Gatekeeping Behavior Analysis for Information Credibility Assessment on Weibo

Microblogging sites, such as Sina Weibo and Twitter, have gained significantly in popularity and become an important source for real-time information dissemination. Inevitably, these services are also used to spread false rumors and misinformation, usually with the unintentional collaboration from innocent users. Previous studies show that microblog information credibility can be assessed automatically based on the features extracted from message contents and users. In this paper, we address this problem from a new perspective by exploring the human input in the propagation process of popular microblog posts. Specifically, we consider that the users are the gatekeepers of their own media portal on microblogging sites, as they decide which information is filtered for dissemination to their followers. We find that truthful posts and false rumors exhibit distinguishable patterns in terms of which gatekeepers forward them and what the gatekeepers comment on them. Based on this finding, we propose to assess the information credibility of popular microblog posts with Hidden Markov Models (HMMs) of gatekeeping behavior. The proposed approach is evaluated using a real life data set that consists of over ten thousand popular posts collected from Sina Weibo.