Keystroke Patterns as Prosody in Digital Writings: A Case Study with Deceptive Reviews and Essays

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1469–1473, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Keystroke Patterns as Prosody in Digital Writings: A Case Study with Deceptive Reviews and Essays Ritwik Banerjee Song Feng Jun S. Kang Computer Science Stony Brook University {rbanerjee, songfeng, junkang} @cs.stonybrook.edu Yejin Choi Computer Science & Engineering University of Washington yejin@cs.washington.edu Abstract In this paper, we explore the use of keyboard strokes as a means to access the real-time writ- ing process of online authors, analogously to prosody in speech analysis, in the context of deception detection. We show that differences in keystroke patterns like editing maneuvers and duration of pauses can help distinguish be- tween truthful and deceptive writing. Empiri- cal results show that incorporating keystroke- based features lead to improved performance in deception detection in two different do- mains: online reviews and essays. 1 Introduction Due to the practical importance of detecting deceit, in- terest in it is ancient, appearing in papyrus dated back to 900 B.C. (Trovillo, 1939). In more recent years, sev- eral studies have shown that the deceiver often exhibits behavior that belies the content of communication, thus providing cues of deception to an observer. These in- clude linguistic (e.g., Newman et al. (2003), Hancock et al. (2004)) as well as paralinguistic (e.g., Ekman et al. (1991), DePaulo et al. (2003)) cues. Recognizing deception, however, remains a hard task for humans, who perform only marginally better than chance (Bond and DePaulo, 2006; Ott et al., 2011). Recent studies suggest that computers can be sur- prisingly effective in this task, albeit in limited domains such as product reviews. Prior research has employed lexico-syntactic patterns (Ott et al., 2011; Feng et al., 2012) as well as online user behavior (Fei et al., 2013; Mukherjee et al., 2013). In this paper, we study the effect of keystroke patterns for deception detection in digital communications, which might be helpful in un- derstanding the psychology of deception and help to- ward trustful online communities. This allows us to in- vestigate differences in the writing and revisional pro- cesses of truthful and fake writers. Our work thus shares intuition with HCI research linking keystroke analysis to cognitive processes (Vizer et al., 2009; Epp et al., 2011) and psychology research connecting cog- nitive differences to deception (Ekman, 2003; Vrij et al., 2006). Recent research has shown that lying generally im- poses a cognitive burden (e.g., McCornack (1997), Vrij et al. (2006)) which increases in real-time scenar- ios (Ekman, 2003). Cognitive burden has been known to produce differences in keytroke features (Vizer et al., 2009; Epp et al., 2011). Previous research has not, however, directly investigated any quantitative connec- tion between keystroke patterns and deceptive writing. In this paper, we posit that cognitive burdens in deception may lead to measurable characteristics in keystroke patterns. Our contributions are as follows: (1) introducing keystroke logs as an extended linguis- tic signal capturing the real-time writing process (anal- ogous to prosody …

Content truncated to 500 words