Culturally Grounded Personas in Large Language Models: Characterization and Alignment with Socio-Psychological Value Frameworks

This research investigates the fidelity of Large Language Models (LLMs) in simulating human behavior, specifically focusing on how accurately synthetic, culturally-grounded personas align with established socio-psychological value frameworks. The study aims to conceptualize and generate LLM personas using variables derived from the World Values Survey (WVS) and assess their alignment with the Inglehart-Welzel Cultural Map and Moral Foundations Theory, thereby evaluating their interpretation of stable cultural differences.

The accurate reflection of cultural and moral values by LLM personas is critical for their reliable application in global contexts, impacting areas from strategic communication to policy simulation. Ensuring LLMs can genuinely represent diverse cultural perspectives is foundational for building trustworthy and effective AI systems that avoid bias and promote equitable outcomes.

The study addresses the uncertainty surrounding LLM-generated personas' ability to accurately reflect world and moral value systems across diverse cultural conditionings.

Forward consideration, not a verified fact.