Detecting Deception in Computer Mediated Communication: A Social Structural Perspective

Despite the widespread use of Computer-Mediated Communication (CMC) for effective collaboration and interaction, CMC has become a growing hotbed for deception due to its provision of ubiquity, anonymity and open environment. Deception is an increasing threat to our society and to the daily communication of both individuals and groups. This dissertation aims to provide a new venue for understanding deception and for detecting deception through the identification, extraction, and application of social structural behaviors of deceptive communication. To this end, the dissertation consists of three major studies. The first study conceptualizes deception in terms of social structure by drawing on interpersonal deception theory and social network theories and proposes a research model of structural properties of deceptive communication: centrality, cohesion and similarity. Viewed from the social structural perspective, structural behaviors are denoted as the relationships between different individuals (entities) or as relatively stable patterns of relationships. The second study examines the impact of time on structural behaviors of deception based on interactional adaption theory and characteristics of temporal networks. The third study addresses the problem of automatic deception detection by extracting the structural features of deceptive communication and by combining the structural features with linguistic features. In addition, it evaluates the generality of the structural features identified from synchronous CMC to asynchronous CMC. The findings of this dissertation extend existing theories and research on explaining the effect of deception intent on structural behaviors of communication in multi-fold aspects. First, this dissertation extends the context of deception theories from interpersonal interaction to social interaction by addressing the interactional dynamics in group communication that is composed of one deceiver and multiple receivers. Second, the dissertation operationalizes the structural behaviors of deceptive CMC and extracts them from two different types of networks: static and temporal, and empirically validates the behaviors with real-world data. Third, the dissertation improves the performance of automated deception detection by incorporating the structural behaviors of deception