Belief Embeddings with Large Language Models
2026-03-17
Citation#
Lee, Byunghwee, Rachith Aiyappa, Yong-Yeol Ahn, Haewoon Kwak, and Jisun An. 2025. A semantic embedding space based on large language models for modelling human beliefs. Nature Human Behaviour.
Overview#
This paper develops a semantic embedding space based on large language models to model human beliefs and positions in public debate. The project connects representation learning with computational social science and provides a framework for studying relative belief structure at scale.
Why I Think It Matters#
- It creates a computational representation of beliefs rather than reducing opinion to simple labels
- It links language model representations to social and political judgment
- It opens a path toward measuring belief change and persuasion in a structured way