II. Wrongs to Unforeseeable Victims

I suggested that the Relational Fault Requirement, which takes its inspiration from Palsgraf, implies that Ali does not wrong Cat in Quarry 3. However, in a recent paper, Jed Lewinsohn offers an analysis of Cardozo’s judgment in Palsgraf that, if extended to cases like Quarry 3, has a different implication. Lewinsohn interprets Cardozo as articulating a principle he calls Danger Zone, which holds that, if an agent believes or should believe that some act imposes an unreasonable risk of harming someone or other who instantiates a “danger-making property,” then the agent owes a duty of care to all and only those who even unforeseeably instantiate that property.Footnote ⁵ A danger-making property is roughly one that carries a relevant risk when that property is instantiated, such as being in the physical proximity of an explosion or drinking from a contaminated water source. A person who instantiates this property is owed a duty of care and may be wronged if this duty is breached, even if it was unforeseeable that this person would instantiate the property.

Importantly, this principle generates duties via these danger-making properties rather than the ex ante probability of harm. The ex ante probability of Cat being hurt by the explosion is extremely low since they are paragliding miles away, but once Cat is in the quarry, they are owed a duty because they instantiate a danger-making property. Similarly, Mrs. Palsgraf would have been able to claim against the defendant if, say, the box had fallen on her feet, regardless of whether the ex ante probability of her being in the vicinity of the box was very low. If this is right, and if we extend this idea to the realm of moral duties, the Palsgraf decision provides support for the intuition that Cat is wronged in Quarry 3.

Lewinsohn’s primary interest is in defining which victims of a wrong are “properly aggrieved,” meaning that they possess a set of entitlements in relation to the wrong such as the right to demand an apology, compensation, and so forth. Directed wrongdoing and proper aggrievement are separate concepts, but since I think they typically go together, for my purposes I will treat Danger Zone as applying to both. Lewinsohn appeals to his analysis to argue against a range of theories that seek to ground proper aggrievement in some relation between the wrongdoer’s act and that individual’s interests. A general theme of the argument is that none of these interest-based theories can accommodate Danger Zone (and thus Palsgraf) because an appeal to the ex ante probability of harming any particular person cannot predict who will be properly aggrieved. Two people with equal ex ante probabilities of harm might find themselves on different sides of the danger zone boundary. This would happen if, say, Bea and another hiker, Delta, who are equally at risk, choose different paths, only one of which leads into the quarry. His conclusion is that the grounds of proper aggrievement are mediated through impersonal danger-making properties and cannot be accounted for using the “highly individualistic notions that have dominated philosophical discussions of relational normativity.”Footnote ⁶ The best explanation for how we demarcate the class of the properly aggrieved is a conventionalist one, which depends ultimately on a set of social and institutional practices.

I agree with Danger Zone’s implication that Cat is wronged in Quarry 3 (and is properly aggrieved). While it is unforeseeable that Cat would be in the vicinity of the explosion, Ali’s wrongful failure to take care means that he wrongs anyone who is in fact in the vicinity. But this judgment about Quarry 3 violates the Relational Fault Requirement. Had Ali not been at fault in relation to anyone, the case would be analogous to Quarry 2 and Cat would not have been wronged. It is only because Ali’s failure to erect the fence is wrong in relation to hikers like Bea that we have the intuition that Cat is also wronged.

We could try and get around this point by saying the following. Ali is at fault in relation to Cat because he did something inherently dangerous (setting an explosive) which in fact harmed Cat. It does not matter that it was not foreseeable that Cat would be in the quarry; all that matters is that it was foreseeable that Ali would harm Cat if they turned up in the quarry because of the dangerousness of his conduct. This is sufficient to conclude that Ali is at fault in relation to Cat. The problem with this suggestion is that it defines relational fault too broadly (if we were talking about the duty of care in negligence, we could say it defines this too broadly for similar reasons). It is true in both Quarry 2 and Quarry 3 that Ali would harm Cat if they turned up in the quarry, yet in the former case Ali clearly commits no wrong. In fact, the highest degree of care would not guarantee that Ali would not harm Cat if they turned up in the quarry, given the potentially harmful nature of explosives. We cannot appeal directly to the inherent dangerousness of Ali’s conduct to explain why he wrongs Cat, because inherently dangerous conduct is only wrongful if it imposes risks on others. Detonating a bomb on Mars is not faulty in relation to anyone because no one is subject to an unjustified risk. On the other hand, if by inherently dangerous we mean to imply that the activity poses an unjustified risk to someone, we are back to the claim that Ali’s wrong against Cat in Quarry 3 piggybacks on his fault in relation to foreseeable victims like Bea. Ali’s act of setting an explosive and failing to erect a fence imposes an unjustified risk on hikers, not paragliders. The conclusion that Ali nevertheless wrongs Cat must, therefore, violate the Relational Fault Requirement.

