In Conscientious Driver 1, it would be permissible for the pedestrian to kill the driver. Now imagine this slightly different version of the case.

    Conscientious Driver 2: There are two pedestrians and one driver. It’s the driver’s 40th birthday and he has 40 years and a day of good life ahead of him. The pedestrians are older than the driver. Each is 60 and each have exactly 20 years of good life ahead of them. The collective 40 years of good life the pedestrians have ahead of them that is just as good as the 40 years of good life the pedestrian in Conscientious Driver 1 had ahead of her.

    In Conscientious Driver 2, the pedestrians not only suffer the harm of death, but also the side-effect harms you identify. Since, by stipulation, narrow proportionality is barely met in Conscientious Driver 1, and since the collective harm suffered by the pedestrians in Conscientious Driver 2 is collectively worse than the harm suffered by the pedestrian in Conscientious Driver 1 (the difference being due solely to the side-effect harms), it seems that narrow proportionality is not met in Conscientious Driver 2. Consequently, it would be unethical for the pedestrians to kill the driver. But this seems wrong to me. If it’s permissible for the pedestrians to kill the driver in the first case, then it seems permissible in the second case.

    Am I mistaken in thinking this is an implication of your view? When you describe narrow proportionality, you do so in terms of the harm of death (pp. 126-127), but I interpreted this description to be a product of the type of case you’re considering. Are you including the side-effect harms in the calculation of narrow proportionality? I took narrow proportionality, plausibly construed, to include all harms caused by the act in question. At least, I did given my reading of McMahan’s “Proportionate Defense” (2014). So, as I understand your view, killing the driver would be permissible in Conscientious Driver 1, but not in Conscientious Driver 2 and that seems wrong to me.

    Thanks again for an interesting paper. I am glad that you’re defending a version of McMahan’s view. I think it is an ingenious and viable view, even if it turns out to be wrong. Either way, it certainly seems to merit defense.

  17. Sandy Steel says:

    Hi Kerah,

    Nice to meet you as well and thanks for your reply.

    I haven’t read the Overland piece (I’m a relative outsider to this area) but I will take a look – thanks for the reference.

    It seems to me that if securing a person’s liability to defensive harm is a setback to that person’s interests, then depriving a person of securing another’s liability to defensive harm (at least where the other is posing a wrongful threat to their interests) could also plausibly be a setback to that person’s interests. In both cases, the permissibility of a person’s protecting their interests is significantly affected.

    But I agree, as you suggest, this (making liability more difficult to secure) is often a harm that it is very difficult to avoid imposing – one can do so merely by ‘being there’. However, if a person, say, intentionally situates himself in such a way as to make defensive action impermissible, it seems plausible that he could be required to bear some additional burden. Perhaps where a person causes this harm through being a minimally responsible threat (rather than just being there), it might also be justified to require them to bear some additional burdens. I’m not sure – though it does, admittedly, seem improbable: the fact one is minimally responsible for a threat of harm, H (the risk of death to the pedestrian), doesn’t make one minimally responsible for harm of an entirely different character, X (the risk of making liability more difficult to secure), just because X is also a causal upshot of one’s threat.

    (The latter risk might also (often) be a foreseeable risk of posing a threat of harm when one knows others are posing that threat. So one might (often) be minimally responsible not only for the wrongful threat of physical harm/death itself, but also the harm of making liability more difficult. Admittedly, however, this would only be the case where a person person knew or ought to have known of the existence/likelihood of other (overdetermining) threats. And it is deeply implausible that the judgment that defensive action against the 1000 MRTs is impermissible depends upon whether each MRT knew/ought to have known of the existence of the others.)

  18. Hi Kerah,

    It’s nice to e-meet you, too! I hope our paths will cross in person soon. (I was up at Queen’s last autumn giving a paper at the Colloquium in Legal & Political Philosophy. I’m sorry we didn’t get a chance to meet then.)

    I look forward to reading your paper in PPR. I promise I shall keep an open mind as to the liability account.

    There’s one thing I should clarify about my criticism of the liability account: it’s not so much that I think the account is incorrect – it’s just that I think a reasons-based account provides a richer, broader range of concepts to understand and explain the kinds of cases that perplex us when we think about self-defense. I prefer a reasons-based account simply bc I think it’s more illuminating – it helps us answer more questions about the moral positions of the various players.

