Suppose a disaster could be prevented by foresight, or narrowly averted by heroic action. Which one is more
awesome? Which one is better?
Preventing disaster by foresight is more likely to work than narrow aversion by heroic action, so the the awesomeness of foresight working gets multiplied by a larger probability than the awesomeness of heroic action working when you decide to take one approach over the other. This advantage of the action that is more likely to work belongs in decision theory, not your utility function. Your utility function just says whether one approach is sufficiently more awesome than the other to overcome its decision theoretic disadvantage. This depends on the probabilities and awesomeness in the specific situation.
Preventing disaster by foresight is more likely to work than narrow aversion by heroic action, so the the awesomeness of foresight working gets multiplied by a larger probability than the awesomeness of heroic action working when you decide to take one approach over the other. This advantage of the action that is more likely to work belongs in decision theory, not your utility function. Your utility function just says whether one approach is sufficiently more awesome than the other to overcome its decision theoretic disadvantage. This depends on the probabilities and awesomeness in the specific situation.