The second claim I want to make in relation to Quarry 3, and contra Lewinsohn, is that the wrong Ali commits against Cat is a nonconventionalist feature of relational normativity. I will explore two ways of understanding this wrong and conclude that it is the Relational Fault Requirement that needs to be revised rather than the notion that the wrong is nonconventional.

III. Harming Through a Wrongful Act

To achieve this aim, we might endorse something like the following:

Here I understand fault to encompass negligent, reckless, and intentional wrongdoing. In Quarry 3, Ali’s failure to erect the fence is negligent, but if it had been reckless or intentional this would also satisfy Nonrelational Fault. Often these faulty acts will be wrongful in themselves, especially when they impose unjustified risks on others as Ali does in Quarry 3 (this assumes that the imposition of pure risks can be wrongful, which is a view I have defended elsewhere),Footnote ⁷ or when they constitute inchoate wrongs. I leave open the possibility that an act can be at fault in one of these ways without amounting to wrongdoing. However, I will focus here on acts that themselves constitute moral wrongs, whether to others or no one in particular.

Another important clarification is that, since Nonrelational Fault is a view about moral rather than legal wrongs, the wrongdoer must be negligent (or reckless or intentional) only in the moral sense, which may differ from how the law defines these concepts. For example, the objective standard of care enshrined in the law of negligence means that some people who are not morally at fault, because they lack the capacity to live up to this standard, are nevertheless legally liable, and I do not intend to include these people in the scope of Nonrelational Fault. Footnote ⁸

Nonrelational Fault sits alongside relational fault as a distinct route to directed wrongdoing. A directed wrong of the sort specified by Nonrelational Fault can piggyback on two kinds of nonrelational faulty acts: those that are at fault with respect to a third party and those that are at fault tout court. I will focus on acts like those in Quarry 3 and Palsgraf that are at fault with respect to a third party and will mention cases where acts are at fault to no one in particular where relevant (note that I consider Quarry 3 and negligence cases in general instances of relational fault even though we often do not know the identity of the foreseeable victims, or precisely how many there are. As long as there are some foreseeable victims and the risk imposition is unjustified, the act is faulty in relation to them). Like the Relational Fault Requirement, I do not intend Nonrelational Fault as a universally sufficient condition for directed wrongdoing. It is a sufficient condition only for those wrongs that have some fault requirement, especially negligent, reckless, or intentional harming.

With these clarifications out of the way, Nonrelational Fault implies that Ali wrongs Cat in Quarry 3. More generally, it implies that a person is always wronged if they are harmed by an act that is at fault in relation to a third party. One objection to Nonrelational Fault is that, in Quarry 2 and Quarry 3, the relation between Ali and Cat is exactly the same. In both cases, it is not foreseeable that Cat will be in the quarry because they are paragliding miles away. How, then, can it be true that Ali wrongs Cat in Quarry 3 but not in Quarry 2?

It is true that Ali harms Cat to the same degree in Quarry 2 and Quarry 3, and Cat’s ex ante probability of harm is also the same. So if we think directed wrongdoing is determined exclusively by the actual or expected harm imposed on the victim, this objection is powerful. Some views seem vulnerable to this objection, as Lewinsohn shows.Footnote ⁹ For example, in some contractualist views, a person’s objection to a given principle of action is determined only by its impact on their personal interests, where these are understood in terms of actual or expected harm. On such views, if Ali does not wrong Cat in Quarry 2, we must reach the same conclusion in Quarry 3. Footnote ¹⁰

There are two responses to this problem that offer alternatives to Lewinsohn’s suggestion that the explanation is a matter of convention. One is to reject the narrow conception of personal interests, and another is to argue that facts independent of interference with personal interests sometimes ground directed wrongdoing. I prefer the second solution, but both are more plausible than adhering to the narrow conception of personal interests. Certainly, we cannot infer that Cat’s interests are a setback to the same degree in Quarry 2 and Quarry 3 from the fact that Ali’s relation to Cat is the same in the two cases because this is not true. In Quarry 3, Ali harms Cat through a negligent act, and in Quarry 2 he does not. It is true that, in Quarry 3, Ali’s act is negligent because of the risk it imposes on others, but this does not change the fact that this relation—that of harming through a negligent act—holds specifically between Ali and Cat.

Even the “individualist restriction” adopted by contractualists and criticized by othersFootnote ¹¹ can be modified to accommodate this idea. If we expand the notion of personal interests to include whether one’s interests are impacted by an act that is independently negligent, reckless, or showing culpable intent, then Cat’s personal interests are more severely affected in Quarry 3 than in Quarry 2. This is not such an unusual or arbitrary extension of the notion of personal interests. My primary concern with others’ conduct is that they do not harm me, but I also care that they do not harm me through a moral failing on their part, even if that failing is not directed at me. This is especially evident when the failing is highly culpable. All else equal,Footnote ¹² I would rather have my arm broken by an honest accident than a botched attempt to kill the person standing next to me. In the latter case, my injury results from behavior that I, and all others, have strong reasons to condemn. It is something that morally should not have happened, and that should have been prevented by the person who caused it.

