Compensation and Necessity in War

3.1. Infringing and Compensating as Complex Acts

Consider the following case.

No Compensation

A culpably wrongful Aggressor will shoot and kill Defender unless she throws a bomb at Aggressor, which she consequently does at T1. Doing so foreseeably kills Aggressor, eliminating the threat he poses. However, next to Aggressor is Bystander–an uninvolved innocent whose arm foreseeably breaks in the ensuing explosion. At T2, Defender violates her compensatory duty toward Bystander by gratuitously refusing to compensate him.

The deontic status of what Defender did at T1 remains indeterminate until T2. Her refusal at T2 to compensate resolves the deontic status of her earlier action at T1. To be clear, the claim is not that her refusal at T2 retroactively determines the deontic status of what she did at T1. Rather, it is only from T2 onward that what she did at T1 has the deontic status that it does. More specifically, Defender’s refusal to compensate at T2 means that her actions at T1 now violate the necessity constraint. To see why a wrongful failure to compensate affects necessity, we need only compare the actions available to Defender:

• a. attack-and-then-fulfill-compensatory-duties

• b. attack-and-then-violate-compensatory-duties

Option ‘a’ results in less harm to Bystander than option ‘b’. Given the availability of option ‘a’, option ‘b’ violates necessity.

One might point out the following. In assessing necessity, harms to Bystander aren’t the only harms that count. The harms to Defender matter as well. (If not, then necessity would prohibit Defender from throwing the bomb regardless of whether she compensates Bystander. After all, refraining from defending herself is the most effective way of preventing Bystander from suffering harm).²⁷ Though options ‘a’ and ‘b’ both avert the harm that Aggressor would otherwise inflict on Defender, option ‘a’ results in some harm to Defender: the deprivations that compensation causes her. But because Defender owes compensation, that harm is ‘moralized’ in that we partially discount that harm in the necessity calculus. The result is that opting for ‘b’ violates necessity even when including the harmful deprivations that compensation causes Defender.

The claim that the deontic status of what we do at one point in time can depend on what we do later on is an utterly banal feature of the moral landscape. There are many acts which can be properly evaluated only in tandem with other, subsequent acts. Consider this mundane case: it is permissible for me to eat a meal at a restaurant (at T1) only if I subsequently pay for it (at T2). The deontic status of the decision to eat there depends on whether I pay. The two events morally interdepend.

The point might be put like this: if I dine-and-dash, what I have done wrongly is not dining by itself (it is permissible to dine so long as I pay) nor is it dashing by itself (it is permissible to dash if I did not dine), but rather the two in combination. The same goes when Defender compensates at T2 for a rights-infringement she committed at T1. As a result, we might not know until T2 whether what she did at T1 is indeed part of a deontically permissible whole.

Still, it might be alleged that I have engaged in a kind of mereological gerrymandering in my description of the actions available to Defender. According to this criticism, I have inaptly conjoined two distinct actions–one at T1 and one at T2–into a single, diachronically-described act. This move might be thought inapt for the following reason. When we descriptively conjoin multiple acts only one of which is wrongful, any moral evaluation of the conjoined act will be procrustean. Suppose I go to the gym, eat dinner, and then drive home intoxicated. We might ask whether what I did is permissible. More specifically: is it permissible to exercise-then-eat-then-drive-intoxicated? Presumably doing that is wrongful. But this negative verdict on the deontic status of a complex act obscures the moral status of its constituent parts. Clearly, the exercising and eating are permissible; drunk driving is not.

It might, then, be alleged that I have committed the same legerdemain in my descriptions of the actions available to Defender. In each of ‘a’ and ‘b’, I referred to a pair of distinct actions by describing them as a diachronically conjoined whole. This speciously suggests that what Defender does ex post can affect the deontic status of what she does ex ante. It might seem, then, that Defender’s refusal to compensate does not mean that her earlier actions now violate necessity. Rather, Defender’s decision to inflict defensive violence was permissible, and the subsequent decision to refrain from compensating Bystander was impermissible. Or so it might seem.

