Geometric Utilitarianism — LessWrong