UniGO: A Unified Graph Neural Network for Modeling Opinion Dynamics on Graphs

Hao Li, Hao Jiang, Yuke Zheng, Hao Sun, Wenying Gong

Track: Social networks and social media

Keywords: Opinion Dynamics, Graph Nerual Networks, Graph Coarse, Social Networks

Abstract: Polarization and fragmentation in social media amplify user biases, making it increasingly important to understand the evolution of opinions. Opinion dynamics provide interpretability for studying opinion evolution, yet incorporating these insights into predictive models remains challenging. This challenge arises due to the inherent complexity of social interactions, the diversity of opinion fusion rules, and the difficulty in capturing equilibrium states while avoiding over-smoothing. This paper introduces UniGO, a unified framework for modeling opinion evolution on graphs. By abstracting various opinion dynamics models into a unified graph-based structure, UniGO captures both common features and complex fusion rules. Using a coarsen-refine mechanism, UniGO efficiently models opinion dynamics through a graph neural network, mitigating over-smoothing while preserving equilibrium phenomena. Additionally, UniGO leverages pretraining on synthetic datasets, which enhances its ability to generalize to real-world scenarios, providing a viable paradigm for large-scale applications of opinion dynamics. Experimental results on both synthetic and real-world datasets demonstrate UniGO's effectiveness in capturing complex opinion formation processes and predicting future evolution. The pretrained model also shows strong generalization capability, validating the benefits of using synthetic data to boost real-world performance.

Submission Number: 1623