Towards the Evaluation of the Role of Embodiment in Emotions Elicitation

In the human-robot interaction (HRI) literature, there is a focus on the ability of a robotic system to display emotional expressions to make them easily recognizable by the users. However, only few studies addressed how the multimodal behaviours of a robot may cause a specific emotional reaction in a user. In this work, we deploy a late fusion model for emotion recognition using facial and bio-signal inputs, and we use such a model in a user study aiming at evaluating the role of embodied interaction in emotion elicitation. To this extent, we evaluate the ability of a robot to change the elicited emotion with respect to the one elicited by a standardized stimulus (e.g., a video). Initial results show that the robot’s behaviour is effective in changing the valence of the perceived emotion, but not the arousal that is lowered.