LLM-Driven Social Simulations Face Significant Epistemic Validity Challenges
A review article mapping the methodological vulnerabilities of LLM-driven agents highlights that surface-level narrative fluency often masks underlying epistemic invalidity ...