    Liability accounts (it seems to me) obscure the fact that killing another human being is, always and at least, pro tanto wrong – and they fail to explain the moral residue that continues to bear rational force even in cases where a killing is fully justified.

    Now, perhaps folks who are drawn to liability accounts simply disagree with me regarding one of two things: (1) that killing a human being is always, at least, pro tanto wrong; or (2) that a moral residue remains, even in cases of fully justified killings. Both of these claims strike me as true and important – and I think a reasons-based account does a better job than the liability account in explaining them.

    That said, if someone thinks both of these claims are incorrect, then I suppose a liability account does a good enough job of explaining what they want to be able to explain about self-defense. (In that case, I would say that the liability accounts are not so much “incorrect” – they just fail to illuminate some salient features of the cases.)

    Again, thanks for the interesting discussion. (I’m impressed by your stamina over these past few days!)

  19. Hi Michelle,
    for what it is worth, I think one needs to keep a few things apart. First, you talk about a “liability account.” I think you mean by that what I elsewhere (https://philpapers.org/rec/STESOA-5) called “rights-forfeiture theory,” namely the view that liability or rights-forfeiture provides a defeasible justification for the infliction of harm. However, it does not, for reasons also pointed out by you: people do not only have rights, they also have intrinsic value, and this value does not disappear only because a right against being harmed disappears (see also my recent paper in Law and Philosophy). However, the rejection of rights-forfeiture *theory* is compatible with accepting that people can forfeit rights or become liable to the infliction of harm. Liability can play a *part* in the justification, but it is not a justification itself (not even a defeasible one). Moreover, there are different views about how people become liable. McMahan’s moral responsibility account is just one such view — it is not *the* liability account — and it is, for reasons I mentioned, a particularly implausible one. If defenders of that account want to show otherwise, they would have to address the strongest counter-examples, not the “conscientious driver” or “multiple threat.” If those other example cannot be overcome – and I am pretty sure they can’t – conscientious drivers and multiple threats do not even matter.

  20. Vera Bergelson says:

    Dear Kerah,

    Thanks for the fascinating and thought provoking paper. Even if I do not completely agree, I thoroughly enjoyed your ingenious argument. Here are a couple of points of disagreement. Basically, I am not persuaded that each minimally responsible driver (D) acquires additional liability by the fact that she is one of the group of minimally responsible Ds who also need to be killed to save the life of the pedestrian (P).

    – You argue that wide proportionality would not be violated if Ds were culpable instead of being minimally responsible. “Things would be different if the threats were fully or partially culpable. (A culpable threatener is blameworthy; her culpability is a function of the magnitude of the harm she threatens, and her degree of blameworthiness for threatening that harm.)” I wonder why the same logic should not apply more generally, to the minimally responsible Ds too. Should it apply, the minimally responsible Ds would certainly have a very low degree of blameworthiness (if at all) but the formula should still stick and, if killing culpable Ds does not violate wide proportionality, neither should killing of minimally responsible Ds.

    Indeed, why should minimally responsible actors be treated differently? Consider a group of 1000 legally insane homicidal maniacs attacking an innocent victim (V). I would think it justifiable and permissible for V to kill each of those attackers despite their lack of responsibility.

    – I doubt that each D is additionally harmed by P’s need to kill all 1000 Ds. To be so harmed she must have a right that P not kill her (or not kill her in that way). But where would that right come from? D has forfeited her right not to be killed to the extent that killing is necessary for saving the life of P. And it is necessary. The fact that that killing is not also sufficient for saving the life of P is irrelevant. If P were about to be killed by a falling rock, she would still be justified in killing in self-defense.

    – I wonder whether harm to P in your scenario can be reduced just to the loss of life. Perhaps it also involves a violation (even if non-culpable) of P’s right by each D. If so, each D’s liability does not depend on other Ds’ liability. She may be killed because posing a threat to the life of another human being combined with a strong possibility of destroying that life outweighs defensive killing of the attacker. If we view harm this way (or other more nuanced way), the actual death or saving of life of P loses its paramount significance, no additional liability is imposed on Ds, and wide proportionality is not violated.

    All in all, great paper!

