Bayesianism for humans: prosaic priors — LessWrong