Accounts of Last Resort and Their Problems

Before exploring several prominent accounts of last resort, I make two general points. First, last resort is premised on a moral distinction between violence and nonviolence. It represents the intuition that nonviolence is morally preferable to violence. As Thomas Hurka writes in explaining a standard view of last resort, “if the just causes can be achieved by less violent means such as diplomacy, fighting is wrong.”Footnote ¹ Likely, one reason last resort endures is that many make the mistake of concluding that because violence is generally morally worse than nonviolence, it is always so. This raises the question of what qualifies as violence. There are at least three prominent definitions of violence. Some argue that it is the direct use of physical force against the body of another person, whereas others think that it is any violation of human rights.Footnote ² A third category, which Johan Galtung defends, is known as “structural violence.”Footnote ³ He defines structural violence as the “cause of the difference between the potential and the actual, between what could have been and what is,” in terms of people's ability to realize their goals.Footnote ⁴ As an example, he argues that there is (structural) violence if someone dies of tuberculosis as a result of not being able to afford readily available treatment.Footnote ⁵

I use the first, most restrictive, definition of violence for several reasons. First, it adheres closer to our intuitions and commonsense uses of the term than the idea that all rights violations or all suboptimal social outcomes are violent. A second reason I favor the first definition over the second is that it permits describing an act as violence even while allowing for theories of liability to defensive harm because the target of the harm has forfeited his right not to be attacked. Consider the following hypothetical situation. A police officer shoots and kills a hostage taker, which is necessary and proportionate in order to save the two innocent people the hostage taker is lethally threatening. Proponents of the second definition of violence would not call the policeman's action violent because the hostage taker has forfeited his right against attack. However, I think a better way to explain this situation is to say that the policeman's shooting is violent (a description), but in this case it is not a rights violation and likely justifiable, all things considered, because it was necessary and proportionate to save the lives of innocent persons (a moral assessment). The first definition of violence is preferable to the second because to say that no one has committed violence if someone is shot dead stretches the definition of violence beyond recognition. Importantly, however, the first account still allows for theories of liability to defensive harm, which are central to determining which harms are permissible in and outside of war.

A third reason I use the most restrictive definition of violence is that scholars and practitioners sometimes separate military action from other forms of coercive action. For instance, the International Commission on Intervention and State Sovereignty (ICISS) notes that “coercive measures short of military intervention ought first to be examined, including in particular various types of political, economic, and military sanctions.”Footnote ⁶ The definition of violence as direct physical harm is central to just war theory and is a foundation of determining what counts as war. Nonetheless, my argument holds even if one prefers either of the other two definitions of violence. If one believes that all rights violations qualify as violence, or if one views all structural harms (of innocents) as violence, my argument would still remain intact because one could frame it in terms of minimizing rights violations or minimizing such structural harms.

A diverse group of just war theory scholars and politicians views last resort as a standard jus ad bellum precept. Michael Walzer writes that “one always wants to see diplomacy tried before the resort to war, so that we are sure that war is the last resort.”Footnote ⁷ He claims that “it is obvious, for example, that measures short of war are preferable to war itself whenever they hold out the hope of similar or nearly similar effectiveness.”Footnote ⁸ Cécile Fabre argues that a key jus ad bellum precept is that war “must be the option of last resort.”Footnote ⁹ Brian Orend argues that “one wants to make sure something as momentous and serious as war is declared only when it seems the last practical and reasonable shot at effectively resisting agression.”Footnote ¹⁰ In a discussion of standard just war theory precepts, Hurka writes that “war must be a last resort” in order for it to be just.Footnote ¹¹ Jeff McMahan states that “the fourth principle of jus ad bellum is last resort.”Footnote ¹² James Pattison, in a discussion of the “categorical approach” to just war theory precepts (which he does not countenance), writes that one jus ad bellum precept is that “if intervention is not the last resort, then it should not occur.”Footnote ¹³ U.S. politicians, including those vying for the presidency, have likewise referenced last resort in debates about under what conditions a war is just.Footnote ¹⁴ In the 2015 State of the Union address, President Obama said that “the American people expect us to only go to war as a last resort, and I intend to stay true to that wisdom.”Footnote ¹⁵ It is clear that last resort is widely accepted. But what exactly does it mean?

