Testing for Rationalization — LessWrong