Abstract
Video recordings 151 online dyadic conversations involving individuals playing a deception-based interrogation game were collected for analysis of temporal features involving facial expression. The interrogation game involved two participants (one interrogator and one witness) engaging in an interrogation in which the witness was directed to either lie or tell the truth regarding evidence that was presented to the witness. Analysis identified that interrogators paired with dishonest witnesses demonstrated a higher variance in their smile-related lip corner puller action unit expression than interrogators paired with honest witnesses. We propose to apply HMM-based multimodal models to identify the temporal patters in facial expressions in dyadic communication and hope to discover differences in such patterns between honest and dishonest communication.