Consequentialist cognition — LessWrong