The law of effect, randomization and Newcomb’s problem — LessWrong