University CopyCat Game: Observational Learning with Robot Tutee

Abstract

The presented study investigates gesture imitation in human-robot interaction (HRI), employing EMAH robotic system, on an anthropomorphic social robot Ameca, and Google’s MediaPipe Holistic framework. This integration of human pose, face, and hand detection aims to advance the interaction capabilities of social robots. Of the 12 gestures implemented, 5 are ultimately utilized for examining the impact of gesture mimicry, tutor gender, previous robot experience, and feedback strategies on learning within HRI. A within-subject lab study was conducted in a university setting, with 15 participants (9 males, 6 females) engaged in a game-based scenario with EMAH. Results highlight a generally positive perception of EMAH’s mimicry for ‘curious’ and ‘unsure’ gestures, identifying a need for improvement in ‘backoff’ gesture mimicry. Gender analysis revealed female tutors perceived EMAH’s performance more uniformly positive. Moreover, tutors that displayed greater willingness to reteach indicated that having no prior experience with robots influences engagement levels. Feedback strategies, including verbal feedback, facial expressions, and encouragement, emphasize the importance of communication in educational HRI scenarios. This research contributes to the field of HRI by demonstrating the potential development of socially intelligent robots capable of learning from human instruction within educational settings.