Harm, responsibility, and enforceability

ABSTRACT

In this article I respond to the eight critical essays in this issue that evaluate the claims in my book with Gerhard Øverland, Responding to Global Poverty: Harm, Responsibility, and Agency.

KEYWORDS:

• Global justice
• harm
• exploitation
• doing and allowing
• duties to assist
• global poverty
• Peter Singer

I am very grateful to the contributors to this symposium on Responding to Global Poverty: Harm, Agency, and Responsibility for their thoughtful and searching comments. They raise many more interesting and challenging questions than I can address adequately here, but I will try to engage with their central concerns about our arguments in the book.

Our book is about moral responsibilities to address the hardships of those who are in a very bad state. Responsibilities, as we define them, are moral reasons that are quite stringent. I may have a moral reason to be very gracious to a stranger, but I do not have a responsibility to do so, whereas I do have a responsibility not to break their arm, and to prevent them from drowning. Responsibilities vary, however, in just how stringent they are. We characterize the stringency of responsibility in terms of how it constrains agents, and what it demands of agents who acted against these constraints. A responsibility is constraining to the extent that a person cannot justify acting against it by appealing to the costs to themselves of acting in accordance with it or by appealing to other valued moral ends that their conduct will bring about if they act against it. A responsibility is demanding insofar as those who have acted contrary to it are required to take on quite significant costs to address the effects of their having done so.

Harming, exploiting, and giving rise to cost

Siba Harb and R. J. Leland raise objections to our treatment of exploitation and, more generally, to our treatment of the concept of harm. They begin by claiming that, on our view, ‘Affluent countries allow harm to befall the poor and even enable such harm, but they do not cause harm to the poor’ (Harb and Leland, 2019, 1). Against this, they argue that by exploiting the poor we cause them harm, invoking non-comparative accounts of harm. Non-comparative accounts of harm affirm that it is a sufficient condition of A’s harming B that A’s actions cause B to be in a certain type of bad state. It need not be a worse state than B was in previously, nor a worse state than they would have been in absent A’s intervention.¹ And since being exploited involves being caused to be in a bad state – in ‘agency impairing situations’, as Harb and Leland put it – an exploited person can be harmed even if they are made better off than they would have been without the exploitation (Harb and Leland 2019, 4).

Harb and Leland note that, while on our view exploitation is morally problematic, we do not regard responsibilities with respect to exploitation to be as stringent, all else being held equal, as responsibilities not to do or enable harm. They argue, however, that since exploitation constitutes a kind of harming, responsibilities arising from the exploitation of others are comparably stringent to those relating to harming people in these other ways. In addition, they argue that because exploitation is a kind of harming, responsibilities arising from engaging in exploitation are, contrary to our account, much more stringent than those arising from culpably failing to assist. Harb and Leland hypothesize that we have gone astray in our treatment of these issues because we have accepted a rather crude counterfactual account of harm, according to which the only manner in which we can contribute to harming a person is by making them worse off than they would otherwise have been (Harb and Leland, 2019, 4). It is this understanding of harm, they think, that leads to our failure to treat exploitation as a kind of harm.

Let me begin by setting aside what I regard as only apparent disagreements between us. First, we do not hold the view that affluent countries are not causing harm to the poor, but may merely be failing to prevent it. Harb and Leland seem to assume that, on our view, enabling harm is a way in which we fail to prevent harm, rather than causing it. Some theorists do treat enabling harm as a species of allowing it (see e.g. Rickless 2011). But we argue, on the contrary, that enabling is a manner in which an agent can contribute to harm. If an affluent country sells arms to a dictatorial regime that uses them to terrorize its population, it contributes to harm by enabling it to occur. In this case, the enabling takes the form of knowingly providing others with the means of doing harm – which we argue is a particularly problematic way of contributing to harm (Barry and Øverland, 2016, 168–169).

Second, we are not committed to a counterfactual account of what it means to contribute to harming others. Indeed, our treatment of individual responsibility in overdetermination cases commits us to the view that you can contribute to harm without this harmful outcome depending counterfactually on your conduct (Barry and Øverland, 2016, ch., 11). You can be a contributor if you are part of the set of people who actually brings an outcome about. If 10 people push a boulder off a hill, crushing those below, when the efforts of any five of them would have been sufficient for this result, each contributes to these deaths without making a difference as to whether they will occur. The incompleteness of crude counterfactual accounts of harm is also evident in cases of late pre-emption – if Jill shoots Bob she harms him, even if it is true that had Jill’s shot not hit Bob, Andy’s shot would have.² So we do not reject the view that you can contribute to non-comparably bad states without making a difference to whether they occur. Let us grant that contributing to non-comparably bad states is a form of harming. But that still does not settle the question of how stringent our responsibilities are not to engage in it. Some instances of harming make those who are harmed worse off than they otherwise would have been, while others do not; and it is an open question whether these different types of harming are morally equivalent.

Harb and Leland are correct that we do not regard responsibilities not to exploit to be as stringent as responsibilities not to do or enable harm. Our reason for holding this position relies on a central normative concept in our book, that of giving rise to cost. Giving rise to cost is a comparative notion. A person gives rise to cost when his or her location, movements or agency have as a consequence that a person will suffer some deprivation – extreme poverty or death in most of the cases we discuss in the book. If a person gives rise to cost, then removing that person from the scene will make it the case that an innocent person will not suffer that deprivation. What distinguishes those who exploit (in the cases that interest us) from those who do or enable harm is that exploiters do not typically give rise to cost, unlike doers and enablers of harm.³ It is not the presence or relevant action of an exploiter, in the cases we discuss, that causes the extreme need of the exploitee. The presence of the exploiter, on the contrary, provides an opportunity for the exploitee to escape her predicament. It is in this sense that those who exploit are akin to those who can assist people in need but who have not contributed to the plight of these people. As Harb and Leland make clear, we focus in the book on only one class of cases of exploitation, but there are others. Many of those who exploit in the real world do give rise to cost – they simultaneously exploit and enable harm by preventing others from assisting the victims without extracting benefits from them. On our view, such people have more stringent responsibilities than those who exploit without giving rise to cost.

Why should it make a moral difference whether a person gives rise to cost? Well, a person who is now in a state of severe need can say to those who have done or enabled this harm to her, ‘look, I would be doing just fine if you were not around, but as a result of your acting in various ways I am now in a state of extreme need. You made it the case that I am now in peril.’ This provides a powerful reason to ascribe stringent responsibilities to doers and enablers of harm. It does not follow from this, of course, that those who exploit do no wrong. Indeed, we assume that those who exploit in the cases we discuss are indeed wronging those they exploit. They are wronging them because they extract benefits from them as a condition of helping them escape their state of need, rather than simply assisting them when they have a responsibility to do so.⁴ Those who fail to assist people when they have a responsibility to do so also wrong those they fail to assist, even though they do not give rise to cost. In these senses, those who culpably fail to assist are akin to those who exploit.

