A cornerstone of traditional just war theory is that combatants on both sides of any conflict are legitimate targets of attack. This is not because they are engaged in wrongdoing but because they pose a threat to others. So long as they follow the jus in bello rules of proportionality and discrimination, those who fight for an unjust cause and those who fight for a just cause are moral equals: both may permissibly kill and both may permissibly be killed. This is the doctrine of the ‘moral equality of combatants’ (MEC) and it has long been the standard view. A revisionist challenge, however, argues that one is not liable to be killed if one fights for a just cause. People have a presumptive right against being killed, after all, so why should a soldier fighting for a just cause lose that right for posing a justified threat against an unjust enemy? Jeff McMahan makes the case for this revisionist view in Killing in War. He explicitly rejects the MEC and instead endorses an asymmetric view of the morality of combatants. Critics object that this revision makes just war impossible because without the MEC it is impossible to justify our convictions about legitimate and illegitimate targets. In this article, I defend the revisionist view against this objection. My defence, in turn, leads me to propose a new and more adequate account of enemy status.
1. The Dilemma
McMahan rejects the MEC for a variety of reasons. His central argument is that a soldier fighting for a just cause (hereafter, ‘just soldier’) has done nothing to make himself liable to be killed.1 Much of Killing in War (KIW) is spent showing how traditional defences of MEC fail to overcome this simple point. If just soldiers haven't lost their right to not be killed, then enemy soldiers violate that right if they kill them. In that case, soldiers fighting for an unjust cause (hereafter, ‘unjust soldiers’) cannot actually discriminate at all, for if it is wrong to kill just soldiers, unjust soldiers have no legitimate targets whatsoever. So it's impossible for unjust soldiers to follow the jus in bello requirement of discrimination. Hence, a soldier's side must have just cause for her to be capable of acting justly in war. The presumption of moral symmetry between soldiers is thus abandoned.
That a just soldier has done nothing to surrender her right to not be killed derives from McMahan's view of liability. One becomes liable to be killed only if there is some wrong for which one is responsible. One loses the right to not be killed and becomes liable to defensive killing, for example, when one is sufficiently morally responsible for a threat of serious and objectively unjust harm to another person. But liability comes in degrees. One is more or less liable in accordance with however more or less responsible one is for the wrong in question. Accordingly, any response to a wrongful threat must be proportionate to the degree of the wrongdoer's responsibility (what McMahan calls ‘narrow proportionality’).
The worry is that McMahan's revisionist account is incompatible with just war. The objection centres on the difficulty of determining liability in war. If liability is not based on posing a threat but is instead based on responsibility for a wrongful threat, then the determination of such responsibility could be problematic. At one extreme, nearly all on the unjust side (including civilians) might be liable to attack. At the other extreme, virtually no one might be liable (including enemy soldiers). Either possibility fails to deliver just war.
A few have mentioned this problem for the rejection of the MEC, but Seth Lazar recently developed a powerful version of the objection in great detail (Lazar 2010). Lazar concurs with McMahan's rejection of the MEC, but argues that this rejection coupled with McMahan's model for liability results in an internal problem – namely, that the ability of just soldiers to discriminate between legitimate and illegitimate targets is lost. In individual self-defence cases we assign liability to those responsible for a threat of wrongful harm. Yet assigning responsibility for an unjust cause to actors in war will be more difficult. The MEC resolves this difficulty: all soldiers on both sides are liable to be killed. But without the MEC how can soldiers know whom it is permissible to kill? Lazar contends that McMahan has no means to block the conclusion that either too many people are legitimate targets for killing (total war) or that almost no one is a legitimate target (contingent pacifism). This is Lazar's ‘Responsibility Dilemma’ (hereafter, the ‘Dilemma’) for McMahan's account and, by extension, against all accounts that reject the MEC.
