Doublethink (Choosing to be Biased) — AI Alignment Forum