Four Accounts of Last Resort

The first account, which I call strict last resort, is the simplest. It requires all peaceful options to be tried before war is permissible. As James Pattison puts it, “last resort is often interpreted literally so that every option short of the use of force must be attempted” before the resort to violence is permissible.Footnote ¹⁶ Helen Frowe argues that a standard view of last resort is that “a war can be just only when all other means of averting a threat or seeking redress have been exhausted.”Footnote ¹⁷ There are two main features of this account. One is the premise that nonviolent options are always preferable to violent ones. The underlying reason why this first account of last resort is attractive is simply because war foreseeably causes tragic harms to innocents and their property. If the same just end can be achieved in ways that avoid these harms, without high costs to the party striving for the just ends, they must be taken instead of war. The second feature is that strict last resort requires an actor to attempt all possible alternatives before violence is permissible. On this account, an actor must try diplomacy, sanctions, threats, and anything else one could imagine before the use of force would be permissible. For instance, attempting to reason with the offending party but not imposing any more coercive measures, such as sanctions, would not satisfy this criterion.

Despite the initial plausibility of this account, upon closer scrutiny there are definitive objections to it. One major problem with strict last resort, as Michael Walzer argues, is that “taken literally . . . [this view of] ‘last resort’ would make war morally impossible. For we can never reach lastness, or we can never know that we have reached it. There is always something else to do.”Footnote ¹⁸ Because actors could always attempt additional options or allow more time for existing efforts to achieve a just aim, a strict interpretation of this view would require pacifism. But pacifism, by definition, prohibits violence ever being permissible, and thus it cannot be the best option because it would require nonviolent options that in some circumstances would likely produce a greater number of severe harms to innocents than would war. One should take seriously a variety of nonviolent options, including, in extreme circumstances, acquiescing to demands. But just war theory can already account for this because only some wars would likely meet all just war theory precepts even when the just cause is met. To circumvent one prong of this objection, one could reply that an actor could impose all nonviolent measures in rapid succession. For instance, it is theoretically possible within just a few hours to impose progressively more coercive measures, such as diplomacy, sanctions against key individuals, and then comprehensive sanctions. Yet this cannot be what such an account would require, because most policy options take some time to work. A more plausible account of last resort would therefore require giving policy options a reasonable time to take effect (as I discuss below).

A related problem with strict last resort is that many options short of war would in most circumstances have little to no chance of success in any reasonable amount of time. For instance, UN, bilateral, or multilateral sanctions against Syria's President Bashar al-Assad and his regime were unlikely to succeed in ending the Syria civil war and Assad's mass atrocities by themselves, at least not in the short or medium term. Yet strict last resort required imposing such sanctions before a resort to force were permissible, thus losing precious time during which lives might have been saved by other means. More broadly, if all options short of war are very unlikely to have any reasonable chance of success, and going to war sooner would likely save a greater number of innocents' lives, waiting to try options other than war is wrong. These objections greatly weaken the first account of last resort.

A second account—what I call necessity last resort—holds that last resort means that war is permissible only if it is necessary to achieve a just cause. Hurka argues that one view of “the last resort condition is really an ad bellum necessity condition.”Footnote ¹⁹ McMahan writes that last resort “means that war must, in the circumstances, be necessary for the achievement of the just cause.”Footnote ²⁰ (McMahan supplements this view, as I discuss below.)