McMahan tries to avoid the contingent pacifism horn – that just soldiers cannot permissibly kill unjust soldiers – by arguing that the vast majority of unjust soldiers are, indeed, liable to be killed in war. He argues against various mitigating considerations that can be offered for unjust soldiers’ liability and concludes that these considerations, on the whole, fail as full excuses – and that, in any case, excuse is compatible with responsibility and hence with liability. Thus, just soldiers are right to take unjust soldiers as liable to be killed. But if that's the case, Lazar argues, this far-reaching liability should also extend to many who are traditionally considered non-combatants. That is, many civilians on the side of an unjust cause will be causally related to the unjust cause to a similar degree as many of the soldiers by activities such as campaigning and voting for pro-war candidates, paying taxes, and in general creating a favourable political climate for war. And if the excuses protecting the soldiers from liability fail, then so too will such excuses fail to shield the non-combatants from liability; and we've opened the door to total war. McMahan is aware of this problem and tries to block it by arguing that non-combatants escape liability because most of them are insufficiently causally and morally responsible for the threats the war poses. But in that case, Lazar replies, this argument should succeed for many of the unjust soldiers as well. If this is right, then just soldiers cannot rightly attack anyone and victory is impossible; and we're back to pacifism. In short: if McMahan makes unjust soldiers liable, then so too will many unjust non-combatants be liable; hence, total war. But if McMahan argues that unjust non-combatants are insufficiently liable to be killed, then that makes many unjust soldiers insufficiently liable as well; hence, contingent pacifism. Lazar argues the Dilemma has made the just war tightrope too slender: the rejection of the MEC causes just war theory to lose its balance and fall to one extreme or the other.
2. The Contingent Pacifism Horn
The Dilemma is based in part on the claim that unjust combatants are not liable. It is also based on the related but distinct claim that just combatants cannot have sufficient knowledge of unjust combatants’ liability to make it permissible to kill them. Much of Lazar's work focuses on this later claim. Indeed, Lazar insists this epistemic bar for just soldiers is so high that it is virtually unattainable. As he writes, for just soldiers to know their adversaries are liable to be killed ‘they must know at least their adversaries’ personal histories, the context of their decision to fight, their connection to a particular threat, their capacity for responsible agency, their beliefs and intentions, and that their own cause is just’ (Lazar 2010: 187). If this is right, then just combatants cannot attain permissibility to kill unjust combatants and, thus, contingent pacifism wins the day. But such claims of epistemic uncertainty are too strong. A great deal of uncertainty is involved in determining liability to be killed in war, but it is not inscrutable; at least not in all cases.
Responding to this kind of epistemic uncertainty raises difficult issues. If the epistemic bar is as high as Lazar demands, it creates a standard that most would find implausible for other cases of liability attribution, such as personal self-defence. Presumably Lazar accepts that it is possible to have sufficient evidence to permissibly kill an attacker in personal self-defence. It is worth exploring, then, what knowledge is required for legitimate liability attribution in both self-defence cases and war. Imagine walking peacefully down the street when a stranger charges you, attempting to stab you with a large knife. It is certainly possible there are relevant facts inaccessible to you. You would, indeed, need to know the long list of attributes Lazar gives above to know with certainty that the stranger is liable to defensive harm. But this mere possibility of mistake does not negate the evidence you have for the stranger's liability and thus does not negate the permissibility of attacking the stranger in self-defence. To have sufficient evidence that the attacker is liable to make it permissible to use defensive force you do not need to know the attacker's entire personal history, the context of his decision to attack you, all his beliefs and intentions, and so forth.
The following knowledge could be relevant to the permissibility of defensive force in ascertaining the attacker's liability: Have you done anything to make yourself liable to his attack (e.g. are you unjustly threatening him or some other)? Does the attacker have some other justification for attacking you (e.g. does it appear that by killing you he is preventing the deaths of innocent bystanders)? Is the attacker a morally responsible agent (e.g. can you see he is under some kind of mind control)? Most think there is sufficient evidence for each of these questions in what I've already said: a stranger charges you while trying to stab you and there is no reason to believe that he is morally justified in attacking you, that he is not responsible for his action, etc. Most believe that you are justified in presuming the attacker to be liable and, thus, in defending yourself. If killing the stranger were both necessary for averting the unjust threat and proportional to the potential harm, then it would be permissible to do so.
For each of the kinds of relevant knowledge just discussed, the epistemic difference between personal self-defence and war is a matter of degree, not kind. All that needs to be added to the list for the war case is knowledge that one's cause is just. Lazar rightly shows it is difficult to be justifiably confident that one's cause in war is just; there are countless reasons why one might be mistaken, including deception by one's government. But it is possible to overcome these uncertainties and have sufficient evidence for a just cause. If so, and if one also has evidence for the liability of one's adversaries, it can be permissible to kill in war.