Of course, we need not spell this thought out in terms of personal interests. We could go with the second option and hold that being harmed by a faulty act plays an interest-independent role in grounding directed wrongs. This option will appeal more to those who, like me, adopt a pluralistic attitude toward the grounds of directed wrongdoing, rather than trying to reduce them to a single general idea. Having raised these options, I will remain ecumenical instead of trying to adjudicate them any further. Our next task is to ask, aside from explaining the intuitive difference between Quarry 2 and Quarry 3, why we should think Nonrelational Fault is plausible.

IV. The Appropriateness of Guilt and the Inappropriateness of Mere Agent-Regret

Consider agent-regret. In Bernard Williams’ classic formulation of agent-regret, its “constitutive thought” is that I am the agent of harm, rather than that I have done something I morally should not have done.Footnote ¹³ In Quarry 3, I suggest it would be inappropriate for Ali to experience mere agent-regret, as this would treat his harming Cat as equivalent to his causing an innocent accident, like in Quarry 2. He is a culpable agent of harm, not an innocent one, even though the culpable wrong that brought about that harm is directed at a third party. The flip side of this claim is that it would not be inappropriate for Ali to feel full-blooded guilt toward Cat, since he did something morally wrong which hurt Cat. Similarly, Cat is entitled to be angry about the fact that they would not be in this situation if Ali had adhered to his moral obligations. This forms the basis of a distinctive complaint that is not available to the victim of an innocent accident. In Quarry 2, for instance, where Ali does nothing wrong and yet harms Cat, it would be unfair to expect Ali to feel full-blooded guilt rather than agent-regret.

Another way of seeing why mere agent-regret is inappropriate is that it would be unwarranted for third parties to talk Ali out of his negative feelings. While we expect those who cause innocent harm to feel agent-regret, we also think it is appropriate for others to try and lessen their pain by reassuring them that they were not at fault.Footnote ¹⁴ This is not true of guilt, where we would look askance at anyone who tried to talk a culpable wrongdoer out of feeling guilty. This difference is an important way to distinguish between agent-regret and guilt. In Quarry 2, it would be appropriate for a third party to try and diminish Ali’s sense of agent-regret with sentiments such as, “It wasn’t your fault,” “You did nothing wrong,” and “No one else would have done anything differently.” In Quarry 3, where Ali wrongly disregards the safety of hikers like Bea, similar sentiments would be inappropriate. This is even more evident if we increase Ali’s culpability. If he acts recklessly or intentionally in not erecting the fence, it seems even more inappropriate to talk him out of his guilt.

At this point, one might press the following objection. Granted, Ali ought to feel guilt rather than mere agent-regret in Quarry 3, but this is because he is at fault in relation to hikers like Bea. Either he should feel a directed form of guilt, but to the hikers rather than to Cat, or he should feel guilty in a nondirectional sense. In either case, the appropriateness of guilt and the inappropriateness of mere agent-regret can be explained without the assumption that Ali wrongs Cat.

The problem with this reply is that it mischaracterizes the phenomenology of the case. In Quarry 3, it would be odd if Cat were to confront Ali, nursing the life-changing injuries that he inflicted through his careless conduct, only to be told that he did not feel the least bit guilty toward Cat. Similarly, if Ali did feel guilty toward Cat, it seems inappropriate for third parties to contest this, saying, “Your action wasn’t wrong in relation to them, so direct your guilt elsewhere!” To this, I think Ali could, and should, respond, “I just ruined this person’s life because of my negligence, and you think I have no reason to feel guilty about that? Maybe we shouldn’t hang out so much…”

To make the same point another way, consider the following cases:

In Two Responses 1, I take it that Ali should apologize to Bea rather than Delta. He has more reason to apologize for wrongfully harming Bea than for innocently harming Delta (if you do not think we ought to apologize for innocent harm at all, this gives you all the more reason to distinguish the responses owed to the two victims). I think Two Responses 2 is analogous. Ali should apologize to Cat rather than Delta. To apologize to Delta instead, or flip a coin, would ignore the culpability that was implicated in the harm he did to Cat. This suggests that what he does to Cat is closer to a standard case of directed wrongdoing than it is to a case of innocent wrongdoing. His response goes beyond mere agent-regret and is therefore best understood as a form of guilt.

Appealing to nondirectional guilt to explain the phenomenology does not fare better. If Ali has reason to feel guilty only about the faultiness of his conduct, rather than because he wrongs anyone in particular, this seems like a self-indulgent form of guilt. There are contexts when preoccupation with one’s own moral character is appropriate, such as when a person is engaging in moral self-examination. When it comes to guilt about a particular wrongful act, though, nondirectional guilt misses something important. If anything, it would be worse if Ali did not feel guilty in relation to Cat or Bea than if he were to feel it only in relation to Bea. Ali should feel guilty not just because he acted wrongly tout court, but because this wrong caused harm to another person. Ali’s fault may not be directed at Cat, but once it causes harm to them, his guilt should be so directed.