Harb and Leland worry that by treating exploitation on a par with culpably failing to assist we are somehow downplaying or denying its wrongness. But we do not claim that ‘Exploitation is permissible since the harm of basic needs deprivation is worse than the harm associated with accepting the exploitative option’ (Harb and Leland, 2019, 8); we explicitly reject any such claim in the cases we consider. Moreover, we argue that culpably failing to assist can be a very serious offence. In some extreme cases, we claim even that those in need might be permitted to kill those who are failing to assist them as a means of compelling them to act (see e.g. Barry and Øverland, 2016, 42–43). So responsibilities not to exploit are quite stringent, even if they are not as stringent as those involving giving rise to cost by doing or enabling harm.

Harb and Leland think that those who exploit act worse than those who culpably fail to assist. But why should we think this? I grant that exploiters harm the exploitees, whereas those who fail to assist do not harm them. Yet while exploitation may constitute a non-comparative harm, since it causes people to be in a non-comparatively bad state, it also confers a comparative benefit on them, one that they voluntarily seek. Failing to assist, on the other hand, may not be a non-comparative harm, but it also fails to provide a comparative benefit. Harb and Leland are surely right that exploiting people ‘characteristically involves putting the exploited in some agency-impairing situations, situations that involve demeaning choices or ongoing asymmetric vulnerability’ (Harb and Leland, 2019, 4). But culpably failing to assist leaves people in even worse situations, thus also leaving them to face demeaning choices and asymmetric vulnerability. If we are to compare cases of exploitation, where the exploited party is somehow locked into ongoing vulnerability, with cases in which those who go unassisted are not, we are no longer comparing like cases with like. Do exploiters act worse than those who culpably fail to assist? We think that it is not at all obvious. Would a person in need have more reason to resent a person who fails to offer them assistance unconditionally, but does provide them with a means of escaping their situation so long as they work on their behalf, than a person who fails to offer them assistance at all? Would we judge the exploiter more harshly than someone who simply leaves the person in need to suffer? I suspect that the answers may sometimes, but not always, be yes. There may be particular forms of exploitation that strike us as so degrading to the person who is exploited that it is morally worse to exploit them than to leave them in an even worse state. But not all cases of exploitation are of this sort. And of course, failing to assist can take on degrading forms as well – such as when a person looks on with amusement as someone drowns when they can save them with the slightest effort.

As Harb and Leland note, those in need would have more reason to welcome the presence of exploiters than of those who simply culpably fail to assist them. This surely gives some clue to the moral status of these categories of wrongdoers. Another clue is provided by considering what might be called secondary responsibility. Secondary responsibility has to do with what you ought to do, given that you are not going to do the thing you really ought to do. Once you have limited your choice set by ruling out alternatives required at the bar of primary responsibility, what then ought you to do? In the cases we are discussing, the primary responsibility of both the exploiters and those who culpably fail to assist is to simply help those in need without demanding anything in return. Secondary responsibility is clearly a rather different kind of obligation since even those who comply with it are not acting as they really should. When we say that if you are going to fight an unjust war, you ought to avoid killing civilians, we are not saying that if you abide by that precept you are worthy of moral praise; you have clearly not done what you really ought to do. But still, there is a meaningful sense in which you have complied with a relevant kind of responsibility.⁵ And we should certainly encourage you to comply with your secondary responsibility when it becomes clear you are not going to act on your primary responsibility. In the cases we are considering, the critical question is: what should a person do given they are not going to do the thing that they really ought to do (assist the person in need)? If the options in this limited set are to exploit them and make them better off or do nothing and leave them worse off, the morally preferred option may well be to exploit, at least so long as the form of exploitation is not extremely degrading.

Agents of responsibility

Holly Lawford-Smith notes that we refer to the affluent many times in our book, but seem to pick out different things with this term, suggesting that ‘we are unclear on the question of who it is that has moral responsibility for poverty’ (Lawford-Smith, 2019, 1). More worryingly, she thinks that any of the three ways we might specify the idea of the affluent creates problems for various conclusions we try to reach regarding responsibilities to address poverty abroad.

Our aim in the book is in the first instance to engage individual affluent people. We encourage them to consider their reasons for action with respect to poverty, including reasons to assist the poor when they can do so at moderate cost; to refrain from contributing to poverty; and to offer compensation when they have failed in their responsibilities to poor people. We do, however, sometimes find it useful to refer to uncoordinated groups of affluent people, where the groups are defined by some feature that is relevant to our analysis. One such group is the group of affluent people who fail to take on moderate cost to address poverty; another is the group of affluent people who promote policies and conduct that aggravate poverty or prevent the poor from improving their lives. And sometimes we discuss the policies of affluent states (where the majority of affluent people live), because it is often, though not always, through the intermediation of the policies of affluent states that affluent individuals contribute to poverty abroad.

According to Lawford-Smith, to be in any way responsible for the occurrence of poverty, some entity must have both caused it and exercised control over it. ‘What we need to know now’, she writes, ‘is who causes, and has control over causing, global poverty’ (Lawford-Smith 2019, 1). She worries that none of the candidates referents for the term ‘affluent’ that are employed in our work can cause or have control over causing global poverty. She thinks it makes little sense to say that uncoordinated groups of individual affluent people can either contribute to poverty or fail to assist the poor:

To assess whether a prevention comes at moderate cost, to prevent something, to create an obstacle or undertake an action … all require a level of coordination, and indeed the capacity to act intentionally, that uncoordinated groups simply do not possess. (Lawford-Smith 2019, 3)

It only makes sense to speak of a group as bearing a responsibility if it has an internal structure and relatively sophisticated capacities. But is this a plausible position? Consider first the case of assistance-based responsibilities. Suppose that poverty can be alleviated if each member of some uncoordinated group A of affluent people makes a very slight contribution to poverty relief, but no member of A does so. Surely we can say here that A could have prevented poverty at moderate cost, but failed to do so. The cost was moderate, since the uncoordinated group could have achieved the goal had the individuals that constitute it taken on the modest costs that each was morally required to bear. In this case, A need not develop sophisticated capacities or plans; its members have only to make modest financial contributions to agents (e.g. charity groups, local institutional actors) that could provide assistance or directly to the poor themselves.

We argue in the book that the responsibilities of those who culpably fail to assist can compound over time. Lawford-Smith disputes, however, that this could hold true for uncoordinated groups. Let us imagine that an emergency situation arises as a result of extreme drought and there is a risk of famine. Suppose that no member of A takes on cost to help avert the famine when this can be achieved at very little cost at time t₁, and that members of A are not taking on cost to address extreme poverty elsewhere. Witnessing this, it seems fair enough for a commentator to say, wow, the affluent are really failing in their duties to assist. When, at t₂, people start to perish in the famine, it seems perfectly reasonable to say that A bears some responsibility for this, since any of its members could have helped to avert this problem at an earlier time without exceeding the costs that they were required to bear. Since they culpably failed to prevent famine at t₁, it seems plausible that they must now bear greater costs to address it than they were required to bear in the first instance.

Consider next responsibilities based on contribution to poverty. Suppose that A is an uncoordinated group, but its members live for the most part in wealthy societies. Members of A pay taxes to their governments, participate in the political and social lives of their societies, and so on. What further defines membership in A is that its members support punitive tariffs on exports from poor countries and encourage economic policies that cause severe environmental damage which disproportionately affects the poor abroad. Now suppose that, because the members of A are numerous and influential, the governments of the countries in which they live implement policies that impose extreme hardships on the poor. Here it seems perfectly sensible to say that A contributes to harm. How? Because each of its members has provided the means for various collective agents to enable or do harm to the poor. Were it not for the efforts of members of A, different policies would have been adopted that did not have such negative effects on the poor. Correspondingly, on our view, all members of A have stringent contribution-based responsibilities to address these harms.

