x
Consequentialist preferences are reflectively stable by default — AI Alignment Forum