Loss aversion is not what you think it is — LessWrong