Models of preferences in distant situations — LessWrong