Debiasing as Non-Self-Destruction — LessWrong