Optimal Torture

Should we torture military prisoners? Let's ignore moral issues and just consider what we should do if our goal was to maximize the welfare of U.S. soldiers.

One anti-torture argument you could make is that by not torturing we would be promoting universal human values that discourage torture. But as Bryan Caplan writes, this is probably a weak argument since "Most people are going to listen to the tribalist who declares, 'All I care about is saving American/German/Russian/Roman lives!'"

Bryan also writes that by not torturing prisoners we encourage enemy combatants to surrender. This is true. But by not torturing prisoners we also decrease the cost to people of becoming a U.S. enemy and so we increase the number of people who will fight us. Of course, this only applies to enemies who must recruit fighters and can't simply draft solders. Consequently, the U.S. should be less reluctant to torture enemies in our current conflict than we were during WWII.

My guess is that to minimize the chance of captured Americans being tortured we should torture captured enemies if and only if these enemies come from forces that torture Americans. Such reciprocity would create an incentive for enemies not to torture our captured troops. (Of course, for moral reasons we might decide not to torture enemies even if abstaining from torture harms our solders.)

It might seem strange that I would write about optimal torture. But one of my PhD dissertation advisers (Michael Chwe) wrote a paper on torture in principal-agent models.