Co-Representation with Social Robots: Evidence from the Social Inhibition of Return effect

As artificial intelligence continues to advance, people increasingly expect humanoid robots to demonstrate sophisticated social behaviours. To meet these expectations, it is crucial to design robots that align with human models of social cognition. This study explores the extent to which humans co-represent actions performed by social robots and how prior attitudes influence this process. We use the social inhibition of return (SIOR) paradigm, where individuals are slower to respond to targets in locations previously acted upon by themselves or by a partner. In a within-subjects design, participants engaged first with a robot co-actor and then with another human co-actor. Overall, our findings suggest that interacting with a robot fosters positive perceptions of the robot and evokes co-representation of their actions. Additionally, we found a trade-off between self- and social-inhibitory processes when interacting with a robot, in contrast to interactions with humans. Furthermore, prior attitudes, particularly the tendency to attribute human-like qualities to robots, were paradoxically found to reduce the SIOR effect. We discuss these results in relation to the uncanny valley effect, as well as broader implications for how human and robot factors that drive people’s co-representation of robot behavior may support more natural and effective human—robot cooperation.