A Self-Respect Feedback Loop — LessWrong