Effects of embodiment and gestures on social interaction in drumming games with a humanoid robot

We present results from an empirical study investigating the effect of embodiment and minimal gestures in an interactive drumming game consisting of an autonomous child-sized humanoid robot (KASPAR) playing with child participants. In this study, each participant played three games with a humanoid robot that played a drum whilst simultaneously making (or not making) head gestures. The three games included the participant interacting with the real robot (physical embodiment condition), interacting with a hidden robot when only the sound of the robot is heard (disembodiment condition; note that the term 'disembodiment' is used in this paper specifically to refer to an experimental condition where a physical robot produces the sound cues, but is not visible to the participants), or interacting with a real-time image of the robot (virtual embodiment condition). We used a mixed design where repeated measures were used to evaluate embodiment effects and independent-groups measures were used to study the gestures effects. Data from the implementation of a human-robot interaction experiment with 66 children are presented, and statistically analyzed in terms of participants' subjective experiences and drumming performance of the human-robot pair. The subjective experiences showed significant differences for the different embodiment conditions when gestures were used in terms of enjoyment of the game, and perceived intelligence and appearance of the robot. The drumming performance also differed significantly within the embodiment conditions and the presence of gestures increased these differences significantly. The presence of a physical, embodied robot enabled more interaction, better drumming and turn-taking, as well as enjoyment of the interaction, especially when the robot used gestures.