Carter is now discussing the problems inherent in the twin principles enshrined in both just war theory: proportionality and discrimination.

Proportionality means that you should use the method that will produce the fewest casualties. But taken to its logical extreme, this would dictate that commanders are obligated to take an objective in a way that kills 100 of their guys and 200 of their guys, instead of in a way that kills 10 of your guys and 1,000 of their guys. Indeed, some proponents of international law do argue this. But it is morally thorny (and you can then start asking what minimizes net casualties over the course of the war, rather than current ones). Moreover, if you have a theory of war that tells commanders they should sacrifice large numbers of their men to save the enemy, you will have a theory of war that is never applied to an actual war.

The problem of discrimination is also difficult. We think of just war theory as telling us that you can't target civilians, but in fact it says you can't target "non-combatants"--that's why you can't shoot prisoners. But this becomes extremely difficult. Who is a non-combatant? The cooks? A general back at headquarters? Infantry troops who happen to be asleep?

He dives a little bit into the issue of prisoner treatment. Like pretty much everyone else in the room, he's an anti-torture hardliner. But he also points out how little most of the people arguing about the Geneva Convention actually know about its provisions. For example, we may all agree that you shouldn't hurt people to get them to cooperate--but the laws of war also forbid giving prisoners so much as a Hershey bar in exchange for cooperation. This makes sense under the logic of the convention, which as much as anything, is a list of ways for warring states to avoid wasting energy on tit-for-tat practices. But it has little to do with the moral intuitions that drive most of us to get angry when people violate the Geneva Convention.

We want to hear what you think about this article. Submit a letter to the editor or write to