This second account of last resort seems plausible as well. Why would one enact a violent policy if it is unnecessary to achieve a just aim? This account holds that if a just cause can be achieved by means other than war, war is impermissible. The benefit of this account compared with the first is that it would not require actors to attempt policy options that have no chance of success. In a hypothetical situation where there were only two policy options, such as war and sanctions, the first account would require attempting the sanctions whereas the second would not.

But necessity last resort suffers from problems similar to strict last resort. This second account does not explicitly require actors to try nonviolent options only if they have a reasonable chance of success; it requires that they have only some chance of success. One way to interpret this account is that nonviolent options must have a zero percent chance of success, and war must have some probability of success greater than zero. War must be the only possible way to achieve a just cause. A problem with this view is that epistemic limitations would make it very difficult to know when both nonviolent options and war would meet these demanding conditions. Even if one interpreted this account in a way that brings it closer to the third account (which I discuss next) in that it would prohibit all nonviolent options that did not have a reasonable chance of success, there is an additional, deeper problem. If violent and nonviolent options have the same chance of success, and even if both were certain to achieve a just cause, this account would require nonviolent options even if they killed far more innocent people than violent means because war would not be necessary to achieve the just aim. These are decisive objections against this second account.

The third account of last resort includes a reasonable chance of success component. For this reason, I call it reasonable chance last resort. This account requires that all peaceful options that have a reasonable chance of success be given adequate time to achieve their just aim before the use of force is permissible.Footnote ²¹ A. J. Coates is one proponent of such a view. He writes that “what the principle [of last resort] enjoins is the exhaustion of effective alternatives to war. The obligation to employ sanctions or other non-military methods is conditional upon their efficacy.”Footnote ²² McMahan concurs. He supplements his necessity requirement with a reasonable chance of success clause. He argues that last resort holds that “it would be wrong to go to war if there were an equally effective but peaceful means of achieving the just cause.”Footnote ²³ Larry May also believes that a reasonable chance of success is a necessary condition of last resort. He argues that “last resort does not dictate that we must always choose the least violent means, but only the least violent and equally efficacious means, time permitting.”Footnote ²⁴

There are two reasons why reasonable chance last resort is better than the previous two accounts. First, internal to this account is the reasonable chance of success criterion for nonviolent options, which avoids Walzer's objection that last resort might require pacifism because we could never reach “lastness.” Second, those options with a reasonable chance of success only need to be given a reasonable amount of time to achieve their just end. Resort to the use of force is only prohibited during the time period that is reasonable for nonviolent options to work. Thus, this third account is clearly more sensible, convincing, and sophisticated than the earlier ones.

Despite the third account's improvements, however, there is a decisive objection against it. The central problem with reasonable chance last resort is that it does not stipulate that harms to innocents must be minimized. Nonviolent options may be disproportionate, indiscriminate, and unnecessary, but even this third account of last resort would require them to be attempted before a resort to force is permissible, assuming that they have a reasonable chance of success. This most plausible account of last resort, therefore, is still morally indefensible because what should matter is how many innocents are harmed and how severe those harms are, not whether harms are inflicted violently or nonviolently.

Another way even reasonable chance last resort might fail in protecting the greatest number of innocents is by requiring the postponement of war in order to give nonviolent means with a reasonable chance of success a reasonable amount of time to work. Imagine that one determines that a reasonable chance of success is 51 percent, and in one situation sanctions meet this requirement. Imagine a war is 95 percent likely to achieve a just cause in this situation, and would impose the same risks on innocents as the nonviolent policy. Suppose that sanctions, which are deemed to have a reasonable chance of success (60 percent), are tried and fail. Suppose furthermore that the countries that imposed the sanctions could have instead resorted to war, and that the war would have had a significantly higher chance of success (90 percent) and would have harmed the same number of innocents as the sanctions. Imagine also that during the time the sanctions were imposed the leader of the targeted sanctions realizes that a war might be coming and therefore increases his state's military capacity so much that a later just war would certainly result in a greater number of harms to innocents than if the sanctioning states had fought a just war earlier. In this case, an earlier war would have been preferable to sanctions because it had a higher chance of success and it would have harmed the identical number of innocents. Furthermore, imposing sanctions may have made a later war impermissible because the chances of success could have fallen to below a reasonable threshold.Footnote ²⁵ To illustrate this point, Coates cites Churchill, who argued that had those states that opposed Hitler risked earlier a “little war, a great war might have been avoided.”Footnote ²⁶ Last resort would be problematic in this case because it would require the nonviolent option to be attempted and exhausted, such as Neville Chamberlain's accommodation with Hitler.

