Minimization of the spread of acute rumors in social networks

Social networks offer the capability to spread inspiring ideas and adapt innovations at unprecedented speed and convenience. Although such an information-spreading capability is invaluable, social networks can also rapidly disseminate misinformation to a large number of people, with dire consequences. In this study, we focus on the spread of a specific type of misinformation: acute rumors, which we define as misinformation with the potential to spread quickly in the network and mobilize individuals to take harmful actions that may go beyond social platforms. We present a new diffusion model for acute rumors, which extends the classical linear threshold model by considering the base idea of the individuals regarding the topic of the rumor and the emotional impact its content creates on them. Based on this diffusion model, we propose a new centrality measure that can accurately detect individuals with a high potential to enhance the spread of acute rumors. In a comprehensive numerical study, we evaluate the performance of our proposed centrality measure by comparing it to existing traditional centrality measures in the literature. Our results attest to the superior performance of our centrality measure not only in terms of detecting the individuals with the highest potential to increase the reach of acute rumor but also in finding the correct ranking among the individuals in the network regarding their potential to contribute significantly to the dissemination of acute rumors. While the proposed diffusion model and centrality measure offer valuable insights into who is likely to spread an acute rumor, they do not fully capture how platform-specific dynamics and individual-level behaviors shape the overall diffusion process. To address the limitations of the diffusion model in capturing message variability and algorithmically shaped user behavior, we develop an agent-based model (ABM) inspired by the structure of Twitter/X. The model is calibrated using empirical data and incorporates multiple forms of user engagement. Within this framework, we evaluate several control strategies intended to reduce the spread of acute rumors while maintaining user interaction on the platform and avoiding real-world physical actions triggered by rumor diffusion. Simulation results show that the targeted intervention, which adjusts the reading order for a selected subset of users with strong or polarized opinions identified through the proposed centrality measure, outperforms all other strategies and achieves the most effective balance between limiting rumor diffusion, preventing physical actions, and maintaining user interaction on the platform.