V. The Broadness Problem

Although I think Nonrelational Fault helps to explain how Ali wrongs Cat in Quarry 3, it faces challenges. The central problem is that it is very broad. Rather than giving us a way of distinguishing between Quarry 3 and Palsgraf, it implies that the victims in both cases were wronged because they were harmed by an action that was at fault in relation to others. Yet it is more intuitive that Cat is wronged in Quarry 3 than that the plaintiff was wronged in Palsgraf.

One response to the broadness problem is to appeal to the distinction between causation requirements and other elements of a wrong. One reason why cases like Palsgraf can be confusing is because, even if we agree with the result, it is not clear whether it should be based on Cardozo’s rationale that the plaintiff was owed no duty, or on the idea that the relevant causal conditions were not satisfied. Consider the following case, which brings out these issues.

Nonrelational Fault implies that Ali wrongs Cat in Quarry 4 because Ali’s act is at fault in relation to Bea, and Cat ends up being harmed as a result. But it is less intuitive that Cat is wronged in Quarry 4 then it is in Quarry 3. In Quarry 3 Cat is injured directly by the explosion, whereas in Quarry 4 Cat is injured by a landslide produced by a series of unusual events that are triggered by the explosion, not unlike the causal chain in Palsgraf. The intuitive difference between the cases remains even if we stipulate that Cat’s ex ante probability of harm is the same in both. Perhaps the explanation as to why we should distinguish these cases is that, in Quarry 4, the relevant causal requirements are not satisfied because the causal route is too attenuated. Exactly what the relevant causal requirements are is of course a complex question that we cannot hope to settle here, but it should encourage us to revise Nonrelational Fault into the following:

If highly unusual causal routes are ruled out, Nonrelational Fault 2 can distinguish between Quarry 3 and Quarry 4. Reflection on these cases shows that we should tread carefully around Palsgraf. It is easy to confuse causation with noncausal elements of wrongdoing such as foreseeability, because unforeseeable outcomes are often those that arise via unusual causal routes. Nevertheless, causation and foreseeability are distinct elements of wrongdoing that should not be conflated. The important point for our purposes is that some of the concerns about broadness might be addressed once we factor in causal requirements. Nonrelational Fault 2 can therefore accommodate the decision in Palsgraf if it can be explained in terms of lack of appropriate causation, although admittedly it is difficult to know whether this is so until we have a fleshed-out account of causation and a definitive analysis of Palsgraf, neither of which is likely to be forthcoming.

In any case, relying on causal principles will not solve the broadness problem completely. In fact, this problem is deeper than we have so far acknowledged. Once we accept that one can be wronged in the directed sense by an act that is at fault in relation to another, or to no one in particular, it is not obvious why we should not apply this view in cases where the reasons why the act is wrong have nothing to do with the outcome that occurs. Consider:

In Fraud and Priceless Artifact, Cat is an unforeseeable victim who is harmed by a wrongful act. In both cases, though, the grounds on which the act is wrong not only have nothing to do with the act’s relation to Cat, but they also have nothing to do with the type of harm that Cat suffers. In Fraud, the wrong is grounded in the intentional violation of property rights rather than the infliction of psychological harm, and in Priceless Artifact Ali’s act is not a directed wrong at all but a violation of impersonal values.

There are a few options here. The most expansive, and perhaps most radical, is to bite the bullet and insist that directed wrongs occur in all these cases. Though some may dismiss this out of hand, I think it is more arguable than it first appears. It is supported by the fact that, in all of them, the essence of the complaint articulated above still applies. In Fraud and Priceless Artifact, it is understandable that Cat would be angry at Ali, even though what makes the act faulty had nothing to do with the prospect of harm to Cat. It is still true that Ali was not morally free to harm Cat and did so anyway. It would be appropriate for Ali to experience guilt in relation to Cat rather than mere agent-regret. If he were to view himself as equivalent to someone who innocently caused harm, I suspect we would find something lacking in his response, a failure to take responsibility for the outcome of his actions.Footnote ¹⁵ The insufficiency of agent-regret does not depend on causing the kind of outcome that made the act wrongful; it depends on the fact that the agent is culpable for the wrong that leads to harm.

Bear in mind that I am not necessarily arguing that Palsgraf was wrongly decided, for two reasons. First, as we previously clarified, we are interested in directed wrongdoing in the moral sense rather than legal liability. There are many reasons why liability to compensate should be more restrictive in scope than directed moral wrongdoing; for example, to prevent it from becoming over-burdensome, disproportionate, or difficult to predict (we will return to these issues later).