These are not merely abstract possibilities. There is the familiar example of the Polish soldier fighting an invading Nazi soldier. That the Pole is fighting for a just cause is the only further thing the Polish soldier would need sufficient evidence for beyond what's needed for justification in an ordinary case of self-defence. And, certainly, unprovoked aggression could count as evidence – perhaps even sufficient evidence – for that conclusion. A more recent example could be a Kuwaiti defending against the Iraqi army in 1990. Granted: for it to be permissible to kill in war one needs to determine that the particular enemy is liable because he is making a sufficient contribution to an unjust cause, that he is a responsible agent, etc. But these same epistemic hurdles exist in personal self-defence. No doubt some of these uncertainties are greater in war. In self-defence cases, for example, it will usually be easier to determine that an attacker is making a sufficient contribution to an unjust threat, in contrast to an unjust soldier carrying a gun he may never fire. But it remains true that these uncertainties still arise in self-defence cases and the difference between the cases is one of degree, not kind.
The insistence on the impossibility of ascertaining unjust soldiers’ liability creates an important dilemma for the Dilemma itself. If one maintains that it is impossible for just combatants to have necessary knowledge of unjust soldiers’ liability, then so too will it be similarly impossible for anyone to have the knowledge necessary for permissible killing in self-defence. For, with the exception of knowing they have a just cause, all the uncertainties just soldiers face determining their adversaries’ liability in war are also faced in any self-defence case. So, to maintain the Dilemma against the revisionist view, we must insist on a high level of certainty regarding an adversary's liability in war, which will also commit us to rejecting the permissibility of self-defence in a wider range of cases than most of us would be willing to do. If we cannot accept this, we must grant that the epistemic threshold can be crossed in war as well. Admittedly, there will usually be more uncertainty in war cases than in self-defence cases; but that does not defeat the possibility of successful war-time liability determination. To avoid this problem, the Dilemma's advocates might claim that self-defence justifications are never based on liability, but are instead grounded in an agent-relative permission or a lesser evil defence. But that seems a high price to pay; better to reject the Dilemma.
A further resolution to the Dilemma becomes available if we revise McMahan's views on liability. McMahan claims that a person can act permissibly according to the evidence available to them, yet, if it turns out their evidence is mistaken and the act is impermissible relative to the actual facts, then they can be responsible and, hence, liable. That is, McMahan thinks liability turns on acts that are fact-relative impermissible as opposed to merely evidence-relative impermissible.2 This distinction is based on an assumption that there is no univocal way to understand ‘wrong’ or ‘permissible’. As Derek Parfit explains:
I favour an evidence-relative (ER) view of moral wrong-doing and permissibility. McMahan explicitly rejects the view that ER permissibility excludes liability. An adequate discussion of these issues lies beyond the scope of this article. However, if we claim that ER permissible action cannot be a ground of liability to defensive action, we can better handle many of the Dilemma's worries.
Some act of ours would be wrong in the fact-relative sense just when this act would be wrong in the ordinary sense if we knew all of the morally relevant facts … and [another] wrong in the evidence-relative sense just when this act would be wrong in the ordinary sense if we believed what the available evidence gives us decisive reasons to believe, and these beliefs were true (Parfit 2011).
This is so for at least two reasons. First, under this view, just soldiers would be liable only if they killed without ER permissibility. So they would not incur liability for killing an unjust soldier they reasonably believed to be liable, even if it turned out the unjust soldier was, in fact, non-liable. Second, by investigating what evidence unjust soldiers have regarding the justice of their cause, the just side could determine what level of ER permissibility the unjust soldiers could be reasonably taken to have. From this the just side could then determine which soldiers should be viewed as liable and judiciously fight under such constraints. This evades much of the Dilemma's force by making liability determination more epistemically attainable since it turns on evidence available to an actor, not on facts they cannot access. This diminishes the force of claims about the difficulty of ascertaining liability.
