A case study in fooling oneself — LessWrong