Welcome to what we expect will be a very interesting and productive discussion of Patrick Tomlin‘s “Proportionality in War: Revising Revisionism.” The paper is published in the most recent issue of Ethics, and is available here. Gerald Lang kindly agreed to contribute a critical précis, and it appears immediately below. Please join in the discussion!
Gerald Lang writes:
Patrick Tomlin’s “Proportionality in War: Revising Revisionism” (Ethics, 2020) provides subtle and significant challenges to some revisionist axioms in just war theory. Everyone working in these complex debates will want to read and study this paper. I have learned a great deal from it. However, I do have doubts about some of the moves Patrick makes, and this seems like the place to rehearse them.
I start with some stage-setting and basic exposition. Just war theory has two central dimensions: “jus ad bellum”, which is concerned with the morality of going to war, and “jus in bello”, which is concerned with the morality of fighting in war. In traditional or orthodox just war theory, the evaluations of these dimensions are entirely independent from each other: if ad bellum has already condemned country A for going to war against country B, but A goes to war anyway against B, it is still an open question whether A’s combatants fight permissibly or impermissibly at the level of in bello. We will therefore have to budget for the possibility that A’s combatants, who ought not to be fighting in the first place, might nonetheless be fighting permissibly. How can that be? Haven’t A’s combatants already lost the case for participating in this war? True, if you have already cleared the ad bellum bar, there is no guarantee that you will then clear the in bello bar. But it is tempting to think that, if you have already failed the ad bellum test, there is nothing you can do to achieve success under in bello. It is too late for that.
Revisionist just war theory, spearheaded by the pioneering work of Jeff McMahan, is troubled by such discrepancies. It wishes to bring these dimensions of assessment into productive contact with each other by insisting that the same critical standards must govern both ad bellum verdicts and in bello verdicts. Moreover, these standards, for McMahan, are drawn from ordinary morality. There are no sui generis standards of moral assessment for war. Revisionism is thus usually coterminous with reductionism. Once we commensurate critical standards in this way, drawing in each case on the tenets of ordinary morality, the verdicts on in bello will not risk getting out of alignment with the verdicts of ad bellum, at least at the level of deep morality.
Patrick provides a more fine-grained analysis of these revisionist critical interventions. On his view, the traditional ad bellum proportionality exercise has a number of functions: (1) it helps to determine whether the resort to war is permissible; (2) it assesses the war as a whole; and (3) it offers a moral assessment. By contrast, traditional in bello proportionality assessment pursues the following tasks: (4) it helps to determine whether acts of war are permissible; (5) it assesses individual acts of war, rather than the war as a whole; and (6) it offers a non-moral assessment (p. 36).
Patrick thinks that, of these various claims, it is (6) in particular which the revisionist literature has targeted. On his view, revisionists wish to replace (6) with the following claim (p. 37):
Revisionist in Bello Proportionality
(6*) is a moral assessment (of the same kind as the ad bellum proportionality assessment).
Important corollaries of this revisionist programme are these (pp. 39-40):
(7) If every act of war is proportionate, then the war as a whole must be proportionate.
(8) If every act of war is disproportionate, then the war as a whole must be disproportionate.
And the more general versions of these claims—with (7) and (8) counting as more specific instances of them—are these (p. 41):
(7*) If each individual act within a harmful course of action is proportionate, then the course of action must be proportionate.
(8*) If each individual act within a harmful course of action is disproportionate, then the course of action must be disproportionate.
McMahan thinks these claims are obvious, and that their truth is more or less guaranteed when we commensurate standards for evaluating ad bellum questions and in bello questions. If there is just one single set of standards for assessing ad bellum and in bello, then the various verdicts we reach in these sub-domains cannot fail to be mutually consistent.
One of Patrick’s central aims, which takes up the bulk of his paper, is to show how each pair of claims—(7) and (7*), and (8) and (8*)—is actually false. Some of the cases he uses to argue for these conclusions may be on the outlandish side, but the results still have real news value, since most revisionists will be in little doubt that (7) and (8) are true on broadly conceptual grounds.
