Convenient explanations — LessWrong