Features of Computer-Mediated, Text-Based Messages that Support Automatable, Linguistics-Based Indicators for Deception Detection

Abstract

ABSTRACT: Researchers are assessing the viability of automating deception detection in TAC (text-based, asynchronous, computer-mediated) messages. Viability is conditional on whether deceivers incorporate different features in their TAC messages than truth tellers do and on whether these differentiating features support deception indicators suitable for automation. We propose that deceivers incorporate different features in their messages to realize impression-management goals. They avoid message features people associate with lying (defensive-targeted features) and include message features people associate with truth telling (promotive-targeted features). In a series of three studies, we assessed features that may be part of deceivers' impression-management strategies and then assessed whether deceivers' messages contain different levels of these features than truth tellers' messages. The results show that deceivers construct different messages than truth tellers do and that the differences relate to promotive-targeted features. These findings suggest that researchers look to promotive goals of deceivers for automatable deception indicators in TAC communication.