  21. Hi all,

    Thank you all for sticking with this, and – especially – for your encouraging words.

    Since today’s the last day: I’ll make sure to respond to Travis’s first-round comment (Travis – thank you for joining in!), and should be able to address yesterday’s second-round comments, from Victor, Saba, and Renee. I suspect that will exhaust the timing remaining (if not me). Michelle, Sandy Uwe – please feel free to e-mail me during the week if you’re keen for a response to your latest.

    Vera – hi! You’ve joined while I’ve been typing. I’m glad you’re here; no guarantees, but I’ll do my best to respond.

  22. Teresa Bruno says:

    Hi Kerah, thanks for such a great paper! And thanks to Renee as well for her commentary and all the thoughtful contributions to the discussion. I have a question about why securing a person’s liability to defensive harm is in itself a harm to the person, something that makes her worse off.
    What happens when the pedestrian secures the drivers’ liability to defensive harm is that it is morally impermissible for the drivers to defend themselves against what the pedestrian might do against them to prevent the harm that they are bound to cause. They are rendered morally limited, there is one fewer thing that morality allows drivers to do. But it is not an actual physical limitation. Imagine that just a second before the pedestrian pushes the button that will erase the drivers from the face of earth, one of them realizes what is about to happen and pulls out her shotgun and kills the pedestrian first (this case is set in America friends). That would be morally impermissible, but doable nonetheless. This is just to point to an assumption in the paper, namely that rendering the driver’s defensive reaction morally impermissible makes her worse off. But this assumption ties morality to well-being in a way that many find objectionable. There’s Haybron’s case of Genghis Khan, a ruthless warlord who is responsible for the slaughter of millions. Haybron argues that Khan’s life might have gone very well for him (he lived a long life, was adored by many, etc.) (Daniel Haybron, “Well-being and Virtue”). Sadly, his immorality might not have harmed him. Something similar might have go on in my case above. The driver did something morally wrong, but that increased her well-being greatly, she didn’t end up dead after all. One might disagree with Haybron and want to conclude that immorality does make people worse off. I don’t mean to defend either position here, only to point out that you take a side in the paper. Given that the issue is highly contentions, more argumentation is needed to reach your conclusion. Thanks!

    (just saw your last comment Kerah, no worries about addressing my question though. Since I finished it, I thought I’d send it your way in case it’s helpful. Thanks!)

  23. Hi Victor,

    I really want to address something in the 2nd paragraph of your last; I’ll limit my reply to that.

    Here’s quoting the relevant bit:

    “[T]he moral importance of each MRT’s interests are weakened because of their responsibility for the threat, and so her interests do not ground a duty not to kill her in the single case. But that says nothing about whether these weakened interests in the aggregate ground a duty not to kill, and thus say nothing about whether killing violates their rights.”

    I disagree that the “moral importance of each MRT’s interests are weakened because of their responsibility for the threat.” An MRT’s greater responsibility for the threat than the victim’s weakens her claim not to be killed in defense of the victim. (When the only way to save the victim is to kill the MRT, and other things are equal, the MRT’s claim is forfeit.) But minimal responsibility doesn’t make the MRT’s interests any less important; it doesn’t make her death any less impersonally bad.

    I think – although this is stepping a bit outside the parameters of the present discussion – that claims about the importance of persons’ interests being weakened are best understand as a type of desert claim. (I’m of course borrowing this idea from Shelly Kagan.) When bad things happen to bad people, it strikes us as not so bad. We mind the fact of it less than we mind when bad things happen to good people. (For fun: see this piece http://www.newyorker.com/news/daily-comment/should-o-j-simpson-get-parole). But there’s no reason to discount the impersonal badness of harms suffered by MRT’s in this way. (The harms suffer by culpable threateners are another matter.)

