谣言传播模型与仿真方法研究进展

Research Progress on Rumor Propagation Models and Simulation Methods

摘要: 为梳理复杂社交网络中谣言传播模型与仿真方法的发展脉络，本文围绕经典谣言传播理论、微分方程模型、复杂网络拓扑与高阶交互模型以及大语言模型智能体与动力学约束融合模型展开综述。首先，从Allport-Postman模型、Knapp谣言类型分类和仓室建模思想出发，阐明经典理论对传播变量选取和状态划分的基础作用；其次，分析DK模型、分数阶ISDR模型和SKT型偏微分方程模型在传播机制、记忆效应、辟谣反馈、空间扩散和时空斑图刻画中的建模意义；进一步讨论无标度网络、社区结构、节点中心性、单纯复形和超图等拓扑结构在关键传播者识别、群体协同传播和突发扩散分析中的作用；最后，讨论LLM智能体与元胞自动机、SIR机制耦合在意见演化仿真中的应用。研究表明，谣言传播建模正由宏观经验解释逐步转向动态机制刻画、拓扑结构约束、时空耦合分析和智能体仿真融合，可为复杂网络环境下的舆情预测与干预策略设计提供理论参考。

Abstract: To clarify the development of rumor propagation models and simulation methods in complex social networks, this paper reviews classical rumor propagation theories, differential equation models, complex network topology and higher-order interaction models, and models integrating large language model agents with dynamical constraints. First, starting from the Allport-Postman model, Knapp’s rumor classification, and compartmental modeling ideas, this paper explains the foundational role of classical theories in propagation-variable selection and state classification. Second, the modeling significance of the DK model, fractional-order ISDR model, and SKT-type partial differential equation model is analyzed in terms of propagation mechanisms, memory effects, rumor-refutation feedback, spatial diffusion, and spatiotemporal pattern characterization. Furthermore, the roles of topological structures such as scale-free networks, community structures, node centrality, simplicial complexes, and hypergraphs in key spreader identification, collective synergistic propagation, and abrupt diffusion analysis are discussed. Finally, the application of LLM agents coupled with cellular automata and SIR mechanisms in opinion evolution simulation is examined. The results show that rumor propagation modeling is gradually evolving from macro-level empirical explanation toward dynamic mechanism characterization, topological structure constraints, spatiotemporal coupling analysis, and agent-based simulation integration, which can provide theoretical references for public opinion prediction and intervention strategy design in complex network environments.