Embodiment matters when establishing eye contact with a robot

Eye contact constitutes a strong social signal in humans and affects various attentional processes. However, eye contact with another human might evoke different responses compared with a direct gaze depicted on a screen. Previous embodied gaze-cueing experiments with iCub humanoid robot showed eye contact modulations on gaze-cueing effect using a long stimulus-onset-asynchrony (SOA: 1000 ms, no predictive gaze cue). Instead, eye contact did not modulate the gaze-cueing effect using a shorter SOA (500 ms, no predictive gaze cue). In the present study, we investigated whether a robot’s eye contact on the screen could modulate the gaze-cueing effect by adapting the previous embodied experiments to a screen-based setup. Specifically, in two experiments we examined the impact of eye contact on the gaze-cueing effect for non-predictive cues while we varied the SOA (500 ms and 1000 ms). Our results showed that the robot’s direct gaze did not modulate the gaze-cueing effect (gaze-cueing effect present in all conditions), thereby suggesting that direct gaze presented in a 2D format on the screen has less impact on observers than its 3D embodied version in a physically present robot. Overall, our findings stress the importance of embodied interactions for understanding the mechanisms of social cognition.