The fourth account is what I call reasonable last resort. John Lango requires last resort to have “reasonableness standards.”Footnote ²⁷ William O'Brien writes that “every reasonable peaceful alternative should be exhausted.”Footnote ²⁸ A significant benefit of this account is that it attempts to resolve a potential problem I noted above that James Pattison identifies, namely, that nonviolent alternatives “might cause more harm than military intervention.”Footnote ²⁹ In one passage, Hurka also frames last resort in a similar way. He writes that “a last resort condition forbids war if its benefits, though significant, could have been achieved by less destructive means such as diplomacy.”Footnote ³⁰ Simon Caney attempts to defend this account. He writes that “last resort rests on the moral assumption that agents may resort to a course of action only having considered less awful options first (where ‘awfulness’ is measured in terms of number of rights violations and the nature of the rights).”Footnote ³¹ Lango argues that a reasonable chance of success, proportionality, and awfulness standards should be internal to the precept.Footnote ³² He argues that last resort should incorporate an “impracticality” or reasonable chance of success requirementFootnote ³³ and “a standard of comparative awfulness.”Footnote ³⁴

The main advantages of this fourth account compared with the previous ones are the addition of the proportionality and awfulness standards that Lango and Caney argue should be internal to last resort. Lango suggests that “an alternative measure [to war] does not have to be attempted first if there is no reasonable expectation that it will be less harmful.”Footnote ³⁵ In sum, Lango writes that “to satisfy that [last resort] principle, we have to prove with clear and convincing evidence that each sufficiently detailed planned course of nonmilitary actions either would not achieve the goal or would be disproportionate or would be substantially more awful.”Footnote ³⁶ This emphasis on harms is a real improvement on the earlier accounts.

The central problem with reasonable last resort is that it is redundant with other just war theory precepts, and thus unnecessary. It is standard to include reasonable chance of success and proportionality as independent jus ad bellum principles. There is no standard awfulness requirement, but proportionality can and should cover this criterion, as I discuss below. As I will argue, the best way to account for some of the insights of the scholars who attempt to defend reasonable last resort is to drop last resort as a just war theory requirement.

Why All Accounts of Last Resort Are Problematic

To better understand why the first three accounts of last resort are problematic, consider the following hypothetical example. Imagine that country X has a just cause for war against country Y. Suppose that there are only two policy options that can achieve the just cause, both of which have an equal probability of success and both have an equal risk of resulting in unintended harms to innocents. Assume, too, that in both cases all other just war theory precepts are met, and that there are no partially liable individuals. The first policy option is war: If soldiers from country X violently attack country Y, 50,000 people will be killed, of whom 25,000 will be innocents. The second policy option is economic sanctions: If leaders of country X impose sanctions on country Y, 100,000 people will be killed, of whom 50,000 will be innocents.

Strangely, adherents to the first three accounts of last resort would require the policy option that would foreseeably and avoidably kill double the number of innocent (and liable) individuals than is necessary to achieve a just cause. This cannot be correct. Intuitively, killing far more innocent people than is necessary in order to achieve a just cause is wrong. Given these two options, only the first—violent—option is permissible.

