I’m repulsed that we’re even discussing whether it’s acceptable to waterboard someone 183 times, but the key point in the debate over the effectiveness of torture (as one of Diane Rehm’s guests pointed out Thursday) is whether the “enhanced interrogation techniques” generate new intelligence that would not have been obtained using conventional techniques.

It drives me nuts to hear people claim that we know these techniques work because suspected terrorist X gave us some piece of information after being subjected to torture. Even if those claims were true (many are not), they do not prove that torture works. If the same suspected terrorists would have provided the same information under conventional interrogation in the same amount of time (as former FBI interrogator Ali Soufan claims), then there is no evidence that torture is more effective. However, we do not observe that counterfactual scenario (a problem known as “the fundamental problem of causal inference”) and are thus forced to speculate about what would have happened.

(The debate on this point will almost surely be inconclusive. To evaluate this question scientifically, one would have to unethically randomize prisoners into torture and conventional interrogation conditions. In the absence of such data, both sides are likely to rely on their ideological preconceptions to interpret the anecdotal accounts that are released to the public.)