Enlarger-Spreader-Controller-Taciturn Rumor Spreading Model Based on Multi-Agent Reinforcement Learning Method

Rumors can skew people's perceptions, impact economic progress, and disrupt social stability. Therefore, studying the principles of how rumors spread is both important and valuable. Most current rumor spreading models are based on infectious disease transmission models. They usually focus on the attributes of diffusion actors and consider rumor propagation as a static process but ignore the stakes behind rumors. In the real world, the beneficiaries of a rumor seek to enhance the impact of the rumor by embellishing or enriching the rumor content, while those with a negative stake in the rumor often seek to reduce the impact of the rumor. As this factor has been overlooked, the current efficiency of controlling the rumor diffusion range and speed is low. To improve the situation, we establish Enlarge-Spreader-Controller-Taciturn rumor spreading model which simulates a dynamic rumor spreading process, considering human preference and the benefit relation between actors and rumor as well as the identities. Additionally, a new forgetting mechanism that is suitable for reinforcement learning is proposed in this paper. Finally, we apply our model to various real-world social networks.