Mimicking a robot: Facial EMG in response to emotional robotic facial expressions

Abstract

Humans tend to anthropomorphize i.e., to attribute human-like characteristics (e.g. motivations, intentions, emotions) to non-humans. This suggests that we can interact with non-humans (televisions, computers, robots) in a similar way we interact with humans. Robots, in particular, have physical presence and can be programmed to display social interaction capabilities, i.e. to be social robots, amplifying those similarities. Past studies have shown that social robots in negative situations tend to elicit strong emotional responses and empathy in humans. However, it remains to be tested whether empathy can be felt towards a social robot, set in a situation of positive social interaction. We proposed that facial mimicry, one indicator of empathy, may occur towards a robot in a positive social context, i.e. while the robot is playing a board game with human opponents. Fifty-nine participants (46 females), aged 17 to 27 years (M=19.56, SD=2.11) were exposed to videos of a robotic head (EMYS, the EMotive headY System), previously programmed to display six emotional expressions (joy, surprise, anger, disgust, fear, sadness) and a neutral expression, while playing a board game. EMYS’s facial expressions were shown in two blocks: in the first, no social context was provided and sound was omitted; in the second, a positive social context was provided, which included sound of verbal interaction with humans. In each block, 14 videos were randomly presented. Facial electromyography (fEMG) activity, in response to EMYS’s facial expressions, was measured over the corrugator supercilii and zygomaticus major muscles. fEMG responses were calculated as difference from stimulus presentation to 1 sec baseline. Changes in fEMG reactivity, between conditions, were analyzed comparing fEMG responses to robotic emotional expressions with responses to robotic neutral expressions. In the positive social context condition, results revealed an overall reduction of corrugator supercilii reactivity for the majority of negative emotional expressions (except anger). There was also a significant reduction of the zygomaticus major activity to surprise, compared to neutral, in the positive social context. Overall, our results suggest the important role of the social context in our physiological responses to a robot, and more specifically a reduction of emotional negativity to non-threatening robotic facial expressions, displayed in a positive social context.