Patrick appeals to two cases, or two sets of cases. His Bombing Campaign case is deployed mainly against (7), though he also suggests that a variant of it can be deployed against (8). His Faulty Guns case, which exploits the non-identity problem, is deployed mainly against (8), though, once more, he thinks that a variant of it can also be deployed against (7). I will examine these cases in turn.
- Bombing Campaign
In Bombing Campaign, A has just cause, and is waging war against B (p. 41). There are 100 bombing raids, and these are disproportionate overall: more people are killed than are saved. In more detail, the relevant figures are these: each of Raids 1 to 90 saves one person, and kills ten people, while each of Raids 91 to 100 saves 90 people, and kills ten people. Overall, the bombing campaign kills 1,000 people, and saves only 990 people. Hence, it is disproportionate. For handy reference, I will call any one of the raids in the range between Raid 1 and Raid 90 a “bad” raid, and any one of the raids in the range between Raid 91 and Raid 100 a “good” raid. Bombing Campaign contains 90 bad raids and ten good raids.
Bombing Campaign is pressed into service to challenge (7). Patrick regards each of these raids—from Raid 1 to Raid 100—as proportionate, even though the campaign as a whole is disproportionate. Now why should that be? An obvious first take on this issue is that, at least in fact-relative terms, the good raids between Raids 91 to 100 have a decent chance of clearing the proportionality bar, whereas the bad raids between Raids 1 to 90 look more problematic, given their poor net returns. At the very least, Raids 91 to 100 must have a better chance of qualifying as individually proportionate than Raids 1 and 90. Now in other domains, it can be difficult to disentangle individual contribution from total production or aggregate activity, and this can cause difficulties in the determination of individual entitlement or liability. But these problems do not obviously affect Bombing Campaign, because Patrick has already specified the contribution of each bombing raid in numerically precise terms. Bomber 45, for example, killed ten people, and saved only one person. Delete Raid 45 from the sequence, and there would be ten fewer killings and one saved person less. We leave everything else in place. Isn’t that all we need to know? Apparently not.
Patrick suggests instead that, even in fact-relative terms, each of these raids is proportionate, though the course of action that they together constitute is disproportionate. The crucial issue here is how to interpret the counterfactual involved in the proportionality assessment of any one of these raids. Patrick reminds us that “to assess the proportionality of some harmful act, the relevant comparison is with a world in which the act does not occur” (p. 42). However, the world with which Bomber 45’s raid should be compared is not the world in which none of these raids takes place. The counterfactual measure must include acts performed by others, and the only plausible counterfactual is the completely filled-in world in which the other 99 raids are carried out. So a better way of determining whether Raid 45 is proportionate or disproportionate, drawing on the correct counterfactual, goes like this:
When we remove one act from a sequence of acts like this, no matter which act we remove, we remove the good done by the final act. This is because if Bomber 45 had not flown, Bomber 46 would simply have become Bomber 45, Bomber 47 would have become Bomber 46, and so on, until Bomber 100 became Bomber 99. So, if Bomber 45 had not flown, 90 fewer people would have been saved. (pp. 43-4)
I will call the sequence of raids including Bomber 45 the “original sequence”, and the counterfactual sequence of raids excluding Bomber 45 the “revised sequence”. In the original sequence, there are 90 bad raids and ten good raids. In the revised sequence, there are 90 bad raids and only nine good raids. It is this adjustment between the original sequence and the revised sequence, and the resulting loss of one good raid in the revised sequence, that certifies each raid in the original sequence as proportionate.
Why does Patrick assume that there is a decline in the number of good raids in the revised sequence? These are independent raids, and this fact should matter to the interpretation of the counterfactual. It is not as though Bomber 91, in the original sequence, is more successful because there are 90 bad raids behind him. There was no fixed quota of bad raids to be satisfied before we had the right to expect some good raids. An alternative (and admittedly simple-minded) strategy, as before, is to hold everything else in place when we test each raid for the difference it made to the overall outcome. We know the numbers for the other raids already. With Bomber 45 in the sequence, there are 90 bad raids, and ten good raids. But Bomber 45’s raid was a bad raid. Thus, in the revised sequence, there are 89 bad raids, and ten good raids. What is wrong with this suggestion? Why not embrace this interpretation of the counterfactual?
