Research on False Information Detection Based on Herd Behavior From a Social Network Perspective

Abstract

As social networks become ubiquitous, the rapid dissemination of false information poses a substantial threat to societal stability and public welfare. Although sociological and psychological studies have confirmed the significant role of herd behavior in the spread of false information, traditional detection methods struggle to address the dual challenges posed by decentralized communication modes and artificial intelligence-generated content, as they often overlook the psychological mechanisms at play within groups. This study proposes a multidimensional false information detection model, termed HBD-Net, based on herd behavior, to explore innovative methods for false information detection through the lens of herd behavior propagation mechanisms in social networks. By integrating multidimensional information such as the influence of opinion leaders, popular comments, and friends’ experiences, we construct a robust false information detection model. Experimental results demonstrate its superior performance on both the PolitiFact and GossipCop datasets, particularly excelling on the GossipCop dataset with an accuracy of 93.11%, significantly outperforming other baseline models.