Epistemological Vigilance for Alignment — AI Alignment Forum