Sensors in the wild: Exploring electrodermal activity in child-robot interaction

Abstract

Recent advances in biosensor technology enabled the appearance of commercial wireless sensors that can measure electrodermal activity (EDA) in user's everyday settings. In this paper, we investigate the potential benefits of measuring EDA to better understand children-robot interaction in two distinct directions: to characterize and evaluate the interaction, and to dynamically recognize user's affective states. To do so, we present a study in which 38 children interacted with an iCat robot while wearing a wireless sensor that measured their electrodermal activity. We found that different patterns of electrodermal variation emerge for different supportive behaviours elicited by the robot and for different affective states of the children. The results also yield significant correlations between statistical features extracted from the signal and surveyed parameters regarding how children perceived the interaction and their affective state.