Computational Measures of Deceptive Language: Prospects and Issues

Abstract

In this article, we wish to foster a dialogue between theory-based and classification-oriented stylometric approaches regarding deception detection. To do so, we review how cue-based and model-based stylometric systems are used to detect deceit. Baseline methods, common cues, recent methods, and field studies are presented. After reviewing how computational stylometric tools have been used for deception detection purposes, we show that the he stylometric methods and tools cannot be applied to deception detection problems on the field in their current state. We then identify important advantages and issues of stylometric tools. Advantages encompass quickness of extraction and robustness, allowing for best interviewing practices. Issues are discussed in terms of oral data transcription issues and automation bias emergence. We finally establish future research proposals: We emphasize the importance of baseline assessment and the need for transcription methods, and the concern of ethical standards regarding the applicability of stylometry for deception detection purposes in practical settings, while encouraging the cooperation between linguists, psychologists, engineers, and practitioners requiring deception detection methods.