LessWrong's 'Hyperstition' Doesn't Just Question Rationality — It Calls It a Trap

In a thought-provoking LessWrong post, user alseph explores 'hyperstition'—the idea that believing something false can sometimes yield better outcomes than being rational. The article critiques two key rationalist texts: Scott Alexander's 'Diseased Thinking', which proposes a consequentialist model for moralizing issues like obesity, and Yudkowsky's 'Why Our Kind Can't Cooperate', which argues that more knowledge should never lead to worse performance. alseph counters that embedded agents waste effort debating optimal strategies, and that genuine false beliefs (e.g., in divine punishment) can create cooperative equilibria that rational analysis cannot replicate.

The post highlights a painful paradox for rationalists: they can accurately describe why conservative societies 'work' (through shared, unfounded beliefs) but cannot consciously adopt those beliefs themselves. This creates a 'barrier' where rationalists become the enemy of such beneficial hyperstitions. The piece echoes themes from postmodern theory but applies them to decision theory and social epistemology, suggesting that pure rationality might be self-limiting in real-world embedded contexts. The discussion has sparked engagement on LessWrong with 11 upvotes, signaling its resonance with the rationalist community grappling with the limits of pure reason.