But I think the mereologically conjoined, diachronically described action-combinations in ‘a’ and ‘b’ are indeed morally apt. The short explanation is this: what Defender does at T1 and what she does at T2 morally interdepend in a way warranting evaluating these discrete acts together.

To see why, take the following candidate evaluation of Defender’s conduct: what she did at T1 was permissible and what she did at T2 was impermissible. This evaluation, though, is arbitrary. Consider an equally acceptable alternative evaluation: what she did at T1 was impermissible and what she did at T2 was permissible. After all, if Defender had opted against self-defense and thereby refrained from infringing Bystander’s rights, there would have been nothing wrong with subsequently refraining from compensating Bystander. In evaluating Defender’s conduct, we can either i) “hold fixed” the decision to inflict defensive violence and then evaluate her decision to refrain from compensating, or ii) “hold fixed” the decision to refrain from compensating and then evaluate her decision to inflict defensive violence. There is nothing privileging the first option over the second.

We might think that we should hold fixed the decision to inflict defensive harm rather than the decision to refrain from compensation, because inflicting defensive harm is justified. But that’s precisely what’s at issue: whether inflicting the harm is justified absent compensation. A non-arbitrary evaluation of Defender’s conduct would look like this: she ought to have attacked only if she was going to compensate. Or put differently, a particular combination of acts is verboten: Defender ought to either attack-and-then-compensate or not attack at all. But note that the first option is just the mereologically conjoined, diachronically described action-combination mentioned in ‘a’ above.

The moral is that the complex description is apt after all, because the deontic status of the constituent acts morally interdepend. Defender’s choice–to attack without compensating–is a more harmful option relative to the alternative of attacking with compensating. This means the first option violates the necessity constraint. Since it violates the necessity constraint, that complex action is unjust.

Whether Defender violates necessity depends on whether she compensates Bystander for infringing his rights. How, then, do we evaluate what Defender has done, after she infringes his rights but before she can compensate him? During that interim, given an indeterminate future, there might be no fact of the matter as to whether Defender violated necessity. Until the compensatory window closes, the deontic status of what Defender has done might remain indeterminate. This result is not infelicitous, though. This is because, again, Defender’s actions are in an important sense incomplete until such time as she decides whether to compensate. Before then, the deontic status of what she has done is indeterminate for the rather unremarkable reason that incomplete acts can have indeterminate moral status.

To briefly recap: in No Compensation, we cannot just say that the defensive attack was permissible and the failure to compensate was impermissible. That evaluation is arbitrary given the following equally acceptable alternative evaluation: the defensive attack was impermissible and the decision not to compensate was permissible. A non-arbitrary evaluation of her conduct would proscribe a particular combination of acts: infringing-without-compensating.

This analysis might seem relevant to the decades-long debate between actualists and possibilists. Put roughly, actualism is the view that the moral status of some act ϕ depends on how an agent would in fact act were she to perform ϕ. Possibilism is, put roughly, the view that the deontic status of an act ϕ depends on how an agent could act were she to perform ϕ.²⁸

So, according to actualists, whether it is permissible for Defender to infringe at T1 depends on whether she will in fact compensate at T2. According to possibilists, whether Defender is permitted to infringe at T1 depends on whether she could subsequently choose to compensate at T2. Here is where the views come apart in practice: if at T1 it is a foregone conclusion that Defender will choose not to compensate at T2, then actualism forbids infringement whereas possibilism allows it.

But contrary to appearances, the debate between actualism and possibilism is largely orthogonal to the view I’m defending. Actualists and possibilists reach different verdicts about whether Defender’s conduct at T1 is wrong, depending on how Defender would (according to actualists) or could (according to possibilists) subsequently act at T2. But on the view I defend, a particular combination of acts is wrong: infringing-without-compensating. When that combination occurs, it’s not specifically the infringing that’s wrong. Nor is it specifically the compensatory failure that’s wrong. Rather, the duty Defender violates is disjunctive: a duty to either a) refrain from infringing or b) compensate for the infringement. Because Defender could satisfy the duty either by performing ‘a’ or ‘b’, a failure to perform just one of the two options does not foreclose the possibility of satisfying the duty.

