Two concepts of torture
By More than 50 years ago John Rawls wrote a famous essay in which he defended utilitarian theories of punishment from the charge they justified punishing innocent people under certain circumstances. The idea is that if the central function of punishment is to deter crime (as opposed to a retributive view in which its function is to give people what they deserve) then there are circumstances, at least in theory, in which this primary deterrence function would be served by framing an innocent person. Rawls’ reply was that while it was easy to construct hypothetical scenarios in which this would be true, it was much harder and probably impossible to construct an institution of punishment as a social practice that would allow for punishing the innocent, and yet would still be desirable from a utilitarian perspective.
The analogy to torture should be obvious. Unless you’re a strict deontologist, it’s fairly easy to construct Jack Bauer hypotheticals in which in theory it might be desirable to torture someone. (For instance, would you waterboard George Steinbrenner to save one million children? )
What’s much harder is to imagine a real-world set of institutions and social practices which could somehow limit torture to these very rare circumstances while still producing enough benefits to justify its use. In other words, there’s all the difference in the world between discussions of torture in the abstract, and actually creating a bureacratic machine that wouldn’t, for example, end up waterboarding a mentally ill low-level terrorist 83 times in return for no useful information.