One possible problem is that this interpretation may not be enough to secure the proportionality of any of these raids, even Raids 91 to 100. We have the numbers, but we may not have proportionality. Perhaps, because the bombing campaign failed overall, none of these raids can count as proportionate. That is an undismissable possibility, and I will revisit it in §4. But nor do I see compelling reasons for adopting Patrick’s view that each of the raids is proportionate.
Patrick devises another case, focused on the non-identity problem, for testing (8). Before he gets to that part of his discussion, however—see §3, below—he briefly outlines another variant of Bombing Campaign that is intended to put pressure on (8). In Bombing Campaign 2, as I will refer to it, we have a series of raids—100 as before—which is proportionate as a whole, but where each raid is now supposed to be disproportionate (p. 46).
Bombing Campaign 2 is, in effect, an inverted variant of Bombing Campaign. Getting Patrick’s reasoning clear here may help to shed light on what he thought he was getting out of the original version of Bombing Campaign, as well as helping us to determine the strength of this initial objection to (8).
The Bombing Campaign 2 schedule unfold as follows. Each of Raids 1 to 90 saves 1,000 people, and kills ten people. Those impressive returns dry up in Raids 91 to 100: each of these further raids kills ten people, and saves only one person. 90,010 lives are saved overall, whereas 1,000 lives are lost. The large gap between the lives saved and the lives ended makes this bombing campaign comfortably proportionate overall. However, each of the raids belonging to Bombing Campaign 2 is disproportionate. Why so? This is Patrick has to say: “each individual bombing raid is disproportionate, since if that bomber had not flown, then the bomber would not have killed ten people, and only one fewer person would have been saved” (p. 46). Take Bomber 1, whose raid saves 1,000 people and kills ten people. If we delete this raid from the original sequence, then Bomber 1’s place will be taken by Bomber 2 in the revised sequence. There will be corresponding adjustments all the way up the scale. Patrick’s assumption must be that, in the revised sequence of Bombing Campaign 2, there are only 89 good raids and ten bad raids.
This reasoning leaves me unconvinced. Take, as before, Bomber 45. In Bombing Campaign 2, Raid 45 is a good raid: it is pulling its weight, not dragging down the average. In the original sequence, Bomber 45 kills ten people and saves 1,000 people. If we tackle the counterfactual in the manner suggested by Patrick’s commentary on Bombing Campaign, then we should expect Bomber 46 to take Bomber 45’s place, Bomber 47 to take Bomber 46’s place, and so on. But this is heading towards a result that will fail to uphold Patrick’s belief that Bombing Campaign 2 challenges (8). This is because, at a later point in the revised sequence, Bomber 91 takes Bomber 90’s place. In the original sequence, Bomber 90 kills ten people and saves 1,000 people, and Bomber 91 kills ten people and saves only one person. In the revised sequence, Bomber 91 kills ten people and saves 1,000 people, while Bomber 92 now fills Bomber 91’s shoes, and so all the way up to Bomber 100, who now fills Bomber 99’s shoes. In the revised sequence, in effect, there is no Bomber 100. That is just as well, since Bomber 100 in the original sequence kills ten people and saves only one person.
In summary: in the revised sequence, Raid 91 is a good raid, rather than a bad raid. Thus, in the original sequence, there are 90 good raids and ten bad raids, and in the revised sequence, there are 90 good raids and only nine bad raids. The significance of Bomber 45’s absence from the revised sequence is that there is one fewer raid in which ten people are killed and only one person is saved. Bomber 45 has therefore saved a net total of nine lives in the revised sequence. Even by the logic of the counterfactual test that Patrick is relying upon—a test which, for reasons explained earlier on, we are entitled to quarrel with—Bomber 45’s raid in Bombing Campaign 2 is not disproportionate. In my view, therefore, Bombing Campaign 2 fails to demonstrate what Patrick thinks it does.