Second, the decision in Palsgraf was correct as a decision about the wrong of negligence. The defendant in this case did not commit this wrong against the plaintiff. Instead, I am suggesting that they committed the wrong of harming through wrongful action, which has a distinct scope and grounds. Recognizing this wrong does not require us to reject the ruling in Palsgraf as a decision about negligence. The court might have been right not to enforce this other wrong, or to muddle it with the wrong of negligence. None of this means that Nonrelational Fault 2 is false.

VI. Sufficiently Similar Grounds

Biting the bullet will not appeal to everyone who accepts that Ali wrongs Cat in Quarry 3, so let us explore a narrower way of understanding the wrong in this case. We might say that Ali wrongs Cat not just because Cat is harmed by Ali’s faulty act, but also because what happens to Cat is the same kind of thing that could have happened to hikers like Bea, which made Ali’s act faulty in relation to them. The impact on Cat is sufficiently similar to the grounds on which Ali’s act is faulty in relation to hikers that we are willing to accept that Ali’s fault toward the hikers forms part of a wrong to Cat. By contrast, we have no complaint about being unforeseeably harmed by faulty acts if the reasons why those acts are faulty have nothing to do with their tendency to cause the outcome that in fact occurs. This is why we are uncomfortable with treating fraud or cultural vandalism as an element of the wrong of inflicting physical or psychological harm, while we are more willing to treat risking harm (or doing harm) to third parties as an element of such a wrong, as in Quarry 3. This amendment gets plausible results across the range of cases we have considered. It implies that Cat is not wronged in Fraud and Priceless Artifact but is wronged in Quarry 3.

We might then make the following variation to Nonrelational Fault 2:

Nonrelational Fault 3 also distinguishes between Quarry 3 and Quarry 4. In Quarry 3, Ali’s act is faulty in relation to hikers like Bea, but nevertheless Cat is wronged even though it is unforeseeable that Cat would end up being harmed, because the harm to Cat (physical injury by explosives) is sufficiently similar to what grounds the faultiness of Ali’s act (the threat of physical injury by explosives to hikers like Bea). In Quarry 4, Ali’s act is also faulty in relation to hikers like Bea. However, Cat is not wronged because the harm Cat suffers (being hit by a landslide) is not sufficiently similar to what grounds the faultiness of Ali’s act. In both cases Ali’s act is wrong in virtue of the risk of explosive damage and not landslides, so he wrongs Cat in Quarry 3 and not Quarry 4. Similarly, in Palsgraf, the defendant’s act was wrong in virtue of the risks associated with pushing a passenger onto a moving train, not the risks associated with accidentally exploding fireworks.

One of the main challenges faced by Nonrelational Fault 3 is to explain what counts as “sufficiently similar grounds” in a nonarbitrary way. This could mean that the type of harm suffered is the same although the specific way it comes about is different. This explains cases like Fraud and Priceless Artifact. In these cases, the outcome that occurs (physical or psychological injury) is of a different type to the potential harms that ground the initial wrong (deprivation of property and cultural vandalism). This may not be workable, though, because in Quarry 3 and Quarry 4, the harm is of a similar type: injury to the body. To distinguish these cases, we would have to fall back on the idea that the relevant causal requirements are not met in Quarry 4.

Another possibility is that there are “sufficiently similar grounds” in Quarry 3 because the way the harm occurs to Cat is as foreseeable as the way it threatens to occur to Bea. This casts the grounds on which an act is wrong more broadly, roughly as foreseeable setback to a significant interest. In Quarry 3 Cat is harmed in the same way that Bea threatens to be harmed: physical injury through explosion. In both cases the route via which harm arises is foreseeable. Thus, Cat is wronged even though Ali’s act is faulty because of the risk to Bea. This interpretation has the advantage of allowing some leeway in the precise causal route to the harm. Suppose that when Cat gets into the quarry, Cat does not end up close enough to be harmed by the explosion directly, but the explosion dislodges a rock that falls on Cat. If we think Cat is wronged by Ali in the initial case of Quarry 3, we should reach the same conclusion here because although the specific route to Cat’s injury is different, it is just as foreseeable that they would be injured by debris as by the explosion itself. These variations of Quarry 3 differ from Quarry 4 where the route to the harm is unforeseeable since it involves a series of freak occurrences.

It is not clear that this is workable either, though, because the requirement that the specific route to harm be foreseeable is too restrictive. In the case of Hughes v Lord Advocate,Footnote ¹⁶ for example, a defendant was found liable for burns suffered by boys who had meddled with some equipment left out by Post Office workers. One of the boys tripped over a lamp which fell into a hole and exploded, causing him to fall in and burn himself. The House of Lords found in favor of the boys, Lord Morris holding that burn damage was foreseeable and the defendant should not escape liability just because they could not foresee the precise way in which it came about. This seems a sensible conclusion. If a person creates an obvious danger, they should not escape responsibility because the specific route to the harm was unforeseeable, as long as the outcome was foreseeable in a more general sense. A plausible gloss on the “sufficiently similar grounds” condition should therefore allow that the specific route need not be foreseeable while also holding onto the idea that some unforeseeable harm, such as that in Quarry 4, is ruled out.

