Abstract
The Internet has fundamentally reshaped the formation and diffusion of public opinion in modern society. However, existing simulation studies often face challenges such as insufficient fidelity in evolutionary dynamics, behavioral homogenization, and high modeling costs. This study develops a realistic public opinion simulation system that integrates macro-level diffusion patterns with micro-level individual cognition, addressing the traditional models’ limitations in generalizability and high resource demands. This study proposes an LLM-based multi-agent simulation framework for public opinion dissemination. The framework first constructs behavior probability profiles calibrated against real-world social media data, following exponential and normal distribution models to constrain agent behavior frequency and topical tendencies (macro-dynamics). Concurrently, by utilizing an LLM as the cognitive core, the system enables dynamic semantic content generation via the integration of Public Opinion Simulation Standard Operating Procedure (PSOP) and a Global Information Sharing Pool (GISP) (micro-cognition). Comparative experiments across two distinct scenarios—“Public Policy” and “Food Safety”—show that (1): The framework exhibits significant cross-scenario robustness, replicating the incubation-eruption-decay’ evolutionary pattern, which aligns with propagation dynamics without the need for parameter adjustment (2); Quantitative evaluation shows that the normalized agent behavior distribution entropy reaches 0.69; moreover the Distinct-2 metric for generated content reaches 0.83, substantially mitigating behavioral homogenization (3); Compared with traditional deep learning methods, this framework supports zero-shot cold start, with negligible computational costs for domain adaptation. By unifying the agent architecture, this research enables efficient simulation of public opinion evolution, providing a low-cost, high-fidelity methodological paradigm for public opinion crisis management.
Similar content being viewed by others
Data availability
The datasets analyzed during the current study are available from the corresponding author on reasonable request.
References
Chen, L., Chen, J. & Xia, C. Social network behavior and public opinion manipulation. J. Inf. Secur. Appl. 64, 103060. https://doi.org/10.1016/j.jisa.2021.103060 (2022).
Huang, S., Xiu, B. & Feng, Y. Modeling and simulation research on propagation of public opinion. In IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC), 380–384, 380–384, (2016). https://doi.org/10.1109/IMCEC.2016.7867238 (IEEE, 2016).
Lian, S. & Xue, Y. Sudden diffusion public opinion tendency forecast method: simulation research. Comput. Simul. 33 (7), 335–338 (2016).
Li, C., Li, X. & Duan, S. Modeling, analysis, and applications of complex systems. Abstr. Appl. Anal. 2013, 41–62. https://doi.org/10.1155/2013/798173 (2013).
Zou, J. The construction and operation of an integrated governance system for China’s online public opinion. Journal Nanjing Normal Univ. (Social Sci. Edition) 116–126 (2020).
Devlin, J., Chang, M. W., Lee, K., Toutanova, K. & BERT. : Pre-training of deep bidirectional transformers for language understanding, (2018). https://doi.org/10.48550/arXiv.1810.04805
Xu, F. et al. Are large language models really good logical reasoners? A comprehensive evaluation from deductive, inductive and abductive views, (2023). https://doi.org/10.48550/arXiv.2306.09841
Binz, M. & Schulz, E. Turning large language models into cognitive models, DOI: 10.48550/arXiv.2306.03917 (2023).
Abelson, R. P. & Bernstein, A. A computer simulation model of community referendum controversies. Public Opin. Q. 27, 93–122 (1963).
Chatterjee, R. & Eliashberg, J. The innovation diffusion process in a heterogeneous population: A micromodeling approach. Manage. Sci. 36, 1057–1079 (1990).
Sudbury, A. The proportion of the population never hearing a rumor. J. Appl. Probab. 22, 443–446 (1985).
Leskovec, J., McGlohon, M., Faloutsos, C., Glance, N. & Hurst, M. Patterns of cascading behavior in large blog graphs. In Proceedings of the SIAM International Conference on Data Mining, 551–556 (SIAM, 2007)., 551–556 (SIAM, 2007). (2007).
Wang, G., Chi, Y. & Liu, Y. Studies on a multidimensional public opinion network model and its topic detection algorithm. Inf. Process. Manag. 56, 584–608. https://doi.org/10.1016/j.ipm.2018.12.002 (2019).
Shoemaker, P. J., Eichholz, M. & Kim, E. Individual and routine forces in gatekeeping. J. Mass Commun. Q. 78, 233–246 (2001).
Tan, L., Yang, Y., Han, P. & Li, Y. Agent-based modeling and simulation of the public opinion propagation in the small- world network. In 7th International Conference on Biomedical Engineering and Informatics, 940–944 (IEEE, 2014)., 940–944 (IEEE, 2014). (2014).
Das, A., Gollapudi, S. & Munagala, K. Modeling opinion dynamics in social networks. In Proceedings of the 7th ACM International Conference on Web Search and Data Mining, 403–412ACM, (2014).
Jin, X. & Wang, Y. Research on social network structure and public opinions dissemination of micro-blog based on complex network analysis. J. Networks. 8 (7), 1543–1550 (2013).
Yin, H., Yang, S. & Li, J. Detecting Topic and Sentiment Dynamics Due to COVID-19 Pandemic Using Social Media In (eds Yang, X. et al.) (2020).
