Causing a robot to display happy but not sad facial expression leads to sensory attenuation.

Sensory events resulting from our own actions typically evoke weaker neural responses than externally generated events, a phenomenon known as sensory attenuation. Whether this attenuation also occurs in social contexts-when the outcome of one's action is a reaction of an interaction partner, such as an emotional expression, remains unclear. This EEG study investigated whether causing a humanoid robot (iCub) to display emotional facial expressions leads to sensory attenuation in early visual processing, and whether emotional valence modulates this effect. Participants either chose an object to show to iCub, after which the robot displayed a happy or sad expression, or observed the same expressions triggered by the computer. We found reduced P1 amplitudes for self-generated happy expressions, reflecting sensory attenuation, whereas P1 responses to self-generated sad expressions did not differ from computer-generated sad expressions. Furthermore, higher empathy ratings toward the robot were associated with reduced P1 attenuation for sad expressions. Later components were also influenced by emotional valence (N170, P2, P3) and by whether the expressions were self- or computer-generated (N170, P3), without interaction. Our findings suggest that sensory attenuation can be modulated by the affective relevance of one's actions in social interactions, including interactions with artificial agents.