The precise delineation of Nonrelational Fault 3 will be complicated and no doubt there will be edge cases. However, these issues should not cause us to lose sight of the plausibility of Nonrelational Fault 3. Not only does it give us the right results across the range of cases we have considered, enabling us to distinguish between Quarry 3 and Palsgraf, but it also runs parallel to principles frequently invoked in causation cases. For example, consider:

Intuitively, Ali murders Bea despite the fact that the killing happens via an unforeseeable causal route. This is because Ali achieves his purpose in killing Bea in roughly the manner he intends, so the causal route that occurs is sufficiently similar to the predicted one. Compare this with:

We can stipulate that the ex ante probability of the gun having the defect in Unusual Murder and Bea being struck by lightning in Lightning is the same. It is not the probability that distinguishes these cases but the fact that the causal route that occurs in the first case is sufficiently similar to the predicted one, and in the second it is not. Similarly, in Quarry 3, Ali’s act is negligent because it poses a risk of explosive damage. Although it ends up happening to an unforeseeable victim (Cat), the outcome is sufficiently similar to the risks that made the act wrong in the first place for us to conclude that Ali wrongs Cat. In this way, the “sufficiently similar grounds” condition can be seen as mutually supportive of the cause/coincidence distinction.

To sum up, then, we have seen three main arguments in relation to Quarry 3. First, it is appropriate for Ali to feel guilt toward Cat and inappropriate to feel mere agent-regret. Similarly, he has more reason to apologize for what he does to Cat than for an innocently caused equivalent harm. This argument supports Nonrelational Fault 2, the broad view that we always wrong those we harm via our wrongful actions, as long as relevant causal conditions are satisfied. The first argument also supports Relational Fault 3, the narrower view that Relational Fault 2 must be satisfied in addition to the requirement that the outcome that occurs is sufficiently similar to the grounds on which the act is at fault. Two further arguments also support Relational Fault 3. One is that it gives us intuitive verdicts across the range of cases we have considered. It allows us to explain our initial verdict in Quarry 3 without implying that directed wrongs occur in Quarry 4, Fraud, or Cultural Vandalism. Finally, the “sufficiently similar grounds” condition, though difficult to gloss precisely, is consonant with distinctions we commonly draw elsewhere in our thinking about wrongdoing, such as that between causes and coincidences. Of course, even those who feel the force of some or all these arguments might think the objections are still too weighty to accept them. In the next section I attempt to alleviate the weight of some of these objections.

VII. Burdensomeness and Ex Ante Guidance

So far, I have been stating the argument in terms of directed wrongdoing. In Quarry 3, Ali wrongs Cat. This seems to me a natural way of stating the normative relation between Ali and Cat. There are other things we could say about this relation, for example (1) that Ali breaches a directed duty to Cat, (2) that Ali violates a right Cat holds against Ali, and (3) that Cat can demand some response from Ali such as apology or compensation. Some, such as Nicholas Cornell, argue that wrongs and rights violations come apart, so that it is possible to wrong someone without violating their rights. Accepting Cornell’s view would be convenient for my purposes because in Quarry 3 it is prima facie more plausible to say Cat is wronged than that Cat’s rights have been violated. Accepting the latter claim might be a step too far for some.

Although Cornell’s view is helpful for my purposes, I do not think it is right, at least in one respect. While I have no preference about how best to use the language of rights, Cornell identifies one normative feature of rights that, in his view, does not apply to wrongs: rights constrain ex ante conduct whereas the prospect of wronging someone needs not. To motivate this, Cornell presents several cases. One involves two children who are threatened with harm because A negligently fails to warn them about a fireworks display, and A can now warn only one of the children about the impending explosion. One child has a family who will be grief-stricken over the child’s death; the other child’s death will not cause grief to anyone. Cornell claims that we should not take these third-party harms into account when deciding whom to warn—they have no purchase on our ex ante reasoning. This shows that, though the parents of the child may be wronged, they lack rights against being wronged.

There are some issues with this example. The case should hold all else equal between the two children (there is no additional reason to save one or the other) and focus on the fact that only one death will result in grief. But by virtue of this fact, all else is not equal, since one child is living life with a family and the other is not. We would expect this to make the child without a family worse off, and it is plausible that preexisting well-being is relevant to decisions about whom to save. Another confounding feature is that both children will be wronged as the threat was created by negligence. Although one wrong may cause more harm, it does not follow that one child is wronged more than another. As a general matter, the severity of wronging does not track the severity of harm. We see this when we reflect that, though the murder of a child and the murder of a hundred-year-old cause different amounts of harm (because the child is deprived of more future life), they are equally wrong.Footnote ¹⁷ So it may be that our intuitions about Cornell’s case are affected by the fact that the children will be equally wronged rather than the fact that third-party harms are irrelevant in calculations about whom to save.