  24. Vicente Medina says:

    I would like to thank Travis Timmerman for inviting to comment on Kerah Gordon-Solmon’s paper, “Self-Defence against Multiple Threats.” I enjoyed reading Kerah’s paper. It seems to me that she offers an interesting and rather ingenious argument about the justification of self-defense in the light of her reference to “minimally responsible threats” (henceforth MRT) or what others prefer to call “innocent threats.” One plausible explanation why some philosophers, including Kerah, have chosen to call the threat minimally responsible, rather than an innocent threat, is because the person who threatens another presumably innocent person deliberately engages in a risk like, for example, driving.
    The following are two observations that I hope might help us get a better grasp on Kerah’s argument. While my first observation is not directly related to her argument, it is to some extent relevant to her account of liability. I find her view of the forfeiture of rights, especially the right to life, unpersuasive. She states, “if a person is liable, it means that she has forfeited her right against being killed in certain circumstances, for a certain reason. Killing her does not wrong her, and she has no right to defend herself against it.” Kerah later states, “Liability, by definition, implies forfeiture.” My sense is that one might conceive of a person forfeiting her right to life if one understands rights as claims. But if one understands rights as a natural interest in preserving one’s life or an entitlement to one’s life, it is not clear to me that a person can involuntarily forfeit her right to life in any significant sense. She has a right to life not to be killed even when her behavior threatens an innocent person. The innocent person has also a right to life to defend herself against the person who is threatening her life. Both have a right to life. It is just that the right to life of the innocent person, in virtue of being innocent, outweighs the right to life of the person who is risking her life.
    As Kerah argues, if we follow her way of reasoning, we come up with an apparent counterintuitive situation. So let me rephrase Kerah’s argument by using less number of people involved in her argument.
    Let us assume that rather than having 1000 drivers and one pedestrian, I stipulate having 2 drivers and one pedestrian. By engaging in (MRT) of driving the 2 drivers threaten the pedestrian’s life. According to Kerah, other things being equal, the 2 drivers are liable and the pedestrian is not. Moreover, the only way that the pedestrian can avoid being harmed by the drivers is to kill them all. Killing only one will not be sufficient to save her life. The question is: does she have a right to kill one or all drivers?
    If I am interpreting Kerah’s argument correctly, she seems to agree with Jeff McMahan that killing one driver is justified. However, she seems to argue that killing all is not because it violates the principle of narrow proportionality, namely, the killing of the two would be disproportionate. The reasons she offers for this counterintuitive conclusion are as follows. The second driver is harmed in two ways. If the pedestrian were to kill only the first driver, the second driver could potentially survive and therefore would not be harmed. But that is impossible because according to the necessity condition the only way that the pedestrian could survive is by killing all drivers who threaten her, namely the 2 drivers. In addition, by killing the 2 drivers, the second driver is also harmed because her killing is disproportionate rather than narrowly proportional as Kerah’s account requires.
    My second observation is as follows. As with any consequentialist way of reasoning, in Kerah’s example numbers do matter. In the absence of responsibility and harm to bystanders, assuming other things being equal, the killing of more than one driver who presents a MRT to the pedestrian is questionable. Perhaps I am missing something or misunderstanding Kerah’s account of liability. Nevertheless, I think that what is important for a meaningful account of liability is not only the presence of a threat (be it minimally responsible or not), but the presence of intentionality by the person who threatens another, and whether the person who is threatened is culpable or not. That is, in the absence of intentionality and culpability the puzzle behind MRT can be explained by a purely consequentialist calculus. Thus, the concept of proportionality seems philosophically idle.
    It seems that Kerah argues that, in addition to the consequentialist calculation, one also needs to add the violation of narrow proportionality to explain why it is wrong to kill them all. But the concept of proportionality is rather fuzzy at times regardless whether we construe it as being narrow or broad. That is why I think her ceteris paribus clause assumes too much for an account of liability to be meaningful. By the expression “meaningful account of liability” I mean one that is not only philosophically interesting in a possible world scenario, which hers is, but also philosophically illuminating for real case scenarios, which hers seems not to be.

  25. Hi Travis,

    It’s nice to meet you on the Internet! You make several really good points; but since you’re giving me a pair of marvellous cases, I’ll focus on those.

    Here’s quoting the first:

    “Suppose that you are in a position to prevent either ten drivers from being killed by the pedestrian in a version of Conscientious Driver or prevent a car crash that would kill eleven different people. The pedestrian killing the drivers subjects them to the harm of death, as well as the side-effect harms you identify, while the people who would die in the car crash only suffer the harm of death. All else is equal.

