LLMs Can't Rank Culture Causes

Comparing human polls and several large language models, Robin Hanson found weak correlations and inconsistent rankings when asking them to rate 16 candidate causes of cultural change across two historical periods. The disagreement suggests both that cultural causation is multi‑factorial and that current AI tools give unreliable, nonconvergent causal judgments on complex social history. — If LLMs and quick polls disagree about why cultures change, relying on automated or shallow quantitative summaries to explain cultural shifts risks misleading policymakers, journalists, and educators.