Subtle Forms of Confirmation Bias — LessWrong