RBE PhD Qualifier Presentation
A listening robot: How eye contact affects human-robot interaction
Abstract: In an effective conversation, the speaker uses gaze cues to establish the collaboration while the listener responds with gaze cues. In this research, the authors study how a robot, as a listener, can express that it is paying attention to the conversational partners using these cues. In the first part of the study, we implemented an experimental scenario to measure the gaze cue effect (GCE) through the interaction between human and Reeti, a robot capable of facial and eye expressions. Reeti could move its eyes and head to either establish eye contact or not with a human based on algorithm detecting position of the human eyes. In the controlled laboratory experiment with 22 test subjects in more than 800 valid trials, we have found that GCE is elicited only when the eye contact is established. The measurement and questionnaire results have supported our hypothesis that eye contact presents an essential difference in human-robot interaction, similar to the impact of eye contact between humans in a conversation. The second part of the experiment consisted of conversation scenarios, in which Reeti switched between 3 behavioral modes during the conversation. We aim to study whether our test subjects could distinguish changes in Reeti's attitude. The preliminary results have suggested that the robot’s gaze shifts could influence human speakers’ behaviors during the conversation.
Professor Mike A. Gennert
Professor Jeanine LM Skorinko
Professor Cagdas Denizel Onal (Advisor)