Managing Turn-Taking in Human-Robot Interactions

Abstract

This study deals with turn-taking in human-robot interactions (HRI). Based on 15 sessions of video-recorded interactions between pairs of human participants and a social robot called Furhat, we explore how human participants orient to violations of the normative order of turn-taking in social interaction and how they handle those violations. As a case in point, we present sequences of HRI to show particular features of turn-taking with the robot and also how the robot may fail to respond to the human participants’ bid to take a turn. In these sequences, the participants either complete the turn in progress and ignore the overlap caused by the robot’s continuation of its turn, or they cut short their own turn and restart in the next possible turn-transition place. In all cases in our data, the overlaps and failed smooth turn-transitions are oriented to as accountable and in some sense interactionally problematic. The results of the study point not only to improvables in robot engineering, but also to routine practices of projection and the ways in which human subjects orient toward normative expectations of ordinary social interactions, even when conversing with a robot.