One of the most facile dismissals of torture is that it doesn't work, so why bother? That's tempting, but it's too easy. Torture seems to me very likely to work provided that you can verify the information, which I assume interrogators can in at least some circumstances. Nor is it obvious to me that the quality of information is likely to be lower than that obtained by other means: yes, people will say anything to avoid torture, but they'll also say anything to avoid imprisonment. Maybe the lies will be vivider or more voluble under torture, but it doesn't seem necessarily so that the ratio of lies to truth will increase.
I'd rather see people take the hard stance and say "Yeah, torture may still work, but we still shouldn't use it because it's wrong." Otherwise, you're kind of stuck if someone comes up with a way to make it effective. I've been thinking about this in relation to the much vaunted lie detecting brain scans. Most people have talked about the implications for the criminal justice system--does the fifth amendment still apply? But what I wonder is, what does this mean for torturers? If you can actually tell accurately when someone is lying, torture suddenly becomes very, very effective, doesn't it? And yet, it would still be wrong. So make the case on those grounds. Efficiency is a dangerous red herring.