Towards detecting deceptive intentions on a large scale

Assessing the credibility of verbal statements about someone's intended behavior is becoming increasingly important. At the same time, contexts such as terrorism prevention require large numbers of people to be screened to find the needle in the haystack. This thesis focused on the detection of deceptive intentions on a large scale and identified the verbal approach to deception detection as the most promising candidate to meet large-scale requirements. This work aimed to address three aspects with two empirical chapters each to move the field towards detecting deceptive intentions on a large scale. First, the existing verbal deception detection approach was critically evaluated. A replication experiment, as well as a simulation study, concluded that the field would benefit from larger sample sizes and more replication efforts to establish a reliable empirical base. Second, to bridge the gap between theory and large-scale applications, hybrid approaches were introduced to replace the manual annotation with automated methods. Two studies showed the promise of using theoretically-informed automated methods but also highlighted the role of context in deception detection. Third, two studies looked at identifying truths and lies about future behavior in an automated data collection and data analysis workflow. Truth-tellers and liars were identified above the chance level. This thesis concludes with implications of this work for the field, a discussion of the limitations, and provides an outlook on the future. Overall, this work provided a starting point for research into systems that are applicable on a large scale.