- Non-Identity and Faulty Guns
Most of the critical firepower against claims (8) is generated by a case that reflect the concerns of the non-identity problem. Two ingredients are baked into the Tomlin recipe.
First, there is the following normative claim:
Difference View: there is a moral difference between acts that lead to impersonal loss (which make things worse from the point of view of general welfare, but not worse for any particular person), and acts that make things worse for a particular person. All else equal, and in cases in which all lives are worth living, impersonal loss is not as bad as persons being made worse off. (p. 48)
We need a little more commentary on the Difference View, in order to tackle some understandable questions. The impersonal loss of life is always associated with particular lives, so won’t these supposedly impersonal losses be worse for these particular people? How can impersonal loss and particular losses drift apart in this way?
The broad answer will be familiar from cases exemplifying the non-identity problem, which Derek Parfit discussed in detail in Part IV of Reasons and Persons. Imagine that there is a large-scale policy of environmental depletion (call it Depletion), which sacrifices the interests of future generations for current comforts. The fact that future generations will be worse off than they would otherwise be had different policies been adopted can be registered perfectly well within the rubric of impersonal loss. However, Depletion is also likely to affect which particular individuals are born in the future, since it will influence who goes where, who meets whom, and when they reproduce. Thus the identity of future individuals is, in this sense, a product of the same policy that makes them all poorer. If these individuals still have lives that are worth living, and they would not have existed at all but for the existence of Depletion, then their losses are merely impersonal: these losses are not worse for these individuals, since but for Depletion they would not have existed in the first place.
The second ingredient in the Tomlin recipe is the particular case that exploits the non-identity problem. In Faulty Guns, A attacks B with just cause (p. 49). This war is fought, unusually, only with grenade launchers, which are also beset by a strange malfunction: for every grenade launched, a second grenade is also launched that does not explode immediately, but falls on to the battlefield, where it is buried. If, after the war, a child from country B steps on one of these buried grenades, he or she will be killed. It is assumed that 1,000 children from country B will step on them over the next twenty years, and will die as a result. However, these children would not have existed at all but for the socially disruptive effects of the war, and they have had lives worth living up to the moment of their premature deaths.
Patrick believes that ad bellum proportionality calculations can address only the impersonal losses caused by these second grenades, whereas in bello proportionality calculations can register the fact that these grenades make existing children worse off (by killing them). If we have agreed with the Difference View, we will have to accept there is a part of the moral budget that now affects in bello calculations, but not ad bellum calculations. This makes it possible for a misalignment between in bello and ad bellum to open up: at the level of in bello, every act can be disproportionate, because it kills and thus harms an existing individual, but the disproportionateness of each of these acts might nonetheless be consistent with ad bellum proportionality. In effect, ad bellum proportionality applies a discount to these losses, because it can regard them only in terms of impersonal badness, not as person-affecting harms.
Patrick’s toy case helps us to see how the reasoning is supposed to go (p. 51). Imagine that there are two actions, Action A and Action B:
Action A creates Person 1 and later injures Person 2.
Action B creates Person 2 and later injures Person 1.
Note that the “course of action” is simply the combination of Action A and Action B. There is no disconnection between the course of action and the individual actions that compose it (p. 50). Each of these individual actions, however, exemplifies a different sort of disconnection: Action A injures Person 2 but it did not create him, and Action B injures Person 1, but it did not create her. When we assess each of these individual actions, we can entirely ignore the non-identity effects. We can appraise Action A not just for the impersonal disvalue of Person 2’s injury, but also for making Person 2 worse off. Similarly, we can appraise Action B not just for the impersonal disvalue of Person 1’s injury, but also for making Person 1 worse off.