Proponents of the fourth account fare better. They would prohibit the nonviolent account and permit the violent option because it is less awful. But this is already required by at least one prominent account of the jus ad bellum proportionality precept. Thomas Hurka writes that proportionality can “incorporate the other just war conditions about consequences.”Footnote ³⁷ He argues that there are two aspects of a proportionality criterion: the first requires the relevant expected benefits to outweigh the expected costs; and, if several options meet this first requirement, the second requires choosing the option that will likely have the lowest costs, holding other things constant.Footnote ³⁸ This second component of his proportionality account is important because it prohibits unnecessary (morally weighted) harm. In a discussion of Hurka, Lango acknowledges that proportionality can account for his version of last resort. As he writes, “admittedly, a just war theory might accept a proportionality principle of such complexity.”Footnote ³⁹

Hurka is right that proportionality can “incorporate last-resort considerations.”Footnote ⁴⁰ And Pattison is partially right that his effectiveness account of armed humanitarian “intervention can incorporate some of the importance of . . . last resort.”Footnote ⁴¹ I think “effectiveness” defined as the protection of human rights covers entirely the morally important aspect of the most plausible account of last resort. But this too is already covered by Hurka's account of proportionality. The basic reason reasonable last resort is redundant with this aspect of Hurka's proportionality account is that both require identical policy options for the same moral reason, namely, limiting unnecessary harm. The question then becomes whether we should frame such a principle in terms of proportionality or last resort.

There are several reasons to think it is better to privilege proportionality. First, because there are multiple accounts of last resort that yield not only controversial policy recommendations but in fact require contradictory policies, to include last resort as a necessary precept would leave a great deal of room for mistakes and disagreement. Second, it is not clear how reasonable last resort coheres with typical definitions of “last.” Reasonable last resort does not require war to be last in a series of options, unlike the other three accounts. I do not want to get distracted by semantic debates, but it seems the fourth account is quite some distance from typical understandings of “last.”

Third, an additional benefit of including a comparative harm calculation in proportionality rather than last resort is that it is standard for proportionality to account for the moral status of individuals. By this I mean whether one is innocent or liable to defensive harm to some degree—what Seth Lazar calls “morally weighted harms.”Footnote ⁴² Morally weighted harms discount the moral importance attached to harms inflicted against individuals in proportion to the degree they are liable to defensive harm. As Jeff McMahan argues, proportionality does not require that one must choose the policy that harms the fewest people overall. Rather, McMahan argues that proportionality requires choosing the option that “achieves a net saving of the lives of those who are fully innocent.”Footnote ⁴³ For instance, imagine two policy options, X and Y. Option X would likely inflict a greater number of harms than option Y, but X would harm fewer innocents than Y. Policy X is the better and only permissible option because a fewer number of innocent people are likely to be harmed. That individual moral status matters in moral assessment is captured by jus ad bellum (and jus in bello) proportionality.

Fourth, proportionality better captures our intuitions about why the violent option that harms 50 percent fewer innocent (and noninnocent) individuals is permissible while the other one is not. Proportionality concerns the relevant goods and evils a policy is likely to cause, unlike last resort, which is commonly taken to address whether a policy is conducted militarily or nonmilitarily. Finally, if one adopts this aspect of Hurka's account of proportionality, there is no additional benefit of reasonable last resort. The policy implications are identical if one uses morally weighted harms in reasonable last resort and superior if one uses nonmorally weighted harms.

Now we may return to the hypothetical sanctions example. It is, in fact, grounded in reality. Nonviolent policy options that many politicians and scholars advance as preferable to violent policies can harm as severely and affect as many people as violent ones. Consider the sanctions imposed against Iraq in the 1990s. Joy Gordon, an expert on the Iraq sanctions, writes that “the fundamental goal of sanctions, after all, was containment. The inspections did effectively disarm Iraq; the sustained collapse of the Iraqi economy did prevent Iraq from rebuilding its military capacity.”Footnote ⁴⁴ However, they also may have taken a great human toll. Assessing a variety of studies, Gordon estimates that UN sanctions likely killed 200,000 to 500,000 Iraqis, many of whom were children.Footnote ⁴⁵ I want to emphasize that my point of the potential harm to innocents caused by the sanctions against Iraq is not that war against Iraq would have been preferable; both the sanctions and war could have been disproportionate (or could have failed to meet other just war theory precepts). Rather, my point is that in some circumstances sanctions can cause even greater harm to innocents than would war.