    Now, suppose that the sum of the side-effect harms is greater than the harm of any individual’s death. If this seems implausible, just play with the numbers to make it the case (e.g. make it 1,000 or 10,000 or 1,000,000 drivers who suffer the side-effect harm)…. Given a plausible assumption about one being obligated to generally prevent the greatest harm one can, positing these side-effect harms entails that I am obligated to prevent the pedestrian from killing the ten drivers because they collectively suffer a greater harm than the eleven who die in a car crash. But that doesn’t seem right to me. I think I should clearly save the eleven innocent lives over the ten….”

    I’ll go ahead and up number to 1,000.

    So: On our right, we have a pedestrian who will lethally defend herself against 1,000 MRTs. On our left, we have a car crash in which 1,001 people will die. We can prevent one or the other, but not both. If we prevent the former: 1,000 liable MRTs will be saved; 1 pedestrian plus 1,001 innocent crash-victims will die. If we prevent the latter: 1 pedestrian plus 1,001 innocent crash-victims will be saved; 1,000 liable MRTs will die.

    Assuming I’ve reconstructed the case correctly – the impersonal badness of *having the liability justification on the basis on which one is killed secured* x1,000 is sufficient to make killing all of the MRTs disproportionate, in the wide sense, to saving 1 pedestrian. (Narrow proportionality concerns the rights threateners forfeit against being harmed in defense of their victims; narrow-proportionality calculations are internal to ascriptions of liability. Wide proportionality concerns harms against which no rights have been forfeit, i.e., harms to bystanders.) BUT once we factor into the wide-proportionality calculations saving an additional 1,001 people – well, that decisively tips the balance. (The MRTs deaths are already accounted for in narrow proportionality calculations and so can’t be counted again.) So my view yields the intuitively correct judgment.

    On to case number two….

    It begins:

    “Conscientious Driver 1: The set-up is the same as your version in the paper. There is one pedestrian and one driver. Suppose that narrow proportionality is barely met. The driver and the pedestrian are the same age, equally good people, have the same number of loved ones and so on. But the pedestrian has one more day of good life ahead of her than the driver and this just tips the scales in favor of narrow proportionality. Suppose it’s the driver’s 40th birthday and he has 40 years and a day of good life ahead of him, while the pedestrian had her 40th birthday yesterday and has exactly 40 years of good life ahead of her.

    In Conscientious Driver 1, it would be permissible for the pedestrian to kill the driver.”

    Here, I just have a small, pedantic disagreement. I would resist say that 4 years plus 1 day of life for the driver is a tiny bit better than 40 years of life for the pedestrian. More plausible to say that these goods are on a par.

    Of course, this matters neither to the ascription of liability, nor to the second part of the case, which is where the challenge lies:

    “Conscientious Driver 2: There are two pedestrians and one driver. It’s the driver’s 40th birthday and he has 40 years and a day of good life ahead of him. The pedestrians are older than the driver. Each is 60 and each have exactly 20 years of good life ahead of them. The collective 40 years of good life the pedestrians have ahead of them that is just as good as the 40 years of good life the pedestrian in Conscientious Driver 1 had ahead of her.”

    I’m a bit embarrassed to say – I don’t have a good answer for you on this one! The question comes down how harms to victims aggregate in narrow proportionality calculations. You’ll find good discussion of this in the Just War literature….

    Lastly: thank you so much for your role in making this discussion happen; and for participating in it.

  26. Hi Uwe,

    Thanks for your response. Here are a few quick replies –

    Clearly, there are different versions of liability accounts – I didn’t mean to imply there is only one on offer. (Although I can see why you thought I was suggesting as much. Such is the risk of quick blogging over the weekend.)

    I don’t think liability accounts are particularly helpful in illuminating an adequate range of salient considerations in cases of self-defense (or punishment) – and I’m almost equally as skeptical of rights-forfeiture accounts. (I find rights-forfeiture accounts somewhat less obscuring than liability accounts. See the first part of my chapter, Against Liability, for an explanation of why the concept of liability is particularly ill-suited to the task, given its variety and uncertainty of meanings in legal discourse – from which it was conscripted to do service in moral philosophy.)

