For the ticking-time-bomb scenario, I think Ayn Rand, actually, was close to the right answer. If you try to use any system for something it wasn’t designed for, it’s going to fall apart. If you apply Newtonian mechanics to subatomic high-energy interactions, you’ll get the wrong answer. If you try to build a house with the same carpentry techniques you use for a bookshelf, it’ll fall apart. There are enough conditions tacked on to the front of the time-bomb hypothetical to make it almost totally irrelevant to real life, which is where we need morality to operate.

A large part of the reason for side-constraints—especially constraints on what we can do to bad people—comes from our lack of knowledge: we don’t know that we have the right guy, we don’t know that torturing him will help, we don’t know that deciding to torture him won’t lead us to torture in less acceptable circumstances, we don’t know that the threat is actually valid. If you abstract away all uncertainty you can take more drastic actions, because you’re more sure they’ll work. But that doesn’t help us.