Thus, one defence against the contingent pacifist horn is made by restricting the basis for liability to norms of ER permissibility. If that's right, it can be ER permissible for a just soldier to kill an adversary in war, in some cases, and thereby not make herself liable even if she lacked fact-relative permissibility. But even if one rejects this view, the broader point still stands: the epistemic difference between standard self-defence cases and war will be a matter of degree, not kind; and the uncertainties regarding liability can be overcome.
3. The Total War Horn
The other horn of the Dilemma is that if unjust soldiers are liable, then many non-combatants will also be liable. But then nearly everyone on the unjust side of a given conflict – soldiers, civilians, even children perhaps – could potentially become legitimate targets. Lazar writes, ‘if small, unnecessary contributions, some of which one makes only by being in a particular space, are sufficient for liability to be killed, then many more non-combatants than is plausible will be pulled into the liability net’ (Lazar 2010: 192). McMahan responds by appealing to narrow proportionality. It demands we attack people only if attacking them is necessary to thwart a wrong and we restrict such attacks to the minimal force needed to prevent the wrong. So most non-combatants, because they do not contribute to an unjust cause in the right way or to a sufficient extent, will not be liable to be killed. But the Dilemma's trap is set: for if McMahan uses narrow proportionality to protect civilians from attack in this way, then it should also apply to many combatants and we are back to contingent pacifism.
My defence against the total war horn rests on the McMahanian notions that liability comes in degrees and that liability has the internal requirement of necessity for any permissible response. If that is true, then having sufficient evidence to know when this necessity clause is met will be especially difficult in the case of non-combatants. And, even if some liability did extend to those traditionally considered non-combatants, that (by itself) does not mean they are necessarily legitimate targets for lethal attack. Just forces would have to know that by attacking non-combatants they could effectively thwart the unjust cause. Further, they would have to know that only by killing them could this best be accomplished. If some alternative means could equally thwart the contribution some unjust non-combatants provide (by capturing them, say), then killing would be impermissible. Given such constraints, it's hard to imagine just forces would attain warrant to kill non-combatants very often. The degree of liability most civilians will incur for the contribution which they could be known to be responsible will usually be quite low.
This is primarily because what combatants are actually doing in war is starkly different from what civilians are doing. Combatants are actively engaging in killing other people – so the contribution to the unjust threat they pose is direct and clear. This puts them in the dock for liability in a way the vast majority of non-combatants will not be. Non-combatants’ contributions are indirect and, thus, harder to discern. Moreover, the vast majority of civilians’ actions contribute to unjust war efforts to such a watered-down extent as to be negligible. And, importantly, it is foreseeable that such actions are negligible. Most non-combatants have little reason to believe that anything they do will contribute in any way beyond a negligible one to an unjust cause. So, notice, on an ER account for liability mentioned above, the vast majority of non-combatants will not be liable. But even on McMahan's view of liability, that their contribution is foreseeably negligible will make most civilians essentially non-responsible (and, hence, non-liable) agents.
This will not always be true, however. If an unjust non-combatant does have a less-than-negligible contribution, then they could be liable to some extent and, depending on the necessity and proportionality constraints, could be legitimate targets of attack. Such cases are rare and would probably apply only to high-level decision makers, etc. Even then, it would be difficult for just forces to have the epistemic insight required to determine civilian liability with justifiable confidence. But opening the door for such a possibility does deliver total war, as the Dilemma contends; so the objection fails. The possibility will not mean that all civilians are legitimate targets. At worst, it could mean that some fraction of those previously regarded as immune from attack should, in fact, be liable to attack. And this is not entirely implausible.
So the defence against the total war horn is primarily an epistemic one. But, because of this, it could appear the response is not consistent with the previous response to the contingent pacifism horn. Recall the dialectic thus far: the Dilemma claims that by rejecting the MEC there is too much uncertainty to discriminate that unjust soldiers are liable. I respond that these epistemic limitations can be overcome – that unjust soldiers can be sufficiently responsible for an unjust cause to be liable and that this can be known, in at least some cases. But the Dilemma then presses: if that's the case, then why are civilians not similarly liable for their contribution and why cannot this be known? The reason is because epistemic limitations on ascertaining liability will usually be far greater in the case of civilians. This is because of characteristics endemic to non-combatants that make these uncertainties different in kind, not merely degree. Namely, non-combatants’ contributions are both indirect and (usually) foreseeably negligible. Thus, the necessity and proportionality conditions internal to liability will be radically more difficult to meet and radically more difficult to know that they've been met. As a result, justification for total war does not obtain.
