Investigating the real world impact of emotion portrayal through robot voice and motion

Abstract

In this paper we investigate robot to human Interpersonal Emotion Transfer (IET) in a real world contextualised human-robot interaction (HRI). IET is an umbrella term which describes the impact of emotions in human-human interaction (HHI). This includes emotion contagion and social appraisal effects. These effects are particularly relevant in domains such as teaching, sports, exercise and healthy eating; domains increasingly targeted by socially assistive robotics. As such, we suggest socially assistive robots may benefit from affective communication in the same way as their human counterparts. We show that emotion recognition from robot voice and motion is possible in explicit validation experiments but does not hold in a socially assistive interaction. Our findings suggest that robot to human IET relies on the human having an expectation for, and hence recognition of, robot emotions; mimicry of valanced motion is not sufficient.