Dreaming of Political Bayescraft — LessWrong