4. A Supposed Paradox
Lazar further claims that a paradox develops out of the epistemic uncertainties involved in wartime discrimination, as follows:
It is impossible to discriminate between liable and non-liable combatants, because of the lack of information endemic to warfare.
Many combatants are not liable to be killed, because they are not sufficiently responsible for the threats that they pose.
Anybody who chooses to kill [combatants], knowing both 1 and 2, chooses to kill indiscriminately.
Anybody who chooses to kill [combatants] indiscriminately is maximally morally responsible for the threats that he poses, and so liable to be killed.
It is therefore easy to discriminate between liable and non-liable combatants (Lazar 2010: 197).
The defence I've outlined has a simple response: Lazar's mistake is in conflating the difficult with the impossible. By claiming it is impossible to discriminate liability such absurdity easily results. But if it is difficult but not impossible to discriminate, the paradox falls apart. For, in that case, it's a big leap to label anyone who kills in war as necessarily acting indiscriminately. So 1 is false because it is not impossible to ascertain liability for some combatants and thus to engage in discrimination, at least in some contexts. Because 1 is false, 3 is false. Because 3 is false, 4 does not apply to just soldiers; and 5 does not follow. So there is no paradox. Instead, we know this: it is very difficult to fight justly. But it's not impossible; the just-war tightrope can be walked.
5. A New Proposal
Embracing the revisionist account would require significant changes to how war is waged and jus in bello criteria are implemented. I propose tying combatant status directly to the degreed nature of liability; to re-envision combatant-hood away from an all-or-nothing condition to one of degree. McMahan writes that ‘the extent to which a person is excused for posing a threat of wrongful harm affects the degree of his moral liability to defensive harm, which in turn affects the stringency of the proportionality restriction on defensive force’ (McMahan 2009: 156). We can take this principle and from it derive new categories for an unjust population that could be applied in war. Thus, rather than the binary combatant or non-combatant system of categorization traditionally used, a conflict-by-conflict rubric could be constructed that tracks differing levels of liability for a given set of unjust enemies. The distinctions could range from 1st, 2nd and 3rd-degree combatants and the like (or more, as needed) and similar degrees for non-combatants. The basis for an adversary's degreed status would be tied to what I call ‘reasonable perceived liability’ (RPL) rather than other metrics traditionally used (such as posing a threat). RPL is the best approximate determination of an enemy's degree of liability that could be reached by a just force taking all reasonable efforts to make that determination. Once broad levels of RPL were determined for a population, correlating categories of degreed combatant and non-combatant statuses could then be applied accordingly. Specialized rules of engagement (ROEs) could then be created which best allow a military to achieve victory yet come as close as possible to matching the correct level of response to the RPL for each category.
By deriving enemy status from RPL some traditionally labelled as combatants may be considered some lesser-degreed type of combatant with different ROEs; and vice-versa for non-combatants. Narrow proportionality can press us farther than McMahan suggests to ever increasingly complex and restrictive ROEs resulting in a range of responses for different cases, depending on the RPL of a given enemy.3Pace Lazar, under such ROEs victory could still be attainable in certain contexts.4
To be clear, an unjust enemy's status should track exactly his or her liability. Consequently, one might object that I should not generalize from distinct individual liability to categories of any kind – be it the traditional binary split or my more complex divisions. I agree in principle but believe that by developing complex levels of enemy status we could adequately cover the vast majority of cases so that just soldiers following RPL-derived ROEs could justly engage an enemy with minimized moral risk. There is a limited, finite range of possible choices on how to engage any given enemy population. Thus, we can create discrete levels of enemy status tied to RPL, even if RPL is itself non-discrete. That is, for pragmatic reasons that do not outstrip respect for the enemies’ rights, some generalizations of enemy status are permissible. The more complex divisions I'm proposing are a best approximation of what morality gives us which serve as a heuristic for the best we can reasonably do in efforts to correlate how one is treated in warfare to one's liability.5
So the formal proposal is that we have more categories with simplified differences in engagement for different groups in war. It functions as partial response to the concerns raised by the Dilemma. More substantively, a classic problem in just-war theory is how to classify people who support an unjust war effort in ways such as building tanks and munitions. Walzer and Nagel both wrestled over this question and argued it was acceptable for a just force to kill such persons while they worked but not in their homes (Nagel 1972: 140; Walzer 1977: 146). I'm not defending the position but raise it to show that even these traditional just war theorists recognized that liability in war is more complicated than the meagre two category system can deliver. By creating more divisions within these categories, with different ROEs for each, these and other perennial questions regarding difficult cases could be better resolved. What are we to make of a military lawyer who spends all her efforts fighting to mitigate the war crimes her unjust comrades commit? Should her status be ‘combatant’ and thus liable to be killed by a just force? Or consider civilian target analysts working in an unjust military's command centre. Should they be immune from harm as ‘non-combatants’? Such cases suggest that more complex categories reflecting differing levels of liability are needed.
