Possibly dumb microeconomics question: heuristically, when is expected utility theory a good model, and when do you need to bring in something like prospect theory? I get the idea that people are never "truly" rational, but when does it tend to be a decent approximation?