To see this, suppose Defender fails in her disjunctive duty by performing neither ‘a’ nor ‘b’. Even then, we cannot isolate either the infringing or the failure to compensate as the specific wrong.. This is because Defender could have satisfied her duty in a way consistent either with infringing or with refraining from compensating. If she infringes, the subsequent failure to compensate affects what she has done in this sense: the decision to infringe-without-compensating is wrongful.

How do we translate this into a claim about Defender’s moral liability? I turn to this issue next.

3.2. Preventing Compensatory Violations

I have argued that infringing-without-compensating violates necessity, given the alternative of infringing-and-compensating. That is, Defender has a disjunctive duty: to either compensate or to refrain from infringing. What are the implications for Defender’s moral liability? Generally, an individual who will otherwise violate a duty is morally liable to be treated in adverse ways that will prevent that violation. This is provided that the adverse treatment in question is a) the least harmful means available of preventing the violation, and b) is not too severe when weighed against the importance of preventing the violation. These just describe, in rudimentary form, the constraints of necessity and proportionality.

Because Defender’s duty is disjunctive, the force to which she is liable is disjunctive as well. That is, she can be liable to force necessary to compel her either to compensate or to refrain from infringing. But when does Defender violate necessity? Does it occur at T1, or at T2, or at some other point? What is proscribed is a combination of two acts; does this mean both acts need to occur for the violation to occur and for Defender to be liable for that violation?

Not necessarily. To stymie a wrong, we don’t always need to wait until it commences. There is a rich debate in war ethics and the ethics of defensive violence as to whether and when a party–who will commit a wrongful harm at some point in the future–is liable to preventive force now, given that the party has not yet commenced with that wrongful harm.²⁹ I presume that this would-be wrongdoer can indeed be liable to preventive force if waiting makes it impossible or significantly more difficult to stop the wrong, and if the would-be wrongdoer has made it clear through his own responsible agency that he will so act. (I take these to be jointly sufficient rather than jointly necessary conditions for liability to preventive force). The upshot is that though Defender commits a wrong only once she infringes and refrains from compensating, she can be liable to preventive force if there is no way to compel compensation from her afterward, and if she has made it clear through her own responsible agency that she will violate her compensatory obligations.

One might point out, though, that by forcibly stopping Defender at T1, we prevent her not only from violating her compensatory duties at T2, but also from defending herself against Aggressor. Isn’t this something Defender has a right to do? Why believe that Bystander’s compensatory rights are more important than Defender’s right to self-defense? Recall, though, that the deontic status of what Defender does at T1 will ultimately be determined later on, by whether she fulfills her compensatory duty at T2. Defender has a right to self-defense only if it satisfies necessity–and it satisfies necessity only if Defender fulfills her compensatory duties afterward. The moral here is this: what licenses preventive force is not that compensation is more important than self-defense, but rather that Defender does not have a right to defend herself in a way that will ultimately violate necessity. By violating necessity, Defender impermissibly harms an innocent, which can license defensive force.³⁰ But this view faces a challenge, which the following exemplifies.

Unnecessary Bomb

Aggressor will culpably inflict five units of harm on Defender unless Aggressor is stopped. Defender could forcefully stop Aggressor at no harm to anyone else, by using a taser. But she instead recklessly opts to stop Aggressor by throwing a bomb, which will inflict one unit of harm on Bystander as a foreseeable side-effect.

Because Defender violates necessity, she has made herself liable to defensive force. Since the basis for Defender’s liability is agent-neutral, Defender is morally liable to be forcibly stopped by just about anyone.³¹ But what about Aggressor? Is Aggressor now permitted to counter-defend himself against Defender, on the grounds that successful counter-defense against Defender will prevent her from inflicting unnecessary harm on Bystander?

If so, this would be bad news for my account. But Defender is not liable in this way after all. If Aggressor defensively attacks Defender to prevent her from inflicting unnecessary harm on Bystander, then Aggressor would himself be violating the necessity constraint. This is because there is, by hypothesis, a less harmful way for Aggressor to prevent injury to Bystander: by ceasing his own unjust attack against Defender. By doing so, Defender would have no reason to defend herself against Aggressor, and no reason to act in a way that collaterally harms Bystander. It turns out, then, that Defender is liable to be attacked, but not by Aggressor–which is a welcome result.

