The Necessity of Credibility — LessWrong