There are, of course, other (similarly uncoordinated) groups for which it would not make sense to say that they contributed to poverty or culpably failed to assist the poor. This includes groups of people who do indeed take on moderate cost to help the poor, and who struggle to protect the poor from various policies that might otherwise harm them. In our famine case, imagine that another group of people, H, provide as much as they are morally required to help avert the famine. Even if the famine eventuates, we cannot say that H’s responsibilities are compounding, since the famine is not due in part to H’s culpable failure. In principle, then, we have no trouble distinguishing unstructured groups whose members fail in their responsibilities to the poor from those that do not.

What of affluent individuals? Lawford-Smith allows that individuals can fail to assist, but she denies that they ordinarily do or enable harm. Her argument relies on the idea that, for an individual to have contribution-based responsibilities to the poor, it must be the case that their conduct made the difference as to whether poverty occurred. She writes:

An affluent individual will rarely if ever enable harm to an individual poor person. Even if what she does intersects with what others do in a way that has the effect of enabling harm to the poor, this does not suffice for her enabling harm to the poor (they do, taken together, but she doesn’t). (Lawford-Smith 2019, 3–4)

However, this claim rests on a rather narrow conception of what it can mean for an individual to be morally responsible for harms to which they have contributed. Consider how it would apply to simple cases of overdetermination and pre-emption. Suppose that 10 people, acting independently and without coordination, put waste into a river, raising the toxicity of the water downstream and fatally poisoning Sally. Assume that the waste released by any five of them would have been sufficient to cause Sally’s death, given the toxicity of each load. In this case, no individual makes a difference to whether those downstream are harmed. Surely, however, we should hold each of the 10 morally responsible for contributing to the resulting harms. And this is so whether they are an uncoordinated group of people or are acting together with sophisticated capacities to further some plan. Were any of the 10 to try to get themselves off the hook by claiming that the harm would have occurred even if they had not dumped their own waste, we would rightly scoff. That fact that there are five more of them than are necessary to bring about this harm does not render all 10 immune to compensatory claims. Note also the normative characteristics of the backwards-looking responsibilities of each individual in this case. Each has a very stringent and enforceable responsibility to compensate those whose harm they have contributed to.

Lawford-Smith accepts that affluent countries can indeed do or enable harm, or exploit the poor. She thinks the difficulties lie in shifting from talk of the responsibilities of states to the responsibilities of their citizens. I agree that not all members of a society can be held morally responsible when a state enacts policies that harm others. I disagree that no responsibility can be attributed to those members of society who support such policies. Lawford-Smith thinks that, ‘Of those who vote along with the majority, only some proportion of these will have been necessary to seeing the winning party elected’, and continues:

So first, there’s the chance of any individual being a member of the set of individual citizens who were causally relevant in electing the country’s government. But second, there’s the chance that the country itself is causally relevant in deciding the international trade rules (or reform to them)! (Lawford-Smith 2019, 6)

Again, her argument relies on a counter-intuitively restrictive conception of contributory responsibility, this time within structured groups. Imagine that a group of 10 countries is deciding through a majoritarian procedure whether to keep in place trade rules that harm the poor. Australia goes along with seven other affluent countries and votes to keep them in place. On Lawford-Smith’s account, this makes it causally inefficacious. But surely Australia and all of the other countries that voted in favour helped bring this outcome about, and each bears responsibility for its consequences. Compare this to a case where 10 of us meet to decide whether to hire a hitman to break the legs of one of our rivals. We decide by majority vote. Eight of us vote to hire him, and he carries out the hit. No one of us was a necessary condition for the hiring of the hitman, but all eight of are surely responsible – along with the hitman – for the harm that results, since each helped to bring it about (see e.g. Goldman 1999; Tuck 2008).

Contributing to harm and difference-making

How stringent are our reasons not to contribute to harm when we know that we will not make a difference to its occurrence? ⁶ We argue that the stringency of our responsibilities is reduced, but is still significantly greater than for those who are in no way involved in producing that harm. Our reason for this position again relies on the moral significance of giving rise to cost. In overdetermination cases, none of the individuals, on their own, gives rise to cost, since it is true by stipulation that if any one of them were absent from the scene the deprivations would still occur. Intuitively, it seems easier to override reasons to contribute to harm when you know you will be an overdeterminer by appealing to the cost to the agent of refraining or the greater good she can bring about through non-compliance.

Consider again the case of Sally. Susan is one of the 10 people upstream. She knows that if she puts waste in the river, she will make no difference to the harm Sally suffers, because she knows that enough of the others will add their waste to kill Sally. If Susan knows that by adding her waste she can save two other innocent people who will otherwise die, then it seems permissible for her to do so. It does not seem permissible, however, for Susan to dump her waste if she knows that her contribution is necessary for Sally to die, even if that will result in saving two other innocent people. And this would seem so even if it took the contribution of all of 100 people, or some greater number, to kill Sally. That Susan’s particular contribution depends on other factors to be effective does not seem to alter the stringency of her responsibilities when she knows that these factors obtain. If, on the other hand, Susan’s conduct in throwing her waste into the river does not make anyone worse off, that defeats one very strong reason for objecting to it. There are other reasons to object to it – she will be a member of the set of people who actually bring about Sally’s death – but these reasons can be more easily overridden by reasons to protect other innocent people who will otherwise be harmed (Barry and Øverland, 2016, 240‒243).

Anne Polkamp worries that our account of individual responsibility in cases involving overdetermination is incomplete. More specifically, she thinks that our account will generate reasons for individuals who may contribute to harm in overdetermination cases that ‘may be too easily outweighed by the costs of refraining and the benefits the conduct would bring’ (Polkamp 2019, 1). According to Polkamp, this is because an individual’s contributions to harm ‘can only be very limited in large-scaled overdetermined harm cases’ (Polkamp, 201x, 1). The premise of Polkamp’s critique is that a reason not to contribute to an outcome will not be very stringent in case the magnitude of the contribution will be small due to the fact that there are a very large number of other contributors.

Suppose that Sally is harmed because each of a great many people dump their waste upstream. Suppose further that none of them has a pressing need to do so; it is just a bit more convenient and less costly than taking it to the town dump. Each makes a very small contribution, so it will take very many to harm Sally. Assume that Susan did as all the others do, dumping a small amount of waste into the lake. In this case, it would seem, she has a responsibility to compensate for the harm Sally suffers. The reason is that she should have been willing to bear the inconvenience and minor cost of refraining from helping to bring this outcome about. It would, of course, be fairest if the compensation were to be shared equally among all of the contributors and not fall entirely on Susan. However, it is not clear that Susan would be required to take on only a minuscule fraction of the cost of compensating Sally, should it turn out to be impossible to collect from the others.

