Human-AI Collaboration for the Detection of Deceptive Speech

This paper investigates human trust in artificial intelligence (AI) during human-AI collaboration on a speech-based data analytics task. Human users worked together with an explainable AI algorithm that took as an input acoustic and linguistic measures for the detection of deceptive speech. The working performance of the AI was manipulated resulting in a high performing (HP) AI and a low performing (LP) AI. Trust was measured via self-reported and behavioral measures, which were associated with each other. Various personality characteristics, including openness, neuroticism, and extroversion, moderated one's trust in the AI, but results were mixed in terms of the considered self-reported and behavioral trust metrics.