Personality Classification from Robot-mediated Communication Cues
Robotic telepresence offers a convenient substitute for face-to-face communication by enabling individuals to engage in communication regardless of location, and in various scenarios such as remote education, business meetings and elderly care. In robot-mediated communication, a teleoperator conveys nonverbal communication cues such as head movements, hand gestures, body posture along with audio cues through a robot. Correctly interpreting these nonverbal cues plays an important role in forming impressions, understanding social behaviours and achieving an effective communication. However, it is difficult to form a holistic understanding of how these nonverbal communication cues are interpreted by the interlocutor along with the robot’s appearance , and how the perceptions regarding the teleoperator change as compared to communication in-person. This paper focuses on automatic personality classification from nonverbal communication cues in a telepresence context and provides a comparison with respect to in-person communication condition as illustrated in Fig. 1. We extract a rich set of features and learn the relationship between the extracted features and the personality assessments by training automatic classifiers. Our results show that personality classification from robot-mediated communication cues works better than from audio-only cues except for agreeableness. Facial activity and head pose together with audio and arm gestures play an important role in conveying extroversion and agreeableness.