To eliminate these concerns, consider the following:

This case treats Bea and Cat as equal in all respects but the one we wish to focus on, namely the presence of third-party harms. Although the threat is no one’s fault, Bea and Cat have a right to be rescued if this can be done at little or no cost. Some may be skeptical of a right to rescue, but here I mean only that there is a constraint on Ali’s ex ante conduct. The fact that someone will die unless he performs a costless rescue makes him morally unfree to do anything else. If that is the mark of rights, Bea and Cat have a right to be rescued.

In Third-Party Harm, though Bea and Cat have equal rights to be saved, Ali ought to save Cat rather than Bea as this will prevent further serious harm to five other people. The debilitating PTSD of these five gives Ali some reason to save Cat rather than Bea, and since all else is equal, this tips the balance in favor of saving Cat. To strengthen this conclusion, simply increase the amount of third-party harm. Suppose that ten people will suffer PTSD, then thirty, a hundred, a thousand, and so on. At some point, it will become difficult to insist that the prospect of third-party harm is irrelevant for ex ante decision-making. This suggests not only that the third parties are wronged if Cat is not saved, but that their rights are violated too.

Third-Party Harm eliminates wrongdoing, but similar conclusions follow when we return to the structure of Cornell’s original case. Consider a variation of Third Party Harm which is the same except that the trolley is loose due to Ali’s negligence. Imagine again that five people will suffer PTSD unless Ali saves Cat, then imagine ten people, twenty, thirty, a hundred, a thousand, and so on. None of his affects the gravity of the wrong perpetrated on Cat or Bea—in both cases, they will suffer the wrong of being negligently injured or killed. But the third-party harms give Ali some reason to save Cat rather than Bea, and thus Ali violates their rights if he saves Bea instead.

It might be thought that in saying all of this I am arguing myself into a corner. After all, if wrongs against unforeseeable victims in cases like Quarry 3 constrain ex ante conduct and are therefore also rights violations, does not this make the broadness problem even more acute? At least in Cornell’s fireworks case, the third-party harms are foreseeable. When they are unforeseeable, as in Quarry 3, surely positing rights against their suffering is even more unsupportable. If unforeseeable victims of my wrongdoing have rights against me, this means the class of people who possess such rights is implausibly large and is, by definition, difficult or impossible to predict. The problem is most severe for Nonrelational Fault 2, but it also applies to Nonrelational Fault 3. On the latter view, the “sufficiently similar grounds” condition mitigates the broadness problem, but there is still a potentially large class of persons who might unforeseeably be harmed.

It is important to remember we are working with a specific conception of rights, one that correlates to duties that constrain ex ante conduct. Possessing a right of this kind does not imply a right in any other, thicker sense. Nevertheless, we might think this exacerbates the broadness problem for a few reasons. For one thing, if unforeseeable victims in cases like Quarry 3 have rights, surely this is too many people for me to have in contemplation when I deliberate about how to act. Possessing a right means I must, in some sense, give due consideration to the right-holder, but there is a limit to the number of people who can feature in this way during my deliberation, otherwise, my practical reasoning would be hopelessly cumbersome. What is more, it is not even clear how I could take these potential victims into consideration. Since it is unforeseeable that they might be harmed, there are no clear precautions that can be taken. In this respect, they are unlike the foreseeable victims of my actions where the foreseeability of harm gives me a guide as to what preventive measures I might take.

However, I do not think any of this is a problem for Nonrelational Fault 2 or 3. To explain why, consider:

Clearly, I act wrongly by pressing the button. To whom do I owe the duty not to press it, and who possesses a right that I do not press it? Suppose we say: Everyone in the world. This does not mean that the right is possessed by some collective entity. Rather, it is an aggregation of billions of individual rights. A right/duty pairing exists between me and everyone in the world despite the fact that the one in seven billion chance of dying is far less than the chance of third parties suffering harm in all of the cases we have discussed so far.

Does this answer make rights and duties overly burdensome or incapable of ex ante guidance? I suggest not, because recognizing them does not necessitate my having everyone in the world in contemplation, at least in any granular sense, when I decide how to act. Seven billion people possessing the right rather than one makes no difference at all to my ex ante practical reasoning. In both cases the content of my duty is the same: I must not press the button. There is no additional action that must be taken, or that I must consider at the practical deliberation stage, by virtue of a right possessed by so many people. All these things are also true in Quarry 3. In order to respect the rights of people like Cat who will unforeseeably be injured by the explosion, Ali need not have them in contemplation when deliberating. Moreover, because the duty to Cat piggybacks on the duty to Bea and others like her, it is not more burdensome and does not require further action. Ali can respect Cat’s rights by respecting Bea’s. This is true of all rights conferred by Nonrelational Faults 2 and 3. They can all be respected by respecting other rights that already constrain the agent’s conduct.

