Beyond Static Snapshots: Dynamic Modeling and Forecasting of Group-Level Value Evolution with Large Language Models

Researchers have developed a new LLM framework that dynamically models and forecasts the evolution of group-level social values, moving beyond static snapshots. Using World Values Survey data from China and the U.S., the model integrates historical trajectories, social events, and group attributes. It achieved a maximum 33.97% improvement in accuracy on unseen questions compared to baseline methods, revealing that U.S. groups and younger demographics are more volatile to external changes.