I cannot here examine all the objections that could be raised against my proposal, but will briefly note two. Some will argue that such a scheme makes fighting war impossible by being too restrictive on the just war-fighter; that it would ‘tie their hands’ to such an extent that they could never have military success. Granted, what I am proposing places high burdens on just soldiers – calling on them to use near Herculean-like self-control under extreme circumstances. But, I contend, this is precisely what it means to be a just soldier. And these ever increasing demands placed on just soldiers are what we should expect as our moral reasoning over war improves. Consider the dramatic changes regarding how just soldiers are expected to fight over the past quarter-century alone. Just war doctrine has become increasingly ingrained in Western militaries. We should expect new developments in just war theory will elicit further developments for jus in bello practices; perhaps like those suggested here. Far more is expected of present just war-fighters than was ever expected of putatively just war-fighters a generation previous. When Walzer challenged the realpolitik orthodoxy in 1977 with Just and Unjust Wars, few thought that ethical constraint in war was likely to ever gain consensus much less actionable policy changes. But the fruit of just war theory's restraining influence is now evident in contemporary war-fighting by Western nations.
Moreover, the kind of change I'm proposing may already be occurring. Take NATO's present counter-insurgency (COIN) operations in Afghanistan. NATO forces engage the last vestiges of al-Qaeda quite differently from the way in which they engage Taliban fighters. The former they attack ‘with prejudice’ and, where possible, kill via drone airstrike. The latter they engage more cautiously, avoid high death tolls, and use extreme restraint to avoid non-combatant causalities. The point is that they are treating distinct groups of enemy ‘combatants’ as different kinds of combatants with different ROEs. Admittedly, this distinction is presently done on purely pragmatic grounds (in the hopes of not alienating certain groups), but such distinctions between combatants could be done on the grounds of perceived differences in liability.
Further, recent developments in military technology strive to be ever more discriminate in war, not less. This is contra Lazar who believes modern warfare makes indiscriminate combat more likely due to long-range artillery and ‘dumb bomb’ munitions. But such forms of warfare are becoming increasingly passé. Military weapons of the future enable more just behaviour, not less, because they are far more accurate. Many of these kinds of weapons are already being employed.6
Will expanding the categories for enemy status make battlefield decisions too difficult when soldiers confront military units consisting of individuals from different categories? All just war theorists must answer this question, for even traditional theories distinguish combatants from non-combatants and military units sometimes contain individuals from each of these two categories. I grant that the question is more vexing, however, for the view I have proposed. Two responses are in order. First, like traditional just war theorists, I think an appeal to the doctrine of double effect is helpful here. Soldiers should not intend to harm their enemies to a degree that is inconsistent with their liability. But some such harms, if unintended, may be justifiable. Second, the whole point of expanding the categories of enemy status is to respect the rights of individuals, including those engaged in unjust wars. Doing so will increase the difficulties confronted on the battlefield, but these costs need to be paid if one is to take seriously the rights of an enemy population.
The rejection of the MEC is a needed corrective to just war theory. And, despite fears to the contrary, the tightrope of just war can still be traversed without its aid. Yet to do so we must sharpen our understanding of what justice demands of us in war and we must develop a new set of categories for enemy status, categories that adequately reflect the epistemic and moral factors that McMahan's work has done so much to illuminate.7