There is another sense in which positing such rights might be too burdensome, which is that if people like Cat can claim compensation from Ali in Quarry 3, Ali’s potential reparative burden is much greater than it would be if they were unable to claim. However, we must distinguish the question of whether a right has been violated in the sense that a duty that constrains ex ante conduct was breached and the question of what response is owed as a result. This version of the burdensomeness objection applies only to liability to compensate, but nothing I have said so far entails that those who commit directed wrongs against unforeseeable victims are always required to compensate.

Whether there is a duty to compensate depends on the circumstances. This is because compensatory duties—like all duties—are subject to a stringency threshold. There is a maximum burden a wrongdoer can be required to bear for the sake of fulfilling their duties. For example, if forcing them to compensate would make them destitute and would not benefit the victim a commensurate amount, the duty to compensate does not arise.Footnote ¹⁸ In cases where there are many victims, the stringency threshold is likely to be met more quickly. On the other hand, if Cat is Ali’s only victim in Quarry 3 and compensating them would not be too costly for Ali, it is plausible that he has a duty to do this. So, the burdensomeness of liability to compensate is not a reason to reject the presence of directed wrongdoing and the violation of a right/duty pairing this implies. We can see this clearly by comparing compensation with less burdensome responses such as apology. In Quarry 3, if Cat is paragliding with twenty friends, all of whom are injured in the explosion, it is not implausible to think Ali has a duty to apologize to all of them, even if we do not think he ought to compensate all of them at overwhelming cost to himself. We can explain the lack of compensatory duties in terms of the stringency of Ali’s corrective duties, while the duty to apologize remains because apologizing is less burdensome.

There are other reasons to limit compensatory duties, especially when it comes to legal liability. Liability rules must be applied across a broad range of cases in nonideal epistemic conditions. In particular, the obvious difficulty of predicting liability to unforeseeable victims can create problems for agents in effectively managing their potential liabilities, and when these problems are systemic, rule of law values can become eroded. We cannot appeal to the implausibility of over-broad legal liability as a reason to doubt Nonrelational Fault 2 and 3 since there are plenty of other reasons to limit the liability of those who harm unforeseeable victims.

VIII. Effacing and Double Counting

The cases we have been discussing bear a resemblance to transferred malice cases in criminal law. A standard transferred malice case is like this:

According to the doctrine of transferred malice, Ali is guilty of murdering Cat even though he did not intend to kill them. His intention to kill Bea is “transferred” onto Cat. This doctrine is similar to Nonrelational Faults 2 and 3.

The doctrine of transferred malice is subject to some objections, and it is worth considering whether they also apply to Nonrelational Faults 2 and 3. One problem with transferred malice outlined by Patrick Tomlin is that, if Ali murders Cat, the doctrine mislabels the wrong involved in a way that erases Bea from the picture.Footnote ¹⁹ If Ali’s intention to kill Bea is “used” as an element in the wrong against Cat, making it the single wrong of murder rather than the dual wrongs of manslaughter and attempted murder, then the wrong against Bea is erased from the picture. Insofar as we are interested in the accurate labeling and recognition of wrongdoing, this is a problem. This objection might be raised in Quarry 3 and perhaps gives us further reason to doubt Nonrelational Faults 2 and 3.

This objection does not apply to Quarry 3, though, due to one important difference. According to our analysis, there are two wrongs, one to Bea and one to Cat. Bea is not erased from the picture because the wrong to her is not “used up” in any sense—it exists in addition to the wrong to Cat. Problem solved? Not quite. Now we run into another worry: Double counting.Footnote ²⁰ If Nonrelational Fault 2 or 3 is correct, the wrongful risk imposed on Bea in Quarry 3 is counted as part of the wrong committed against Cat. But if we are also to say that Ali wrongs Bea, this wrongful risk is counted twice, once for the wrong to Bea and once for the wrong to Cat. Perhaps it is unfair that Ali’s wrongful risk should be counted twice in this way.

The double counting objection is persuasive in contexts where we want to maintain a tight proportional relationship between the severity of the wrong and the burden the wrongdoer faces as a result. This is true when issuing punishment.

Being punished twice for the same malicious intent risks heaping a disproportionate burden on the wrongdoer. Although I previously noted that all duties have a stringency threshold, so it is possible for compensatory duties to become disproportionate, it remains the case that the burden of repair need not be tightly proportional to the severity of the wrong in the same way that the burden of punishment should be.

Suppose there are twenty people in the quarry when Ali sets off the explosion. In this case, Ali would wrong each of them and they would all have a claim for compensation against him. This is true despite the fact that Ali’s negligent act might have been no more culpable or unreasonable than if just one person had been injured. In a sense, his unjustified act is “counted” multiple times for each claim that can be made against him, but this is no bar to recognizing all these claims. Unless the reparative burden is so large that it exceeds the stringency threshold of Ali’s duties, there is no objection. In other words, the double counting objection applies to certain responses to wrongdoing such as punishment. It does not apply to other responses such as apology and compensation, or to the prior claim that a directed wrong has been committed.