Red Flags for Rationalization — LessWrong