Decoding Deception and Collusion: Behavioral Analysis of Relational Messages and Interpersonal Relationships in Group Communication

Abstract

Collusion is prevalent and costly, particularly in fraud scenarios. Despite extensive research on deception, there are limited studies on collusive deception, where a group of deceivers with hidden objectives works together to undermine a larger group. This dissertation investigates how secretive collaboration and deception influence relational communication within groups. Two experiments were conducted to explore how collusive deception affects deceivers’ behaviors and interactions with other group members. Analysis of recorded conversations revealed that deceivers often adopt “flight” strategies to evade detection, express less trust, and isolate themselves from other deceivers. Successful deception involves deceivers expressing more trust towards truth-tellers as a bonding tactic. Additionally, the increased overall level of distrust within the group gives deceivers an edge to obscure their actions and achieve their agenda. The second experiment analyzed nonverbal behaviors in a mock hiring scenario using automated tools for face and head feature analysis. Results indicated that deceivers exhibited reduced dominance and appeared less nervous emotionally, yet more nervous cognitively during interactions with fellow deceivers. This points to how different levels of appraisals during collusive interactions influence emotions and cognition. This study further develops deception detection models by incorporating network metrics and leveraging Graph Neural Networks (GNNs), which improve performance in deception detection. Additionally, the integration of advanced language technologies like GPT for automated context-level tagging of transcripts enhances scalability and cost-effectiveness.The findings significantly enhance the understanding of how collusive deception influences verbal and nonverbal relational messages, providing crucial insights into deception research. The tested deception detection system demonstrates the potential for developing comprehensive end-to-end solutions with high efficiency and performance in identifying deceptive behavior.