Just how constraining are responsibilities not to contribute to overdetermined harms? It seems the minor inconvenience of taking her waste to the town dump is not a cost that Susan can appeal to in order to justify acting against the constraint. What would justify it? Perhaps if some serious harm to Susan or her loved ones were at stake. What if dumping her waste in the river was needed to save her job? Even in this case, it is not clear that this would permit Susan to act in a way that amounts to overriding the overdetermination-based constraint. Susan might, all else being equal, be justified in acting against it, but in such a case it still seems that there is a moral remainder (see e.g. Feinberg 1978). Appealing to the cost of losing her job would not, intuitively, absolve Susan of compensatory responsibility to Sally. In such a case Susan would, rather, have engaged in a justified infringement of Sally’s rights (see e.g. Thomson 1990, 91–97). Does the fact that agents like Susan might, all things considered, be justified in acting against her reasons not to help produce harm in overdetermination cases show that they are flimsy, as Polkamp suggests? Our example suggests that even when large numbers of people are involved, reasons to observe such constraints can require that we be willing to bear significant inconveniences to avoid acting against them. It also indicates that even when we are, all things considered, justified in acting against them by appealing to costs to ourselves, we still seem to bear compensatory responsibilities to those whose harms we contribute to.

Let us consider the implications of our discussion for issues like responsibilities not to contribute to global poverty. For ordinary affluent people, there are not, typically, very high costs associated with refraining from activities such as supporting policies and politicians whose efforts impact negatively the prospects of the poor, or of taking on some costs to offset the various ways in which they may indirectly support collective agents that carry out such policies (e.g. through tax contributions to governments). So the responsibilities that we think would emerge from our analysis would require significantly more of most ordinary people, and that is a non-trivial result.

Polkamp has a deeper worry about our approach:

The question arises why being part of the actual set that causes harm is wrong at all. If the actions of a set of people together cause harm, it seems plausible to say that the set has acted wrongly. It does not follow, however, that it is wrong for an individual to act in a way that makes her part of the set. (Polkamp 2019, 8)

I agree that it does not follow. However, it is very difficult to vindicate various widely shared ideas about compensation and the preventive use of force without assuming that each individual has indeed acted wrongly. Suppose that a set of 10 people act in a non-coordinated fashion with the result that Sally is severely harmed, when the acts of any seven of them would have been sufficient to severely harm her. In this case, no individual’s act is necessary for the harm to occur. Polkamp wants to say, rightly I think, that the set acted wrongly – Sally’s rights were violated. I assume that as a corollary of this the set would owe compensation to Sally. It cannot be that just because there were 10 rather than seven of them that acted in this way that Sally’s claim to be made whole is diminished or eliminated. But what will it mean for the set to compensate Sally? Presumably, it will mean that each member of the set is liable to contribute to such compensation. I take it that none of the members of the set would have grounds for complaint were they compelled to offer compensation to Sally, and would not be wronged if this were to occur. But how is it possible that, if none of the members of the set has done wrong, each can have a compensatory responsibility imposed upon them without infringing their rights?

Consider the issue from a different angle, the preventive use of force. If each of the 10 knows what they are doing and the harm that will result, it seems plausible that we can use proportionate force to prevent them from harming Sally without infringing the rights of any of them. Indeed, if it is necessary to save Sally, it seems quite a bit of cost could be imposed on each of the 10 without infringing their rights. Yet, ordinarily, forcibly imposing considerable costs on people who are innocent of any wrongdoing does infringe their rights. We could not, for instance, impose as much cost on a group of innocent bystanders for this purpose without infringing their rights. So it is very difficult to vindicate views that we seem to share about the responsibilities of sets of people without committing ourselves to judgements about the behaviour of the individuals that constitute those sets.

Enforceability

Susanne Burri and Lars Christie contrast two different approaches one might take to the enforceability issue: the Asymmetry View and the Symmetry View (Burri and Christie, 2019, 1). According to Asymmetry, assistance-based and contribution-based responsibilities differ with respect to their enforceability. Symmetry denies this. What does it mean for responsibilities to differ in terms of their enforceability? As they explain it, a responsibility is enforceable

just in case an agent who has culpably failed to live up to it is not wronged if costs are imposed on her to secure the moral good that the responsibility was meant to protect. The enforceability of a responsibility increases with the size of the costs that an agent can be required to bear if she has culpably failed to discharge that responsibility. (Burri and Christie, 2019, 4)

Assuming this understanding of enforceability, they argue for Symmetry and against Asymmetry.

Burri and Christie diagnose what they take to be a tension in the account of enforceability we employ in our book. They claim that in Chapter 3 we seem sympathetic to Symmetry in our discussion of agents who culpably fail to prevent harm, but point out that we do not discuss the enforceability of assistance-based responsibilities in Chapter 9, limiting our discussion there to the implications of failing to act on contribution-based responsibilities, which, they say, ‘suggests that Barry and Øverland endorse an Asymmetry View’ (Burri and Christie, 2019, 2).

In Chapter 9, we argued for a conditional claim, which we called Force Implication. Force implication states that if we possess stringent contribution-based responsibilities to the poor, and then we are liable to the use of defensive force exercised on behalf of the poor. Because this claim is only about contribution-based responsibilities, its truth or falsity does not imply anything about the enforceability of assistance-based responsibilities. Nothing here commits us to Asymmetry, but perhaps our silence can be read as suggestive. Burri and Christie are right, in any case, that we could have been more explicit about our position on enforceability: it is, in fact, a variant of Asymmetry.

According to Burri and Christie, Symmetry holds that ‘contribution- and assistance-based responsibilities are similarly enforceable if we hold fixed the moral good that the responsibilities protect’ (Burri and Christie, 2019, 6), and clarify what they mean by responsibilities being similarly enforceable:

whatever costs you are liable to have imposed on you if you culpably fail in your contribution-based responsibilities to some person, you are also liable to have imposed on you if you culpably fail in your responsibilities to assist them, when all else is held equal. (Burri and Christie, 2019, 5)

According to Symmetry, then, if you make yourself liable to be killed when you culpably fail to discharge your contribution-based responsibilities (say by culpably pushing an innocent person into a pool, knowing that they will drown, when killing you is the only means of saving the drowning person), then you are also liable to be killed if you culpably fail to discharge your responsibility to assist in averting the threat (say by culpably failing to warn a person that they were about to fall into a pool when you know that they would thereby drown).

Should we endorse Symmetry? Burri and Christie point to what they regard as the cost of rejecting it. For instance, they claim that if ‘scope of Force Implication extends only to contribution-based responsibilities, then the Asymmetry View is correct, and assistance-based responsibilities are never permissibly enforceable through the use of significant force’ (Burri and Christie, 2019, 6). But rejecting Symmetry does not carry any such implication. We can reject Symmetry but hold that if we possess stringent assistance-based responsibilities to the poor, and then we are liable to the use of significant defensive force. Indeed, our endorsement of the idea that a person who culpably fails to assist can have significant cost imposed upon them if this is necessary to save the person they have failed to assist is compatible with holding that greater cost can be imposed on those who have culpably contributed to a person’s plight, all else being held equal. We can also reject what Burri and Christie call Symmetry while endorsing a principle that encompasses a different kind of symmetry, namely that whenever some agent has a responsibility to take on a certain amount of cost to assist another in severe need, and does not voluntarily do so, others can impose this cost on that person without wronging them, whether or not the responsibility is contribution- or assistance-based. We can thus treat assistance-based and contribution-based responsibilities as on a par with respect to enforceability in various respects without endorsing Symmetry. So we can, pace Burri and Christie, say many things about the enforceability of responsibilities to assist without endorsing Symmetry.

