Normative uncertainty in Newcomb's problem — LessWrong