The force to which Defender is liable must not only satisfy necessity but proportionality as well. If Bystander stymies Defender, Aggressor will succeed in inflicting five units of harm on Defender. We need to determine whether this satisfies the narrow proportionality constraint.³² This is a constraint on the severity of the harms we can inflict on liable individuals. The harm we inflict cannot be too great relative to the good we achieve by preventing the harm for which she is liable. Suppose a culpable aggressor is intent on maliciously pinching you; the only way to stop him is to kill him. Doing so would be impermissible in that the aggressor is not liable to be killed since it would violate narrow proportionality.

Many factors might enter into the narrow proportionality calculation, such as the doctrine of double effect, the doctrine of doing and allowing, the culpability of the liable party, and so on. To keep things simple, assume narrow proportionality permits inflicting on a culpable aggressor a harm no more than five times the severity of the harm that the aggressor’s attack will otherwise cause. It turns out, then, that stopping Defender satisfies narrow proportionality. (Of course, adjusting the stringency of the proportionality constraint results in different permissions. If narrow proportionality permits inflicting on Defender no more than twice the harm that she would otherwise culpably inflict, then stymieing Defender would be impermissible).

A further reason for thinking that stymieing Defender satisfies narrow proportionality is that in doing so Bystander does not intend the harm that Aggressor will consequently inflict. Even if stymieing Defender means that Bystander must intentionally harm her, the harm that Aggressor consequently inflicts on Defender is harm that Bystander merely foresees rather than intends. Though this by itself does not permit us to “write off” those harms, the moral relevance of intentions and the doctrine of double effect mean that, arguably, those harms count as unintended rather than intended in Bystander’s proportionality calculus.

There is another, closely related reason for thinking that stymieing Defender satisfies narrow proportionality. The harm to Defender comes not from Bystander, but from Aggressor. Because this harm is agentially mediated, it arguably ought to receive discounted weight in Bystander’s proportionality calculus. Indeed, this suggests Bystander can stop Defender even if doing so means Bystander has to harm Defender, in addition to the harms that Aggressor will inflict. But for simplicity’s sake, I set aside the relevance of intervening agency.

The upshot is that narrow proportionality permits Bystander to act in a way that results in harms to Defender that she would otherwise be permitted to defend against if she acted within the bounds of necessity. By needlessly harming Bystander, Defender might morally open herself up to the harms she was trying to prevent.

Violating necessity, then, can serve as a basis for moral liability to defensive force. Recall that an individual can make herself liable to preventive force for violations that have not yet occurred, if she has made it clear through her own responsible agency that she will commit the violation, and if waiting makes it impossible or significantly more difficult to stop the wrong. Recall also that infringing-without-compensating violates necessity, given the alternative of infringing-and-compensating. This entails the following. An individual who defends herself against a wrongful aggressor by infringing a bystander’s rights, makes herself liable to necessary and proportionate preventive force if there is no way to compel compensation from her afterward, and if she has made it clear through her own responsible agency that she will otherwise violate those compensatory obligations.

Thus, in cases like “No Compensation”, Defender can be liable to preventive force at T1 even though what she is doing at that point won’t qualify as a violation until T2. Whether an individual acting in defense will compensate those whose rights she infringes, is of moment not only after the defensive engagement, but during it. The prospect of compensatory failure later on can permit others to stop her now. I now turn to the practical consequences of what I have argued so far.

4.1. Who Exactly Owes Compensation?

In discussing who owes compensation, I focus on state and non-state actors, as opposed to individual combatants.³⁷ So, suppose a country we’ll call ‘Defender’ is waging an otherwise just defensive war against Aggressor, a paramilitary non-state actor operating out of a foreign country, Bystander. Aggressor, once defeated, cannot compensate the victims in Bystander for the otherwise justified collateral harms that Defender inflicts. Thus, the compensatory duty falls to Defender. Or so I have argued.³⁸