So far, I have tried to undermine the Burri and Christie case for Symmetry by pointing out that the costs of rejecting it are not so great as they suppose, but I have not made a positive case against it. The first reason to reject Symmetry, as they define it, is that it seems to fit oddly with any view –, and we endorse such views – that doing harm is worse, all else being held equal, than allowing harm. That is, it seems plausible that the amount of force that can be used against a person when they act wrongly should reflect how serious the wrong is that they have committed. So if it is worse, all else being held equal, to kill a person than to fail to prevent their death, then the force that can permissibly be used to prevent a person from killing should generally be greater than can be employed against a person who will fail to prevent a death. And this seems intuitive. It seems permissible to kill a negligent motorist (they failed to regularly check their brakes) when their car has spun out of control and is heading towards you when this is the only way to prevent their car from permanently disfiguring you; but it seems counter-intuitive that you could kill a bystander who negligently failed to warn you so you could get out of the way of the car, even when this warning is necessary to prevent your death. There also seems to be asymmetry in the costs that an agent is required to assume to make amends for culpably contributing or failing to prevent harm. If you negligently knock someone into the pool, where they will now drown if you do nothing, you are required to make very large sacrifices if this is necessary to save them. The same does not seem to be the case of a person that negligently fails to warn someone who is now at risk of drowning (he heard what he took to be the sounds of a person struggling, but failed to investigate).⁷

The reason why there can be asymmetries in enforceability, on our view, is that those who contribute to harm give rise to cost, while those who fail to prevent it do not. Of course, if the degree of culpability of those who fail to assist and those who contribute to harm differ, things can be quite different. It seems permissible to impose more cost on a very culpable bystander than on an innocent or even mildly negligent contributor. Asymmetry is compatible with this claim. And it is also compatible with the claim that sometimes the enforceability of these different types of responsibilities will be the same. There are cases where we might be justified in imposing the same very high cost – death – on those who fail to assist as on those who contribute. This may be most plausible in cases where the agents are extremely culpable – actively intending the harms to which they either contribute or fail to prevent – James Rachels’ famous bathtub case is a paradigm example. In such cases, the sheer malice of those involved in allowing harm seems to drown everything out (see e.g. Trammell 1975). However, most cases of culpable contribution to or failure to prevent poverty are not like this. Consequently, the enforceability of responsibilities not to contribute will be greater when it comes to poverty than responsibilities not to fail to prevent, all else being held equal.

Doing, allowing, and enabling harm

To understand the significance of contribution-based responsibilities, we need a clear account of the distinction between contributing to and failing to prevent bad outcomes – often referred to as the distinction between doing and allowing harm. There are two main philosophical tasks with respect to such a distinction. The first is analytic – it would be desirable to have a clear and intuitively plausible account of what it means. The second is normative – we need to discern whether it possesses basic moral significance. In Chapter 5 of our book, we take up the analytic task. Rather than trying to draw an exhaustive binary distinction between doing and allowing harm, we argue that it is more fruitful to draw a tripartite distinction between doing, allowing and enabling harm. We provide an analysis of these three relations and supply empirical evidence that this tripartite distinction is also intuitive to non-philosophers.

Fiona Woollard raises some challenges to our analysis of these distinctions, and proposes a different analysis of the doing and allowing distinction.⁸ The differences between our analyses can be illustrated by a suite of cases in which an agent removes a barrier that is preventing harm to another person.

Drive-Away (Victor’s Car): A boulder is rolling towards Victor who is trapped on the slope below. Victor’s car is in the path of the boulder and, if it remains where it is, will bring the boulder to a halt. Bob drives Victor’s car away. The boulder hits Victor and crushes him to death.

Drive-Away (Bob’s Car): A boulder is rolling towards Victor who is trapped on the slope below. Bob’s car is in the path of the boulder and, if it remains where it is, will bring the boulder to a halt. Bob drives his car away. The boulder hits Victor and crushes him to death.

On our analysis, Bob enables harm to Victor in both of these cases. He enables harm by removing an obstacle that would have been sufficient to block a causal process (already in train) that harms Victor. The answer to how Bob was relevant to Victor’s death is simple, and refers to something he did: he drove away a car that was preventing Bob from being harmed. That is why, we think, in neither case should we regard Bob as merely allowing harm to occur. But he does not do harm, on our analysis, because he is not linked to Victor’s death via a complete causal process (as he would be, for instance, if he pushed the boulder down the hill, crushing Victor). Woollard, on the other hand, treats Drive-Away (VC) as a case of doing harm, and Drive-Away (BC) as a case of allowing harm. Woollard draws a binary distinction between doing and allowing harm, and treats enabling harm as a kind of allowing harm.⁹ So on her analysis of doing/allowing ‘all enablings count as allowings, but only some cases involving the removal of a barrier to harm count as enablings’ (Woollard 2019, 8). Whether the removal of a barrier to harm counts as doing harm depends, Woollard suggests, on whether it is part of a sequence of harm (Woollard 2019, 9). To be part of a sequence of harm, on her analysis, some fact about an agent’s behaviour (in this case the fact that Bob removes a barrier to harm) must be substantial or relatively substantial (Woollard, 2015, 35, 80‒1). To be substantial, a fact must be either positive or contrary to our normal presuppositions. To be relatively substantial, a fact must be about the absence of a potential barrier to harm such that either (a) the victim has the claim on the barrier or the victim obtains the privilege (liberty) to use the barrier from the third party; or (b) ‘the barrier does not require the continued use of resources belonging to the agent and the victim has a non-need based claim on the barrier’ (Woollard 2015, 35; 2019, 9). Accordingly, if the car belongs to Victor (as in Drive-Away VC), then Bob does harm to him, since Victor has a claim on the barrier. If the car belongs to Bob, on the other hand (as in Drive-Away BC), then he allows Victor to be harmed (Victor does not have a claim on the barrier, and he does require the continued resources that belong to Bob).

Is Woollard’s treatment of these cases superior to ours? Let us consider first the intuitive evidence. Woollard seems to regard it as an advantage of her approach that it tracks what appear to be important moral differences between cases like Drive-Away (VC) and Drive-Away (BC). She writes: ‘not all removals of barriers are morally equivalent. The non-need based claims (for example claims based on ownership) which the relevant individuals have over the barrier matter enormously’ (Woollard 2019, 12). I agree that not all removals of barriers are morally equivalent. But this does not make it counter-intuitive to treat all removals of barriers as enablings. It is perfectly sensible to regard some instances of enabling harm as much morally worse than others. On our approach, the moral status of some instance of enabling harm (in this case through preventing a barrier from averting harm) depends on further facts of the situation. One such fact is whether the car that Bob drives away is his or whether it belongs to Victor. In Drive-Away (VC) Bob enables harm to Victor in a way that violates Victor’s property rights. This is not the case in Drive-Away (BC). In addition, the cost to Bob of refraining from removing the car is substantially higher if he owns the car (Drive-Away BC) than if he does not (Drive-Away VC), and this too makes a morally significant difference.¹⁰

That different enablings of harm can differ morally from each other but still belong to the same analytic category should not be surprising. The same is true, after all, of clear-cut cases of doing or allowing harm. For example, if Bob allows someone to be harmed when he can rescue her at very little cost, then his behaviour is surely much worse than if he allows her to be harmed when he can rescue her only at the cost of losing both of his legs. But the fact that he can prevent her from being harmed at very little cost and is therefore blameworthy for failing to do so does not make his conduct any less an allowing of harm. In sum, we agree with Woollard that questions of property rights are relevant to the moral status of different cases of enabling harm; we disagree that property rights should figure in determining whether an agent does, enables, or allows a harm to occur.