    That said, I’m sympathetic to your view that culpable aggressors stand in a different moral position than do innocent aggressors (non-culpable threats). On my view (what I call the “exclusionary permission thesis”), culpable aggression by Aggressor (A) grounds an exclusionary permission that Defender (D) may opt to take-up – and, if D does take up this exclusionary permission, then some of the reasons D has not to kill/harm A are thereby excluded. In this way, the culpable aggression of A plays a part in the justification of D’s killing/harming A.

    This idea can be translated into liability accounts lingo by saying that A’s culpable aggression can make A liable to be killed/harmed – and thus, as you correctly observe, “liability can play a *part* in the justification” of D’s killing/harming A.

    My gripe is simply that using the concept of liability (or rights-forfeiture, for that matter) to explain the role that A’s culpable aggression plays in justifying D’s killing/harming A tends to obscure many salient considerations – eg, Did D take-up the exclusionary permission? If so, did D thereby have an undefeated reason to kill/harm A? If so, did D act for an undefeated reason in killing/harming A?, What moral residue (rational remainder) continues to bear force on D’s rational horizons after killing/harming A?, etc.

    To relate these ideas somewhat to Kerah’s paper – One benefit of the exclusionary permission thesis is that it is sensitive to the degree of A’s culpability. So, in the case of minimally culpable threats, the scope of exclusionary permission is unlikely to do much justificatory work on D’s rational horizons. (Eg, even if the 1,000 drivers were minimally culpable, rather than merely minimally responsible, the exclusionary permission thesis is unlikely to support the conclusion that D is justified in killing the 1,000 minimally culpable drivers.)

  27. Hi Saba,
    I’m not sure I completely follow your reconstruction of my response to this case. But I think I know the source of any misunderstanding: I assumed, mistakenly, that in the example, the harm with which each pedestrian was threatened was ALSO downgraded to the harm of a broken leg.

    If (counterfactually) that WERE your case: Each of the two drivers would be liable to the harm of a broken leg. Shooting the boulders would inflict an additional harm on the boulder-preceded driver, namely, the harm of securing the liability justification on the basis of which her leg would be broken. But since the latter harm is not a harm against which the driver has a right, it doesn’t factor into narrow proportionality calculations. It matters in wide proportionality; in particular, it gets weighed as an impersonally bad occurrence.

    As for the harm of a broken wrist, in addition to a broken leg: the broken wrist is a harm against which the driver has a presumptive right, which she does not forfeit. (I’m assuming, as most of us do, that the driver cannot be liable to a GREATER harm than that which she would otherwise inflict on her victim.) The harm that would otherwise befall the pedestrian has already been factored into the liability justification for inflicting on the driver the harm of the broken leg; we can’t count the former a second time to justify inflicting further harms on the driver against which she has presumptive rights. That is: there are no relevant wide proportionality considerations to take into account. The judgment is just that the broken leg + broken wrist is narrowly disproportionate. So the only pedestrian who may permissibly be defended is the one threatened by the boulder-preceded driver.

    (This is slightly different than what I said the other day, but I think it’s right.)

    Anyway: that wasn’t your case.

    So, your actual case:

    Two pedestrians face lethal threats: the first, from a driver who’s car is out of control; the second, from a different driver who’s car is also out of control, plus 999 preceding boulders.

    You can successfully defend either pedestrian, but not both. You can defend the first by inflicting on the first driver the harms of a broken leg and a broken wrist. You can defend the second by inflicting on the second driver the harm of a broken leg.

    Your challenge: whichever driver you harm, you’ll harm to a degree that’s well below the ceiling of her potential liability. Intuitively, in this case, we think that you should defensively harm the boulder-preceded driver; since you could save her victim via the infliction of the lesser defensive harm. BUT it’s not obvious that I can endorse this intuition: I need also to factor in the harm that you would be inflicting on the boulder-preceded driver by shooting the boulders (which is your means of securing the liability justification on the basis of which you would defensively harm that driver).

    This is all marvelously complicated, and I’m not sure what to make of it all. But here’s one factor that I think matters. The harm of having the liability justification on the basis of which one is killed secured is not a harm against which we have general rights. So it can’t be justified by the victim’s liability to it. (Liability implies rights forfeiture; you can’t forfeit a ‘right’ that never existed.) By contrast, the harms of broken legs and wrists – inflicted intentionally or as foreseen side effects – ARE harms against which we have general rights. Such harms ARE justifiable by the victims’ liability to them. So how do we adjudicate among those harms that are external, vs. those that are internal, to liability justifications, in this case? (I don’t have an immediate answer.)