Chen, X., Fan, H. & Duan, S. Public opinion analysis of big data based on machine learning. J. Phys. Conf. Ser. 1302, 022035 (2019).
Saputro, E. Revisiting the research on crisis communication and media use during the time of crisis. In ICSSED 2020, (2020). https://doi.org/10.4108/eai.4-8-2020.2302435
Meqdad, M. N., Al-Akam, R. & Kadry, S. New prediction method for data spreading in social networks based on machine learning algorithm. TELKOMNIKA (Telecommunication, Computing, Electronics and Control) 18, 3331–3338 (2020).
Fatemi, B., Molaei, S. & Pan, S. GCNFusion: An efficient graph convolutional network-based model for information diffusion. Expert Syst. Appl. 202, 117053 (2022).
Feng, Z., Xing, Y. & Wang, G. Distributed opinion competition scheme with gradient-based neural network in social networks. Sci. Rep. 14, 30883. https://doi.org/10.1038/s41598-024-30883-3 (2024).
Li, H., Jiang, H., Zheng, Y. & UniGO A unified graph neural network for modeling opinion dynamics on graphs. In Proceedings of the ACM Web Conference 2025, 530–540ACM, (2025).
Zhao, J. H., Zeng, D. L., Qin, J. T., Si, H. M. & Liu, X. F. Simulation and modeling of microblog-based spread of public opinions on emergencies. Neural Comput. Appl. 33, 547–564 (2021).
Touvron, H., Martin, L. & Stone, K. LLaMA 2: Open foundation and fine-tuned chat models, DOI: 10.48550/arXiv. 2307.09288 (2023).
Gao, C., Lan, X. & Li, N. Large language models empowered agent-based modeling and simulation: A survey and perspectives. Humanit. Social Sci. Commun. 11, 1–24. https://doi.org/10.1057/s41599-023-02072-9 (2024).
Qu, Y. & Wang, J. Performance and biases of large language models in public opinion simulation. Humanit. Social Sci. Commun. 11, 1–13 (2024).
Vora, M., Blau, T. & Kachhwal, V. Large language model for Bible sentiment analysis: Sermon on the Mount, (2024). https://doi.org/10.48550/arXiv.2401.00689
Drescher, L. S., Roosen, J. & Aue, K. The spread of COVID-19 crisis communication by German public authorities and experts on Twitter: Quantitative content analysis. JMIR Public Health Surveill. 7, e31834. https://doi.org/10.2196/31834 (2021).
Xin, W. Psychological analysis of the public opinion in the development of network hot events: A case study of Duo Mao Mao event. Journal Yunnan Normal Univ. (Humanities Social Sciences) (2012).
Weng, L. LLM powered autonomous agents. Online blog post URL: (2023). https://lilianweng.github.io/posts/2023-06-23-agent/
Sid, R. Simulating Society: Monte Carlo Simulations with LLM Agents. URL: https://www.srao.blog/p/simulating-society-monte-carlo-simulations
Nasim, M., Gilani, S. M., Qasmi, A. & Naseem, U. Simulating Influence Dynamics with LLM Agents. arXiv preprint arXiv:2503.08709. (2025).
Yao, J. et al. Social opinions prediction utilizes fusing dynamics equation with LLM-based agents. Sci. Rep. 15(1), 15472 (2025).
Wang, R., Lu, T., Zhang, P. & Gu, N. Data-driven agent-based model for public opinion propagation simulation in cyberbullying. Big Data Min. Analytics 8(4), 794–819 (2025).
Barabási, A.-L. The origin of bursts and heavy tails in human dynamics. Nature 435, 207–211 (2005).
Cha, M., Haddadi, H., Benevenuto, F. & Gummadi, K. Measuring user influence in Twitter: The million follower fallacy. In Proceedings of the 4th International Conference on Weblogs and Social Media (ICWSM), (2010).
Zhou, D., Scharli, N., Hou, L. & Wei, J. Least-to-most prompting enables complex reasoning in large language models, (2022). https://doi.org/10.48550/arXiv.2205.10625
Hong, C. A study on the development cycle of network hotspot events. J. Intell. 31, 45–50 (2012).
Shannon, C. E. A mathematical theory of communication. Bell Syst. Tech. J. 27(3), 379–423 (1948).
Balch, T. Hierarchic social entropy: An information theoretic measure of robot group diversity. Auton. Robots. 8 (3), 209–238 (2000).
Bettini, M. et al. Measuring behavioral heterogeneity in multi-agent reinforcement learning. Journal Mach. Learn. Research (JMLR). 26, 1–59 (2025).
Li, J. et al. A diversity-promoting objective function for neural conversation models. In Proceedings of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT), 110–119 (2016).
Liu, S. et al. Rethinking and refining the distinct metric. In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL), 651–656 (2022).
Author information
Authors and Affiliations
Contributions
The conception and design of the study were jointly developed by all authors. Hua Hu and Peng Cheng Guo were responsible for data collection and simulation. Hai Lan and Qi Huang drafted and revised the manuscript. All authors read and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Lan, H., Hu, H., Guo, P.C. et al. Public opinion dissemination simulation based on large language model multi-agent systems. Sci Rep (2026). https://doi.org/10.1038/s41598-026-44206-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-44206-z