Scholars have conducted systematic studies and have found support for the claim that economic sanctions can harm large numbers of innocents. Specifically, they have found that economic sanctions can worsen respect for the human rights of women,Footnote ⁴⁶ decrease the prospects of democracyFootnote ⁴⁷ (which in turn results in more human rights abuses on average),Footnote ⁴⁸ harm large numbers of people by decreasing health outcomes,Footnote ⁴⁹ and indirectly increase violations of individuals' physical integrity rights.Footnote ⁵⁰ The question is not only whether sanctions harm innocents but whether they may harm a greater number of innocents than violent alternatives that might be able to achieve an identical just cause. Susan Allen and David Lektzian address this question in their study of the effects of sanctions and war on health. They find that “major sanctions and major military conflicts are both seen to significantly decrease HALE (Health Adjusted Life Expectancy), with sanctions having more than twice as large an effect.”Footnote ⁵¹

One might then suggest that sanctions are different in kind from other types of measures because they can be coercive. Sanctions can be imposed under Chapter VII of the UN Charter, whereas less coercive means, such as negotiations, can also be taken under Chapter VI. Noncoercive, peaceful measures can share the same problems as sanctions, however. Consider noncoercive negotiations, which are of course generally preferable to war or any other violent policy. For the sake of simplicity, again assume there are just two options, negotiations or war. Even if negotiations have not been given a reasonable amount of time to succeed and have a reasonable chance of success, if continuing with negotiations would likely result in a greater number of harms to innocents than resorting to war, war is preferable (assuming all other just war theory precepts are met).

Let us consider the Rwandan genocide as an example. Imagine that shortly after the genocide began the best estimates were that negotiations would have a reasonable chance of ending the slaughter but that they would take two weeks. In Rwanda, killers murdered about 800,000 people, almost all of whom were innocents, in approximately 100 days.Footnote ⁵² This means that during the two weeks of negotiations, over 100,000 innocent people likely would have been killed. Now assume that armed intervention would have likely killed half as many people during those negotiations, and imagine also that fewer than half of those killed by armed intervention would have been innocents (that is, more than half would have been belligerents fully liable to defensive harm). As with the sanctions example, the first three accounts of last resort would have prohibited taking military action, whereas the fourth would have permitted the military option only. But like the sanctions example, there is no advantage to the fourth account of last resort in this scenario if one adopts Hurka's view of what proportionality requires regarding choosing the policy option that will likely harm the fewest innocents independent of whether the policy is violent or nonviolent.

Or consider Syria. From July 2012 through the spring of 2013 approximately 5,000 people, including many innocents, were being killed each month on average.Footnote ⁵³ The international community decided against using overt armed humanitarian intervention and instead opted to push for negotiations in combination with sanctions as an attempt to end the civil war and mass atrocities. The second round of peace talks took place in early 2014, and also failed. Was this the right choice? It may have been preferable to intervene militarily instead of continuing with negotiations if intervention could have plausibly resulted in the deaths and severe harms of fewer innocents, or it may have been preferable to use armed intervention in combination with negotiations for the same reason. Approximately 6,000 people were killed each month on average in 2014, again including many innocents.Footnote ⁵⁴ This is not to say that armed humanitarian intervention should have occurred definitively. Rather, my claim is that even if negotiations had not been exhausted—even if they had a reasonable chance to succeed—that is not a good reason in itself to refrain from armed humanitarian intervention. In this case, intervention would not have satisfied several accounts of last resort, but it may have been the better, and indeed the only, morally permissible policy.