I wonder if it's better or worse to construct problems that are implausible from the very start, instead of being potentially realistic up to a certain point where you're asked to suspend disbelief. (Similar to how we do decision problems here, with Omega being portrayed as a superintelligence from another galaxy who is nearly omniscient and whose sole goal appears to be giving people confusing decision problems. IIRC, conventional treatments of decision theory often portray the Predictor as a human and do not explain why his predictions tend to be accurate, only specifying that he has previously been right 99% or 100% of the time. I suspect that format tends to encourage people to make excuses not to answer the real question.) So, suppose instead of the traditional trolley problem, we say "An invincible demon appears before you with a hostage tied to a chair, and he gives you a gun. He tells you that you can shoot the hostage in the head or untie her and set her free, and that if and only if you set her free, he will go off and kill five other people at random. What do you do?" Does that make it better in worse, in terms of your ability to separate the implausibility of the situation from your ability to come to a moral judgment?