Another important respect in which our analysis differs from Woollard’s is that we regard enabling harm as a distinct category from doing and allowing harm, while she treats enabling as a species of allowing harm (ee e.g. Rickless 2011; Foot 1994). She claims that ‘while enabling and clear allowing are morally distinct, they have enough in common to both be treated as cases of allowing for the purpose of drawing a binary distinction between doing and allowing’ (Woollard 2019, 11). We disagree, and not because (as she suggests) we use the word ‘allow’ differently from her (Woollard 2019, 11). Rather, we think that enabling and allowing differ in a crucial respect: enablers of harm give rise to cost; allowers of harm do not. Those who allow harm do not make the victims worse off by their presence, or through anything that they do. If those who allow harm were to have been absent from the scene, the victims of harm would be no better off. That is not true, however, of enablers of harm. The victims of harm would indeed be better off if those who enable the harm were absent from the scene.¹¹ In this respect, enablers of harm are more akin to those who do harm than to those who allow it to occur. This seems a very good reason to treat enabling separately from allowing.

There is another important respect in which our analysis differs from Woollard’s. Her analysis of doing and allowing harm depends for its initial application on our having already done some moral thinking, in particular about the location of property rights. Our analysis does not. Whether this counts in favour of our analysis depends, I think, on the broader question of what we are analysing in such distinctions. We took it as our starting point that the purpose of distinctions like doing/allowing harm is, as Jonathan Bennett put it, ‘to separate out two classes of situation so that we can do some basic moral thinking about them’ (Bennett 1981, 48). We assumed that debates about whether doing harm was worse than allowing harm concerned the importance of a clear and objective distinction that did not itself depend on prior moral judgements. Bearing this in mind, we sought a way of drawing the distinction that is, in Bennett’s words, ‘statable in terms which are clear, objective, and deeply grounded in the natures of things’ (Bennett 1995, 137). It is controversial whether doing harm is worse, all else being held equal, than allowing harm, but it is not controversial whether behaviour that infringes property rights is worse, all else being held equal, than behaviour that does not. So it would be odd if it turned out that our very drawing of the distinction involved something that was obviously morally relevant.

Responsibilities to assist

Kasper Lippert-Rasmussen and Bashshar Haydar raise interesting challenges for our treatment of assistance-based responsibilities.

Stringency

Just how stringent are responsibilities to assist those in severe need? Are we, as a general principle, required to make very large sacrifices when this is necessary to save others in need, regardless of our relationships with them? We argue in our book that, unless certain conditions are satisfied, we are not morally required to make very large sacrifices to save others, including the life of a child in front of us, absent any special responsibilities to them.¹² Instead, we claim that we are ordinarily morally required to make modest sacrifices to save others who are in severe need through no fault of their own. We call our principle of required assistance the Moderate Assistance Principle (Moderate, for short). We argue that the examples of emergency cases that might lead one to accept principles of assistance that are more stringent than Moderate have intuitive force because they possess idiosyncratic features. Peter Unger invented one such case:

Bob’s Bugatti: Bob, who has most of his retirement savings invested in a very expensive Italian sports car, is confronted with the choice of throwing a switch in order to save a child by redirecting a railway trolley, which will result in the destruction of his car, or doing nothing, so that his car remains in mint condition, resulting in the child’s death. (Unger 1996, 136).

It seems that, in this case, Bob is morally required to sacrifice his Bugatti. Moreover, it seems that, in case he refuses, we could forcibly compel him to do so without violating his rights. We drew attention to one noteworthy feature of this sort of case: the protagonist has exposed his assets to undue risk. Ordinarily, things like cars are fair game in emergency rescue cases; and the fact that someone has hinged their financial future on cars that are used in the risky business of driving does not thereby render them immune from use/destruction when this is necessary to prevent the death of an innocent child. We presented other cases – Bob’s Internet Banking, Bob and the Avalanche – to help test whether this feature of Bob’s Bugatti is relevant to our judgements about required sacrifice (Barry and Øverland, 2016, 23–27). In these cases, Bob is again presented with the choice of preserving his assets or sacrificing them to save a child. But in neither case has he exposed his assets to undue risk, and thus, by contrast with the Bugatti case, it does not seem that Bob is morally required to make such a large sacrifice, though it may of course be morally best for him to do so. Moreover, it seems that it would be impermissible to compel him to make the sacrifice without violating his rights. That is, compelling him in these cases might be permissible given a child’s life is at stake, but our justification would be of a lesser-evil sort – we would infringe his rights and surely owe him compensation. Thus, we think such cases lend support to the claim that we should not generalize about the stringency of responsibilities to assist from the Bugatti case to other rescue cases.

Haydar and Lippert-Rasmussen challenge our analysis. Haydar, for example, presents a case (Ali’s Car), in which a factor that we claim is relevant to our judgements in Bob’s Bugatti, is absent (Haydar 2019, 5). Ali’s assets are tied up in his car, which can now be sacrificed to save a child who will otherwise die. But, unlike Bob, Ali is acting prudently – his car holds all his earthly belongings because he is trying to escape a life-threatening situation in Aleppo and this affords him the best chance of doing so (Haydar 2019, 5). For this case to present a challenge to our analysis, we must have the firm intuition that Ali is required to sacrifice all his earthly belongings to save the child. I think, however, that this case really is quite different from Bob’s Bugatti. This can perhaps best be seen by considering it from the point of view of enforcement. Suppose Ali refuses to make the sacrifice. Could we forcibly destroy all of his assets as a means of saving the child without violating his rights? We could, I think, permissibly infringe his rights in case we were prepared to compensate him for destroying his assets after the fact – restoring nearly all of his wealth (nearly all, since like others he is required to bear moderate cost to save the child.)

As we argue in the book, judgements about all rescue cases seem also to be influenced by whether the losses are compensable and whether there is some expectation that those who make large sacrifices in such cases will, in fact, be compensated. That is, it seems as a general matter that our responses to emergency cases track fairly systematically the likelihood of being compensated: when people make large sacrifices in emergency situations, we seem much more willing to compensate them than in those cases where they make large sacrifices to address systemic and enduring problems such as global poverty.¹³ The prospect of being compensated by others for writing a big cheque to Oxfam is much less plausible than for making a sacrifice while responding to an emergency. Haydar is correct, I think, that we should offer Bob some compensation should we compel him to make the sacrifice in the Bugatti case. The difference is that Bob seems required to make the sacrifice, while Ali is not. Ali would seem justified in resisting our attempts to compel him to make the sacrifice if it were fully transparent to him that he would otherwise have to bear the full burden of the resulting loss. The same does not seem to be true of Bob.

