Characterizing Opinion Evolution of Networked LLMs
Abstract
Large language models (LLMs) increasingly interact with one another in multi-agent systems, from simulations of human discourse to influence operations and fully LLM-driven social platforms. These interactions give rise to new regimes of opinion propagation that are not yet well understood. We investigate whether classical opinion dynamics models, which have long been used to explain how interactions shape collective beliefs in human societies, can capture the behavior of LLM networks. Specifically, we quantify the relative contributions of classical opinion-dynamics mechanisms, including neighbor influence, stubbornness, homophily, and bias, to the evolution of LLM opinions. Among the mechanisms considered, a bias term capturing regression toward a model- and topic-dependent opinion emerges as a significant explanatory component, with its inclusion reducing cumulative estimated mean opinion error by up to 88%. We additionally observe similar trends across the evaluated model families, discussion topics, and network structures.
Related
- Characterizing the temporal evolution of Biermann-battery-driven magnetic reconnection in laser-ablated plasmas
- Low-Dimensional Phase Diagram of Higher-Order Networked Systems
- Digital Twins of Networked Dynamical Systems: Machine Learning Prediction of Synchronization Transitions in Coupled Chaotic Oscillators
- Experimental Private Quantum Networked Sensing
- Fundamental Dynamical Units for Physics-Informed Structural Inference from Perturbation Time-Series in Networked Systems
BibTeX
@misc{probine2026characterizing,
title = {{Characterizing Opinion Evolution of Networked LLMs}},
author = {Caleb Probine and Yigit Ege Bayiz and Filippos Fotiadis and Samuel Li and Yunhao Yang and Ufuk Topcu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.18276},
eprint = {2606.18276},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.18276},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Characterizing Opinion Evolution of Networked LLMs
AU - Caleb Probine
AU - Yigit Ege Bayiz
AU - Filippos Fotiadis
AU - Samuel Li
AU - Yunhao Yang
AU - Ufuk Topcu
PY - 2026
DA - 2026/10/10
PB - arXiv
DO - 10.48550/arxiv.2606.18276
UR - https://arxiv.org/abs/2606.18276
AB - Large language models (LLMs) increasingly interact with one another in multi-agent systems, from simulations of human discourse to influence operations and fully LLM-driven social platforms. These interactions give rise to new regimes of opinion propagation that are not yet well understood. We investigate whether classical opinion dynamics models, which have long been used to explain how interactions shape collective beliefs in human societies, can capture the behavior of LLM networks. Specifically, we quantify the relative contributions of classical opinion-dynamics mechanisms, including neighbor influence, stubbornness, homophily, and bias, to the evolution of LLM opinions. Among the mechanisms considered, a bias term capturing regression toward a model- and topic-dependent opinion emerges as a significant explanatory component, with its inclusion reducing cumulative estimated mean opinion error by up to 88%. We additionally observe similar trends across the evaluated model families, discussion topics, and network structures.
ER -