But what about countries uninvolved in the conflict? Do they have compensatory duties? In discussing jus post bellum, Alex Bellamy argues that we should not focus solely on belligerents.³⁹ This is because, often enough, peace settlements are supervised by international institutions and enforced by multinational peacekeeping forces which might not include belligerents in the conflict. Similarly, James Pattison argues that those best capable of assisting with reconstructive efforts should take on that burden, regardless of whether they were party to the conflict in the first place.⁴⁰

We should indeed encourage multilateral, multinational efforts at reconstruction following war. But Defender, by shifting harms from itself to Bystander, incurs a stronger and more stringent duty than others do. But suppose that Defender, by warring against Aggressor, shifts harms from other countries onto Bystander as well. Consider a case where Defender is the sole party fighting a justified defensive war against Aggressor which is attacking multiple countries. In this sort of case, there are other countries who are left better off than they otherwise would be, but who do not participate in the fighting. Do the compensatory duties fall solely on Defender in such cases?

Several authors argue that those who owe compensation for permissible rights-infringements include those who are left better off as a result of the rights-infringement than they otherwise would be.⁴¹ They would have to pitch in, post bellum, perhaps in accordance with a) the severity of the harms shifted from them onto Bystander, and b) their ability to pay. But there is an important exception to this rule. As noted earlier, a party cannot be asked to pay for the costs of helping them avoid harms when we imposed such help upon them involuntarily. Suppose that the party explicitly requested of Defender that they refrain from fighting. They might not want to feature in the justification of the harm that the war causes.⁴² This presumably relieves them of any enforceable compensatory duty toward the civilians whose rights Defender permissibly infringes.

So, for example, suppose in 2005 Coalition forces attack Al Qaeda terrorists in Iraq; the attack maims civilians nearby. Suppose further that these terrorists would otherwise have attacked either New York City, or London, or Paris. Where they would have attacked matters to compensation, since the US and the UK were Coalition members whereas France was not; preventing an attack upon civilians from Coalition-countries would qualify as self-defense, whereas preventing an attack on non-coalition countries would qualify as other-defense.

Given this uncertainty, does this mean France owes some probability-weighted percentage of full compensation? No; again, those who did not ask for or want assistance typically cannot be made to pay for it. The same goes for strikes that prevent attacks against civilians in the region. Those saved by such interventions might owe compensation to those whose rights were collaterally infringed in the course of that intervention, but only provided that they asked for or wanted such assistance and that they are able to compensate.

4.2. Do ‘Condolence Payments’ Erode Civilian Immunity?

One might worry that requiring compensatory payments will erode respect for civilian immunity. Governments might adopt more permissive attitudes toward collateral harm on the grounds that they can just “pay off” the civilians (or their families) afterward.

As a moral matter the argument I have presented does not open up civilians to more attacks. It wouldn’t permit doing what is impermissible under the current norms of warfare. In fact, the argument introduces a restriction rather than a permission: it will often be the case that we cannot permissibly infringe the rights of civilians in war without compensating them afterward (provided we can compensate them).

That being said, the worry, I take it, is this: once we normalize compensatory payments, such a policy will make it easier to spuriously rationalize collateral harm to civilians, on the grounds that whatever harms we inflict will at least be partially compensated later on. This more indulgent attitude toward collateral harm, it might be argued, will make things worse relative to the status quo, even when we factor in the good that the resulting compensation will do.

But I suspect the opposite is true. If anything, a policy of compensating collaterally harmed civilians will further disincentivize such harms. This is because such a policy will, in effect, attach a monetary cost to such harm. For each civilian immiserated or maimed, for each whose family member is killed, we will have to render compensation in return (provided that the unjust aggressor cannot be made to pay). Indeed, this is probably the most effective, feasible way to discourage collateral harms in war since such expenses are unlikely to be viewed upon kindly by administrative and bureaucratic officials who must financially account for them.

So, far from encouraging a more lenient attitude toward collateral harms to civilians, the recommended policy will likely do the opposite by encouraging caution and forbearance. This is badly needed in a culture where foreign civilians are often seen as expendable. Though battling such a culture isn’t the policy’s purpose, it is nonetheless a salutary side-effect–one which speaks further in favor of adopting a policy of compensating victims of rights-infringements in otherwise just self-defensive wars, lest we violate the in bello necessity constraint.

