Detecting Deceptive Groups Using Conversations and Network Analysis

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 857–866, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics Detecting Deceptive Groups Using Conversations and Network Analysis Dian Yu1, Yulia Tyshchuk2, Heng Ji1, William Wallace2 1Computer Science Department, Rensselaer Polytechnic Institute 2Department of Industrial and Systems Engineering, Rensselaer Polytechnic Institute 1,2{yud2,tyshcy,jih,wallaw}@rpi.edu Abstract Deception detection has been formulated as a supervised binary classification prob- lem on single documents. However, in daily life, millions of fraud cases involve detailed conversations between deceivers and victims. Deceivers may dynamically adjust their deceptive statements accord- ing to the reactions of victims. In addition, people may form groups and collaborate to deceive others. In this paper, we seek to identify deceptive groups from their con- versations. We propose a novel subgroup detection method that combines linguis- tic signals and signed network analysis for dynamic clustering. A social-elimination game called Killer Game is introduced as a case study1. Experimental results demon- strate that our approach significantly out- performs human voting and state-of-the- art subgroup detection methods at dynam- ically differentiating the deceptive groups from truth-tellers. 1 Introduction Deception generally entails messages and infor- mation intentionally transmitted to create a false conclusion (Buller et al., 1994). Deception detec- tion is an important task for a wide range of ap- plications including law enforcement, intelligence gathering, and financial fraud. Most of the previ- ous work (e.g., (Ott et al., 2011; Feng et al., 2012)) focused on content analysis of a single document in isolation (e.g., a product review). The promot- ers of a product may post fake complimentary re- views, while their competitors may hire people to write fake negative reviews (Ott et al., 2011). 1The data set is publicly available for research purposes at: http://nlp.cs.rpi.edu/data/killer.zip However, when we want to detect deception from text or voice conversations, the deception be- havior may be affected by the following factors be- yond textual statements. 1. Dynamic. Recent research in social science suggests that deception communication is dy- namic and involves interactions among peo- ple (e.g., (Buller and Burgoon, 1996)). Addi- tionally, the research postulates that human’s capacity to learn by observation enables him to acquire large, integrated units of behav- ior by example (Bandura, 1971). Therefore, a person’s behavior concerning deception or truth-telling can change constantly, while he learns from others’ statements during conver- sations. 2. Global. People may form groups for purpose of deception. Research in social psychology has shown that an individual’s object-related behavior may be affected by the attitudes of other people due to group dynamics (Fried- kin, 2010). Recent studies typically have been conducted over “static” written or oral deceptive statements. There is no obligatory requirement for communi- cation between the author and the readers of these statements (Yancheva and Rudzicz, 2013). As a result, a victim of deception tends to trust the sto- ry mainly based on the statement he reads (Ott et al., 2011). However, in daily life, millions …

Content truncated to 500 words