Haydar worries that in making such judgements we are being ‘a bit too harsh on Bob’ in the Bugatti case (Haydar 2019, 8). He writes ‘it might be plausible to say that Bob deserves less compensation than Ali. It would not be reasonable, however, to leave Bob with the whole burden, while significantly compensating Ali’ (Haydar 2019, 8). I agree that it would be inappropriate for others to leave Bob with the whole burden in the Bugatti case should it be possible for them to compensate him. However, it is not obvious that they are obliged to fully compensate him, or even come close to doing so. This is not because we judge Bob harshly – he has not engaged in wrongdoing by taking on undue risk, or even exhibited a poor moral character. Our judgement is rooted in the more general principle that when you expose your assets to undue risk, you have less of a claim on others to offset your losses should this risk result in substantial loss.

Lippert-Rasmussen worries that the relevant difference is that in cases like Bob’s Bugatti, people are at risk due to freak accidents, unlike in cases like Bob’s Internet Banking. Here Bob is sitting in his remote house doing some internet banking. With the click of a mouse he can transfer his retirement savings to his neighbours to save their critically ill young child (it will be used for a helicopter service and costly operation); if he refrains, the child will die. Lippert-Rasmussen suggests that ‘Bob might think it is unfair that he should pick up the slack from the child’s parents’ (Lippert-Rasmussen, 2019, 9), and offers a new variant of the case (Bob’s Internet Banking 2) in which this feature is absent – the child’s urgent need of medical attention is caused by a spider bite.¹⁴ Hence he holds that in freak accident cases, at least, we owe more than a moderate amount to others. I think Lippert-Rasmussen is correct that whether or not others bear responsibility for a person’s ending up in a bad situation can be relevant to how cost for addressing it should be allocated. If the negligence of parents is involved in a child’s being at risk, then they should be required to compensate others when these others make very large sacrifices to rescue the child. But I am not convinced that Bob’s Internet Banking 2 is very different from our original case. Is Bob really obliged to give up his entire savings to save the boy next door who is bitten by a spider, if he knows that he will not be compensated for his loss? Could we compel him to make such a sacrifice without violating his rights? If not, then this provides a good indication that he is not required to take on such costs in the first place.

Indeed, Bob’s Internet Banking seems equivalent in relevant respects to Bob and the Avalanche, where an unexpected avalanche is certain to bury a child unless Bob acts to redirect the avalanche, which will then destroy the new house in which he has invested most of his savings (Barry and Øverland, 2016, 26). Lippert-Rasmussen himself shares our intuition that Bob need not sacrifice his home to save a child in case he knows that no such compensation is forthcoming (Lippert-Rasmussen, 2019, 12). Further, it is not clear that our judgements about the Bugatti case change all that much if we add to our description of the case that the child has wandered on to the tracks because his parents failed to take adequate precautions when they went to the market with him. It seems reasonable that in this case, the parents owe further compensation to Bob for his loss, but if Bob fails to sacrifice his car (or makes his sacrifice conditional on the parents providing compensation), he still seems to act very wrongly and we could permissibly compel him to make this sacrifice.

In our discussion of Bob and the Avalanche we wrote, ‘if we are not prepared to compensate Bob, we ought not to judge him harshly if he chooses not to redirect the avalanche’ (Barry and Øverland, 2016, 26‒27). Haydar questions what lessons we should draw from this. Our thought here was not, as Haydar suggests, that ‘those who are not willing to share my burden of assisting needy people cannot complain about my failure to take on the whole burden of such assistance’ (Haydar 2019, 7). Our aim was to question whether Bob acts wrongly in case he refuses to sacrifice his house to save the child when it is fully transparent to him that he will not be compensated for his loss. we would not judge him harshly in such a case is a very good indication that he does not act wrongly. There are, of course, cases where agents act wrongly without being blameworthy, but these are typically instances in which the agent lacks some relevant bit of information. Bob’s Avalanche is not such a case. It is precisely because Bob knows that he will not be compensated if he sacrifices his home that it is permissible for him to refuse to make such a sacrifice. And it is because we know that he will not be compensated for his loss that we ought not to judge him harshly.

Lippert-Rasmussen raises a further interesting question: should we distinguish between the cost that an agent takes on in providing assistance to those in need and the difficulty of doing so? Indeed, he proposes a different principle of assistance, one that focuses on difficulty instead of cost.

The Moderate Appeal to Difficulty Assistance Principle: If we can prevent something (very) bad from happening and it is only moderately psychologically difficult for us to do so, then we ought [morally] to do it (unless the costs for us are comparable to the badness of the event we prevent). (Lippert-Rasmussen, 2019, 5)

Lippert-Rasmussen’s principle can be both too demanding or not demanding enough and unfair. Whether we find some sacrifice difficult or not depends very much on our dispositions. So on this principle (and any that appeals to difficulty rather than some objective measure of cost), a person’s concrete responsibilities to assist depend on whether they have a generous or ungenerous disposition. A Scrooge-like person – selfish and wealthy – might find it excruciatingly difficult to take on a modest cost to help another in severe need. He might, on the difficulty account, satisfy his responsibilities to assist without giving much, if anything, to help those in need. A benevolent person, on the other hand, might only satisfy their responsibilities to assist if they sacrificed nearly all of their wealth. These implications seem implausible. Imagine that a benevolent person has sacrificed a great deal of their wealth to help others in severe need and refuses to give more. She acknowledges that it would not be very difficult for her to give more, but decides instead to make some improvements to her modest dwelling. It seems permissible for her to make this choice. Scrooge, on the other hand, acknowledges that making a sacrifice to help a child in danger would cost him only muddy trousers, but notes that he would find that very difficult to bear. Still, it seems impermissible for him to allow the child to drown. Moderate can support both of these judgements, whereas principles like Moderate Appeal to Difficulty cannot (Barry and Lawford-Smith, 2019).

Metrics and methodology

Laura Valentini argues that although Moderate is intuitively attractive, it cannot be vindicated satisfactorily by appeal to the sorts of cases we consider (Valentini 2019, 3). This is because, Valentini claims, cases alone do not allow us to establish: (i) how we should measure the cost that people should be required to bear and (ii) whether responsibilities to assist in one-on-one rescue scenarios are a matter of enforceable right (Valentini 2019, 4–5).

Valentini is right that we had little to say in our book about how costs should be understood in interpreting Moderate. Let us call this the cost metric problem. Insofar as we have not offered a solution to this problem, which has tended to be ignored in the literature on assistance-based responsibilities in general, I think it is fair to say that this principle is underspecified. However, it is not clear how to seek an appropriate solution to this problem other than by appealing to intuition. Suppose, for example, that someone suggests that the cost in question is entirely subjective: an agent can appeal to the fact that it is more than moderately unpleasant for them to make a financial contribution to help others in severe need. How might we evaluate such a proposal? Presumably, we will do so in terms of whether its implications seem plausible to us. In discussing Lippert-Rasmussen’s proposal of a difficulty measure above, I have already suggested that such an approach to the cost metric has implausible implications – it requires too little of the non-benevolently disposed and too much of the benevolently disposed. It is these considerations, drawn from strong intuitive responses to simple cases, that militate towards a more objective metric of costs.

