Abstract
Beliefs form the foundation of human cognition and decision-making, guidingour actions and social connections. A model encapsulating beliefs and theirinterrelationships is crucial for understanding their influence on our actions.However, research on belief interplay has often been limited to beliefs relatedto specific issues and relied heavily on surveys. We propose a method to studythe nuanced interplay between thousands of beliefs by leveraging an online userdebate data and mapping beliefs onto a neural embedding space constructed usinga fine-tuned large language model (LLM). This belief space captures theinterconnectedness and polarization of diverse beliefs across social issues.Our findings show that positions within this belief space predict new beliefsof individuals and estimate cognitive dissonance based on the distance betweenexisting and new beliefs. This study demonstrates how LLMs, combined withcollective online records of human beliefs, can offer insights into thefundamental principles that govern human belief formation.