Credibility-driven rumor spreader and debunker co-evolutionary mechanisms for rumor propagation

Abstract Rumor-propagation models have been an active research topic, while few methods consider the dynamic mutual transformation of the rumor spreaders and debunkers during the rumor propagation. To address the problem, we consider the possibility of co-evolution between spreaders and debunkers due to their suspicion of the message’s authenticity. Specifically, we define three dyadic rules that specify the transformation of ignorants to spreaders, ignorants to debunkers, and the mutual transformation between spreaders and debunkers in the competition of spreading-debunking during rumor propagation. Utilizing the proposed dyadic rule, we establish a new SI r I d rumor propagation model (susceptible ignorants ( S ), rumor spreaders ( I r ), rumor debunkers ( I d )). Additionally, we introduce credibility as an indicating factor for the mutual transformation between spreaders and debunkers, and the credibility itself is updated according to the individual influence of each node, the local changes of neighboring spreader and debunker nodes, and the global changes of spreader and debunker nodes in the entire network. We first evaluate the proposed SI r I d model with simulation experiments in three typical networks, namely, Watts-Strogatz (WS), Erdős-Rényi (ER), and Barabasi-Albert (BA) networks. The results show that the proposed SI r I d model is strongly adaptable to these networks. We then conduct a series of parameter and ablation studies to analyze the proposed model theoretically and experimentally. Finally, we evaluate the proposed SI r I d model with multiple real retweet datasets collected from Weibo and Twitter to verify the generality and adaptability of the proposed model. The results show that our SI r I d can mimic rumor propagation in real-world scenarios.