A semantic embedding space based on large language models for modelling human beliefs

作     者：Lee, Byunghwee Aiyappa, Rachith Ahn, Yong-Yeol Kwak, Haewoon An, Jisun 

作者机构：Indiana Univ Ctr Complex Networks & Syst Res Luddy Sch Informat Comp & Engn Bloomington IN 47405 USA 

出 版 物：《NATURE HUMAN BEHAVIOUR》 (Nat. Hum. Behav.)

年 卷 期：2025年

页      面：1-13页

核心收录：

基　　金：United States Department of Defense | United States Air Force | AFMC | Air Force Office of Scientific Research (AF Office of Scientific Research) [FA9550-25-1-0087] Air Force Office of Scientific Research [HR001121C0168] DARPA Luddy Faculty Fellow Research Grant Programme of the Luddy School of Informatics, Computing and Engineering at Indiana University Bloomington 

摘      要：Beliefs form the foundation of human cognition and decision-making, guiding our actions and social connections. A model encapsulating beliefs and their interrelationships is crucial for understanding their influence on our actions. However, research on belief interplay has often been limited to beliefs related to specific issues and has relied heavily on surveys. Here we propose a method to study the nuanced interplay between thousands of beliefs by leveraging online user debate data and mapping beliefs onto a neural embedding space constructed using a fine-tuned large language model. This belief space captures the interconnectedness and polarization of diverse beliefs across social issues. Our findings show that positions within this belief space predict new beliefs of individuals and estimate cognitive dissonance on the basis of the distance between existing and new beliefs. This study demonstrates how large language models, combined with collective online records of human beliefs, can offer insights into the fundamental principles that govern human belief formation.