How to Defend the Indefensible — LessWrong