Regensburg 2025 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 3: Poster

SOE 3.9: Poster

Monday, March 17, 2025, 17:30–19:30, P4

Characterizing similarities and differences in people's views based on open-ended expressions with LLMs and network science — •Ezequiel Lopez-Lopez and Stefan Herzog — Max Planck Institute for Human Development, Berlin

Citizens have diverse views on critical issues like pandemic management and climate change. Current methods to gather these views are limited: polls and surveys lack nuance, while qualitative approaches do not scale. We developed a computational framework that combines Large Language Models with network science to rapidly analyze citizens' evolving views without oversimplifying them. This novel approach extracts concepts from unstructured text and represents them in networks, offering deeper insights and faster analysis than traditional qualitative methods. We applied this framework to UK citizens' ideas on five UN Sustainable Development Goal-related policy problems. Our results demonstrate the framework's ability to capture nuanced differences across political, demographic, and cognitive variables. This approach has the potential to significantly enhance evidence-based policymaking and citizen engagement in complex societal issues.

Keywords: LLMs; Networks; Policymaking; Open-ended questions; Citizens' perspectives