The assessment belonging to the “course of action”, by contrast, must include these identity-affecting factors: the course of action (Action A together with Action B) both creates and injures Person 1 and Person 2. The course of action can register the impersonal disvalue of the injuries suffered by Person 1 and Person 2, but it cannot say, in all honesty, that it has made them worse off. This, then, is the route to the disconnection between ad bellum and in bello proportionality assessment: the extra badness that is registered in the individual actions belonging to in bello assessment has to be ignored in the ad bellum assessment, which is concerned only with the course of action.
One possible concern with this argument is that Patrick may have under-counted the moral significance of individual actions. In his toy example, Action A makes Person 2 worse off, but it also creates Person 1. Why doesn’t the fact that Action A creates Person 1 make a difference to the assessment of Action A? If it does make a difference, then the person-affecting badness of the injury it deals to Person 2 is offset, at least to some extent, by the person-creating value it has for Person 1. After all, the course of action, construed as simply the combination of Action A and Action B, is entitled to represent this information in its summary of the total effects of Action A and Action B. And Patrick is not in a position to deny that this part of the evaluative inventory arising from the course of action is due to Action A in particular. Exactly the same story holds for Action B, which injures Person 1 and also creates Person 2. Since the course of action as a whole can register the fact that it has created Person 2, and since Action B created Person 2, that person-creating fact can belong to the assessment of Action B.
Patrick seems keen to emphasize that there is no funny business going on in the relationship between the course of action as a whole and the individual actions that compose it: one of its constituent actions must have originally supplied whatever can be attributed to the course of action. But if that is so, then what stops us from importing the information represented in the course of action into our appraisal of the constituent actions?
Now perhaps Patrick wishes to deny that creating an individual can be good for her, on the grounds that, if this individual were not created, then there would be no one for whom being brought into existence was better. Some writers have maintained, in response to this worry, that it can be good for an individual, even if it is not better for her, to be brought into existence. There are of course different things to say about these issues. My main worry, however, is that what we attribute to the course of action is going to be obtained, when all is said and done, from the individual constituent actions. If there is information that belongs to the course of action, then it is likely to have a source in one of the constituent individual actions. I fail to see why the moral book-keeping Patrick endorses here can afford to overlook this truth. The corresponding readjustment may then make Faulty Guns a less convincing vehicle for his challenge to (8).
A variant of Faulty Guns is used to point to analogous problems for (7). Similar problems arise on this front, but I need not labour them here.
- Revising Revisionism
In the final pages of his paper, and under the assumption that (7), (7*), (8), and (8*) are all untenable, Patrick takes stock and reviews some options for revising revisionism. Now I am less confident that Patrick’s arguments against these claims have succeeded. Even so, there are various loose ends to worry about. It is not so clear how ad bellum and in bello proportionality concerns can be fully commensurated.
One revisionist option is simply to concede that (7) and (8) are false, leave as much of everything else in place as possible, and try to tough things out. Patrick thinks that this option is intolerable, because it will lead in some circumstances to a “bizarre standoff between states and their soldiers” (p. 55). An example near to hand is Bombing Campaign 2, in which state A is permitted to order its combatants to fight, because the war as a whole will be proportionate, but in which A’s combatants are required to refuse to fight, because their individual acts will be disproportionate. I agree with Patrick that, if there were cases embodying these discrepancies between every individual action and the course of action—I have, of course, expressed misgivings about his arguments—this standoff would be hard to stomach.
Another broad strategy is to allow either ad bellum or in bello to assume a dominant place in revisionism’s account of the relationship between these dimensions, and to make the other dimension a derivative or junior partner. Here is one possibility (p. 56):
Revisionist ad Bellum Proportionality: a war is ad bellum disproportionate if and only if the acts of war are, taken collectively, sufficiently in bello disproportionate.
This “master principle” establishes the dominance of in bello questions over ad bellum questions: we extract an ad bellum verdict from the in bello verdict. The principle implies that a collection of actions, each of which is disproportionate, will necessarily enforce a verdict of ad bellum disproportionateness. However, as Patrick sees things, this will not tackle the challenge posed by Bombing Campaign. (Again, I have expressed my doubts about Bombing Campaign.)