When we appeal to intuitive judgements about particular cases, however, we do not just ‘report’ them and regard them as fixed points. This seems to be the view that Valentini attributes to us. Indeed, we note that many judgements may be in tension with each other, others in outright contradiction. We, therefore, need to reflect upon these judgements and determine whether they hang together well. This reflective exercise, however, will also necessarily have recourse to other intuitive judgements, including ones about more general principles. So, while it is true that in moral theorizing we can, as Valentini points out, sometimes ‘bite the bullet’, any theory that involves too much bullet biting will not be regarded as credible. Indeed, the general strategy that we take in moral theory is not to simply bite the bullet and move on, but to explain why the intuition against the theory is unreliable, or not as damaging as it first appears – or that rival theories involve still more problematic bullet biting!

Of course, judgements about general principles can also be intuitive (see e.g. Kamm 1998). And it is difficult to see what they could rest on, other than on our spontaneous convictions that they are justified. Philosophers tender arguments and try to explain why some principles should be preferred over others. But in doing this, it seems impossible not to rely upon appeal to intuitive judgements of a general sort – happiness is good, it is unfair to prevent capitalistic acts between consenting adults, saving two is better than saving one, you have duties to your children that you do not have to the children of other people, and so on. This is not to say that such judgements cannot be challenged (see e.g. Taurek 1977). But for challenges to particular judgements to be cogent, they will need to appeal to still other intuitive judgements that are regarded as reliable.

Valentini argues that reliance on intuition does not provide the resources to determine whether responsibilities to assist are what she calls duties of justice – in which case they are a matter of enforceable right – or, instead, duties of humanity, in which case they are not (Valentini 2019, 6). But how then should we decide whether responsibilities to assist in one-on-one rescue scenarios are a matter of enforceable right? Simply stating that such responsibilities are responsibilities of justice on the one hand, or humanity on the other, surely does not settle the issue. Indeed, how to make that sort of division is precisely what is at issue. Nor is it clear to me what ‘broader theoretical considerations’ would resolve the issue. Rather, we need to consider whether forcing people to assist others in various cases would seem to violate or infringe their rights. Suppose that Scrooge comes upon a child in need. The situation is fully transparent to him: he realizes he can save her at the cost of getting mud on his boots; but he is simply unwilling to muddy his boots to do so. Could we impose the cost of muddy boots on him without infringing his rights if this were necessary for us to save the child? It seems, intuitively, that we could and that he would have no grounds for complaint against us. The reason is that he was required to take on that cost in the first place. Consideration of such cases suggest, then, that such responsibilities are duties of justice (in Valentini’s sense).

To contest the proposition that responsibilities to rescue are enforceable (and thus duties of justice), we should consider whether there are cases where we have a genuine responsibility to rescue, but where it seems impermissible to compel us to rescue in case we are unwilling to do so. Despite her reservations about our methodology, Valentini herself provides what she considers to be such a case (Valentini 2019, 9). In the case she envisions (Sam and the exam), Tony is in a position to help a relative stranger (Sam) pass an exam. It will not be very costly for Tony to do so (it will take an hour, but he does not have any plans that would be interrupted during that time), and if he does not, Sam will fall into a severe depression which will destroy his chances of finding a decent job. Valentini claims that while such a person may have a duty to help Sam, it is not enforceable. Note first that this case possesses features that distinguish it from those we discuss in the book. In the cases we discuss, the hardships that result from failure to act cannot be attributed to the conduct of the person who is harmed. Typically, we think that people should be held responsible for rebounding from things like failed exams and should take care not to put themselves in situations where falling short on a test will make such a decisive difference to the rest of their lives. That is not the case when what is at stake is a child drowning. The cost of an hour of our lives is something we can appeal to excuse us from helping virtual strangers to pass exams, because dealing with exam results is the sort of thing we hold individuals responsible for. We do not hold small children responsible for being unable to swim.

We can see the relevance of this feature of Valentini’s original case by considering a modified version of it. Suppose that an evil scientist has arranged that, if Sam fails to pass the exam, a trapdoor will open up beneath him and he will be dropped into a pool with an alligator, which will bite off his leg. In this case, it seems plausible that Tony can be required to take on cost to help avert this harm, provided he really does have nothing better to do. Our hypothesis in the book is that there will be a tight connection between the cost we think a person is morally required to take on to avert a harm and the cost that could be permissibly imposed upon them if this is necessary to do so. We do not hold, however, that we can impose cost on people to compel them to always do what is morally best.

Acknowledgments

Special thanks to Siba Harm, Bashshar Haydar and R. J. Leland for organizing this symposium, and to Anne Gelling and Katie Steele for helpful suggestions on an earlier version of this essay. I gratefully acknowledge the financial support of the Australian Research Council and the Research Council of Norway.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ Non-comparative theories of harm are developed in Harman (2009) and Shiffrin (1999).

² For useful discussions of late pre-emption, see Moore (2010, 421‒424), and Schaffer (2000).

³ An exception is cases where the agents who do or enable harm are overdeterminers of harm – we discuss them at length below in response to Anne Polkamp’s contribution to this symposium.

⁴ For an exploration of this aspect of exploitation, see Snyder (2008).

⁵ Thanks to Seth Lazar for a discussion of these issues. See also Chisholm (1963), and Comesaña (2015).

⁶ This is, we emphasize, quite different from the typical case in which we are uncertain whether or not we will make a difference.

⁷ The same is true, intuitively, when it comes to the force that can permissibly be used against innocent contributors and innocent bystanders, which Burri and Christie do not discuss. Intuitively, you can impose more cost on a person who innocently poses a threat to you to avoid it, than you can impose on an innocent bystander to avoid the same threat.

⁸ Woollard draws in her comment on her important book-length treatment of this issue (Woollard 2015). Unfortunately, our book was already in press when Woollard’s was published. As she points out, our proposals share some interesting affinities since we both take some inspiration from Philippa Foot’s work, even though they diverge in the ways described here. Here I shall refer to her more extended treatment of these issues in the book as well as in her critical commentary.

⁹ Although she regards enabling as a kind of allowing, she thinks that enabling possesses features – the agent who enables harm becomes relevant to harm through their action, rather than their inaction – that make enablings morally distinct from other allowings.

¹⁰ Whether you regard Bob as acting wrongly in removing the car when it is his likely depends on whether you think the cost of a lost car is sufficient to justify his not refraining from driving it away.

¹¹ This is not true in overdetermination cases, which (as Woollard notes) we treat separately.

¹² Sources of special responsibilities to those at risk could be based on various things, such as occupying a special role (lifeguard, childminder), sharing some familial or other associative connection to them, having contributed to their plight, or having failed to prevent them from being exposed to this risk at some earlier time when you could have done so at little cost.

¹³ This point is actually developed in some detail by Haydar in a paper co-authored with Gerhard Øverland (Haydar and Øverland 2018).

¹⁴ Note that, if correct, the distinction between freak accidents and emergencies due to more systemic problems would not vindicate arguments by Singer and Unger that we are ordinarily required to make very large sacrifices to prevent very bad things from happening to total strangers.