Sharing errors with human and non-human agents

Interpersonal action monitoring, i.e., the ability to monitor other people's actions, is essential during face-to-face interactions. Previous evidence from two independent research lines suggests that both how we represent the interaction goal and the human/non-human nature of the co-actor may affect how we process (and react to) their mistakes. Here, we examined in a full-factorial design whether these two factors modulate how we monitor someone else's errors during minimally joint tasks. During functional magnetic resonance imaging (fMRI), participants interacted with a human or the computer while sharing or not the goal of playing a melody together (shared vs. individual goal conditions). We used implied-motion pictures of a human hand to represent the human partner's responses, while a robotic piston represented the computer's ones. Despite the minimal nature of the interaction, multivariate pattern analysis revealed that it was possible to decode the human/non-human nature of the partner from post-error brain activation patterns but only in the shared goal condition. With both partners, post-error behavioral adaptations in this condition were associated with activation of the pre-supplementary motor area and right anterior insula, brain regions responsible for proactive action control. Goal sharing is thus a powerful factor in boosting interpersonal action monitoring with both human and non-human partners.