Exploring Perceptions of Robot Facial Expressions in Service Interactions

Abstract

The problem of emotional alignment in human-robot interaction remains underexplored in real-time conversational contexts. In this paper, we investigate the role of facial expressions in voice-based interactions using the Ameca humanoid robot, comparing conditions where facial expressions are appropriate or inappropriate. In a user study (N = 28), participants interacted with the robot in a simulated ticket-purchasing scenario. Our findings reveal that appropriate facial expressions significantly enhanced pragmatic aspects of the interaction compared to inappropriate expressions, which, although sometimes confusing, could still provoke deeper engagement. Participants noted that the interaction flow felt unnatural as facial expressions could only be rendered after the lip-synced speech, rather than simultaneously. These results highlight the importance of emotional congruence in improving the naturalness of human-robot interaction. This research informs the design and development of emotionally intelligent robots that can better align their non-verbal cues with conversational contexts.