Who believes, who deceives, who defends: assessing online social network health and mitigating false information diffusion.

Abstract

The widespread reliance on social networking sites for news consumption has significantly contributed to the proliferation of false information, which can have serious societal consequences. This dissertation establishes a framework for combating false information by evaluating network health and designing targeted intervention strategies. First, we identify three categories of actors: naive false information spreaders, deliberate false information agents, and those who actively refute false information. Understanding these roles allows for targeted interventions—educating the first group, inhibiting the second, and empowering the third. We further assess the vulnerability of individuals and communities to false information, identifying those at greatest risk, and thus most in need of corrective information. Building on this, we propose a false information debunking technique that identifies an optimal set of refutation spreaders who can actively disseminate corrections to counteract the influence of false information spreaders. At the early stages of false information diffusion, these individuals play a critical role in neutralizing its impact and protecting vulnerable users. Together, the network health assessment and the strategies introduced in this work form a comprehensive framework for mitigating the spread of false information on social networks. In constructing the framework, we have incorporated three key components into our models: user behavior, trust, and network structure. Analyzing users' behavior, such as their history of posts, can offer valuable insights for developing effective strategies to combat false information. Beyond content sharing, social media users build connections through friendships and followers, forming networks of trust and influence. Messages from trusted sources are generally more persuasive. Thus, trust can play a pivotal role in both mitigating the spread of false information and amplifying accurate information. As users form connections, they become embedded within a larger network structure, which in itself contains rich information about user interactions and community dynamics. Given the valuable insights provided by user behavior and network structure, and the critical role of trust in shaping beliefs and actions, these components were incorporated into the development of our false information mitigation framework.