Being legible to other agents by committing to using weaker reasoning systems — AI Alignment Forum