  28. Hi Renee,

    I hope I’ve not kept you in suspense too long! Quoting selectively, first (since your comment is now pretty far upthread):

    “[I] sounds like you consider it indeterminate whether the drivers in Simultaneous are liable, *because* the pedestrian has not yet acted; once she does hit the button, it is no longer contingent whether she will act (and by stipulation, succeed in self-defense), and therefore no longer indeterminate whether the drivers are liable.

    That does secure symmetry between the 1,000 Drivers and Simultaneous, but I’m worried about this move for two reasons:

    1. It seems to imply that if the pedestrian decides not to defend herself, she *thereby* renders the threatener not liable, making self-defense impermissible. But liability-based defensive permissions seem like the sort of thing that can endure despite deciding not to defend oneself, with the success condition understood counterfactually (if she were to defend, she would succeed)…..

    2. 2. More centrally, giving this answer to the Simultaneous case forces us (I think) to say that whether the single driver in the original case is liable is indeterminate until the Pedestrian acts defensively. The single driver’s liability is, after all, also contingent on the success of the Pedestrian’s not-yet performed actions….”

    Here’s the response:

    The type of multiple-threateners case the paper addresses is special, in a way that your Simultaneous variation brings out beautifully. The action itself (pushing the button that kills all 1,000 drivers) secures the conditions of its own justification. (I.e., drivers 1-999 have to be killed for driver 1,000 to be liable to being killed, drivers 2-1,000 have to be killed for driver 1 to be liable to being killed, etc.).

    So, while I absolutely agree that normally, liability justifications for defensive harming are not contingent on whether the victim decides to defend herself (or worse, contingent on whether a 3rd-party defender decides to intervene), in these idiosyncratic multiple-threateners cases, what’s normally true doesn’t hold. Each threatener is liable to be killed just in case all (the others) are killed.

    (In case it’s not obvious, I really like your variation; if I’d thought of it, it would have been in the paper.)

  29. Hi again, all,

    Vera and Vicente –

    I fear I’ve exhausted my endurance. I’m sorry about that. But I’m hoping I’ve addressed, if only indirectly, at least some of each of your worries, in the responses I’ve written to others this afternoon.

    Sandy –

    I agree with everything you said in your last post!

    Everyone –

    Thank you, truly. You’ve made this discussion a ton of fun, and I’m delighted to be given so much material to keep working with well past this weekend. I’m gratified, too, by the collectively high level of sympathetic engagement with the paper. I hope I’ve made joining the discussion worth your while.

    All the best,
    Kerah

  30. Hi Kerah,

    It is nice to meet you too! Thanks for your helpful response to me and to everyone else. We really appreciate you doing this.

  31. Anonymous says:

    Kerah,
    Sorry to get into this discussion late. 2 points: First, I’m not a fan of the nonculpable but responsible-for-the-threat McMahan hypo. If the pedestrian is carrying a lethal weapon and is prepared to use it on a driver whose car nonculpably goes out of control, then the pedestrian is as much of a threat to the driver as the driver is to the pedestrian. (Imagine, if you will, an armed pedestrian on the infield of the Indy 500, prepared to destroy any driver whose car goes out of control.) So I reject the premise of your discussion.

    But second, even if I accept the hypo and the nonculpable driver’s responsibility, are we to say that the pedestrian is not justified in killing the driver on Monday if another driver will swerve towards the pedestrian on Tuesday, whom the pedestrian will again need to kill? And what if still another driver will nonculpably swerve towards the pedestrian on Wednesday, Thursday, etc.–for 1000 days in a row? Your analysis is counterintuitive when applied to this diachronic threat situation. But why is the synchronic situation different. (To be truly synchronic, imagine the pedestrian has a bomb that will kill all 1000 simultaneously.) My sense is that you are imagining a diachronic situation, with the time intervals between threats being small–but then what is the time interval between threats that makes the difference between your 1000 threat hypo and my 1000 threats, 1 per day hypo?

    Best,
    Larry

  32. Larry Alexander says:

    That anonymous comment was not meant to be anonymous. I hit the button too early.

    Larry Alexander

    larrya@sandiego.edu