Embodied social robots trigger gaze following in real-time

摘要

In human-human interaction, we use information from gestures, facial expressions and gaze direction to make inferences about what interaction partners think, feel or intend to do next. Observing changes in gaze direction triggers shifts of attention to gazed-at locations and helps establish shared attention between gazer and observer - a prerequisite for more complex social skills like mentalizing, action understanding and joint action. The ability to follow others’ gaze develops early in life and being able to process gaze signals is a crucial milestone in human development. While human gaze signals are so essential for social interactions that we automatically follow them, it is unclear whether robot gaze cues are followed to similar degrees, and whether they have the ability to establish shared attention between human and robot. Furthermore, most studies on social attention in human-robot-interaction (HRI) use robot images and videos in controlled laboratory settings, which makes it necessary to determine whether gaze following can also be observed in social interactions with embodied robot platforms in real- time. In the current experiment, we use the humanoid robot Meka to examine whether gaze following can be induced in realistic interactions with social robots. The results indicate that Meka’s gaze cues were reliably followed, and that they were able to establish shared attention in HRI. Implications of this finding for social robotics are discussed.