International Journal of Social Robotics manuscript No. (will be inserted by the editor) Eye Movements as a Means To Evaluate and Improve Robots

Abstract With an increase in their capabilities, robots start to play a role in everyday settings. This necessitates a step from a robot-centered (i.e., teaching humans to adapt to robots) to a more human-centered approach (where robots integrate naturally into human activities). Achieving this will increase the effectiveness of robot usage (e.g., shortening the time re-quired for learning), reduce errors, and increase user accep-tance. Robotic camera control will play an important role for a more natural and easier-to-interpret behavior, owing to the central importance of gaze in human communication. This study is intended to provide a first step towards im-proving camera control by a better understanding of human gaze behavior in social situations. To this end, we registered the eye movements of humans watching different types of movies. In all movies, the same two triangles moved around in a self-propelled fashion. However, crucially, some of the movies elicited the attribution of mental states to the tri-angles, while others did not. This permitted us to directly distinguish eye movement patterns relating to the attribu-tion of mental states in (perceived) social situations, from the patterns in non-social situations. We argue that a better understanding of what characterizes human gaze patterns in social situations will help shape robotic behavior, make it more natural for humans to communicate with robots, and establish joint attention (to certain objects) between humans and robots. In addition, a better understanding of huma