A related worry is that requiring compensation–combined with liability for compensatory violations–increases what each side believes to be justified in furtherance of winning the war. The thought might go like this. Inflicting rights-infringing harms on civilians in bello will be justified only if we compensate post bellum. But we will be able to compensate our victims only if we win, which makes it that much more important we win. Otherwise, we violate in bello necessity; mutatis mutandis for the other side. The result is that wars become bloodier–or so it might be argued.⁴³

But this worry is misplaced. It is strange to think that the victorious side will refuse compensation from the vanquished side. Perhaps the worry is instead that the vanquished side will be unable to compensate. But in that case, they have no duty to compensate, given that ‘ought’ implies ‘can’. So, either losing a war is compatible with satisfying our compensatory duties, or it cancels those duties. Either way, our compensatory duties will not yield perverse incentives in favor of prolonging war, even if satisfying those duties is a requisite of satisfying in bello necessity.

4.3. Do We Compensate Individuals or Nations?

If what I have argued is correct, compensation for collateral harms is owed to innocent victims and their families. But what if Defender cannot locate them after the war? Or what if gaining access to them is prohibitively difficult? Given the difficulties of discharging individually owed compensatory duties, maybe it is better to treat the people as a whole–the country–as the party owed compensation. On this view, Defender discharges their compensatory duties by funding schools, hospitals, roads, and infrastructure in general.

Funds aimed at reconstruction might indeed be required where we degrade or destroy civilian infrastructure. But by funding such projects we do not thereby fulfill the compensatory duties we have to the individual civilians we have maimed in the course of the war. They remain wronged even if we confer benefits on their co-nationals.

Case in point: the Department of Defense’s ex gratia payouts are only a small fraction of the payments the US government provided to civilians harmed during the wars in Iraq and Afghanistan. This is because the US distributes most of its assistance through the United States Agency for International Development (USAID) rather than the Department of Defense. In fiscal year 2019, Congress allocated via USAID $10 million for Afghanistan and $7.5 million for Iraq, which is about 65 times and 30 times what the Defense Department distributed that year in ex gratia payments to those countries, respectively. Unlike the payments issued by the Defense Department, USAID payments do not have the purpose of facilitating the cooperation of local populations, and thus benefit a broader swath of civilians. These funds, for the most part, are directed not to particular individuals who have suffered harm during the war, but instead facilitate projects to assist Afghan and Iraqi civilians who suffered physical, economic, and psychological harm during the war.

The USAID distributions are a modest but laudable humanitarian effort. Still, they fail to satisfy the directed duties the US and its allies have toward the victims of collaterally inflicted harm. We do not fulfill our compensatory duty toward someone who has lost a house or a child by funding a hospital in their community, no matter how badly it’s needed. Of course, such community-oriented projects might confer benefits that ultimately redound to the advantage of those owed compensation. But it is only to this extent that such projects satisfy our compensatory duty to that victim in particular.

What of the logistical difficulties involved when it comes to locating the individuals we have harmed? If we are in operational control of the relevant territory post bellum, tracking down victims and their families ought not to be prohibitively difficult. Even absent operational control, the US and its allies have over the past few decades developed the institutions, the intelligence apparatus, and the technology necessary to track down individual high-valued terrorist targets striving to remain hidden. Given these abilities, it is disingenuous to claim that we just can’t locate civilian victims and their families post bellum when they want to be found.

Finding victims of war is presumably easy. But identifying victims of Coalition strikes–distinguishing them from victims of violence committed by terror groups–is much more difficult. Nonetheless, investigative journalists have shown remarkable success at identifying specific victims of specific Coalition strikes, despite the limited resources at their disposal.⁴⁴ State-funded institutions should be able to achieve this sort of success at a much greater scale.

But even once the relevant victims are found and identified, challenges remain. Providing these victims with cash will do little good–and indeed, might do harm–if criminals, bandits, and corrupt government officials subsequently rob them. Ideally, Coalition forces would provide the requisite security apparatus, but history has shown this to be prohibitively difficult. Under these circumstances, compensation might instead take the form of voluntary relocation to more hospitable environments where personalized social services, as well as other forms of compensation (including but not limited to cash payments), can be safely made.