Another candidate for the master principle reverses the direction of dominance. This time, ad bellum is in charge (p. 56):
Revisionist in Bello Proportionality I: an act of war is in bello proportionate if and only if it contributes to an ad bellum proportionate war.
This principle requires the abandonment of (6*), which has been, supposedly, the centrepiece of revisionist theorizing. But the principle seems unsatisfactory in any case, because it is too slack: as long as a war satisfied ad bellum, we would be unable to condemn any individual action as disproportionate. We need more hands-on regulation than that. To this end, here is a revised version of this master principle (p. 57):
Revisionist in Bello Proportionality II: an act of war is in bello proportionate if and only if it makes a proportionate contribution to an ad bellum proportionate war.
While avoiding the errors of Revisionist in Bello Proportionality I, this principle now needs a way of filling out what a “proportionate contribution” is going to be. Moreover, a satisfactory specification of this further condition will no longer be settled by ad bellum.
There are problems, then, wherever you look. Patrick’s final revisionist suggestion is a complex one, with independent provision for both ad bellum considerations and in bello considerations (pp. 59-60). It also requires the revisionist programme to distinguish between assessment of the war “as a whole”, and assessment of the strength of the case for the resort to war, or (1) and (2) in the list above (p. 58). It therefore seems to me to lean in spirit a bit more towards Revisionist ad Bellum Proportionality than one of the Revisionist in Bello Proportionality principles. There are two proportionality principles that have to be balanced in an overall determination of proportionality:
One looks at the harms and goods produced by the “war as a whole.” The other looks at the amalgamated unjustifiability (including the disproportionality) of the individual acts of war. (p. 59)
In some cases, such as Unruly Army (p. 59), in which the “war as a whole” proportionality assessment might favour war, but in which many individually disproportionate acts by ill-disciplined combatants are expected to take place, this second “amalgamated unjustifiability” proportionality condition might make the war impermissible, because disproportionate overall. It thus counts as a highly revisionary form of revisionism.
Now State Official in Unruly Army has the choice whether to authorize its combatants to action. But she surely does not permit them to perform impermissible acts of war. I feel the pull here of a theory which requires independent restrictions on proportionate in bello contribution, just as Revisionist in Bello Proportionality II suggests. At the same time, the proportionality of war as a whole casts a long shadow. Go back to Bombing Campaign, and to Bomber 45. Bomber 45’s raid is part of a series of raids that were together chasing a certain target. The fate of each raid, as far as the proportionality verdict goes, may be dependent on the outcomes achieved by the bombing campaign as a whole. We can say with confidence, just by looking at the individual contributions, that Bomber 95 makes a better contribution than Bomber 45. (That is what entitled me to describe Bomber 95’s raid as a “good” raid and Bomber 45’s raid as a “bad” raid.) Nonetheless, there is a sense in which individual contributions to war still have to be assessed derivatively, in terms of what they together achieve. Bomber 95’s raid is not fully satisfactory, and may not even qualify as individually proportionate, because it was part of a series of raids that cumulatively failed to produce an outcome where the benefits exceeded the costs. If this is so, the proportionality of the war as a whole cannot fail to have a bearing on the proportionality of individual contribution. The significance of what Bomber 45 and Bomber 95 do individually cannot be fixed in the absence of an understanding of what they do together. But it also seems to me that individual proportionality is not fully settled by these ad bellum questions. It may be necessary for being individually proportionate that a raid belongs to a campaign that is proportionate overall, but it is also insufficient. Full commensurability between ad bellum and in bello thus seems difficult to achieve.
It is far from clear that what needs to be provided for in order to respect proportionality can be successfully contained within a single proportionality principle. I suspect that non-revisionists, as well as revisionists, may be following these developments with interest: perhaps, just as they always supposed, there is no satisfactory way of fully commensurating ad bellum and in bello. They each go their own way.