Psychophysiological and Socio-emotional features of Human-Agent Interaction in presence or absence of Depressive Symptoms.

The present work was conducted with the aim of shedding light on the psychophysiological and socio-emotional features of human-agent interaction in presence or absence of depressive symptoms, focusing on four main goals. The first goal was to deepen the literature on the use of electroencephalography (EEG) to assess users’ acceptance and preferences toward robots and virtual agents. Results of the systematic review highlighted that while EEG measures have been used to evaluate user preferences related to the aesthetic features of robots, these measures, and particularly Frontal Alpha Asymmetry (FAA), as indicators of acceptance and preference in human-agent interaction represent an untapped yet promising field. Accordingly, the second goal was to explore the relationship between FAA and willingness to interact with virtual agents, trying to provide insights on the affective and motivational states of users, with and without depressive symptoms, when interacting with virtual agents. Results of the first study indicated a significant effect of agents’ likability on FAA values, as well as differences between participants with and without depressive symptoms. Additional analyses conducted on a larger dataset did not replicate the results concerning agents’ likability observed in the previous study. However, the previously observed asymmetry pattern related to the presence or absence of depressive symptoms was partially confirmed. Thus, further exploration is needed. The third goal was to explore whether and how certain aspects of depressive emotional and social functioning (i.e., impaired processing of emotional stimuli, social avoidance and withdrawal and preference for mood-congruent others) are associated with a different attitude toward virtual agents of different emotional valence and gender, when compared to individuals without such symptomatology. Results showed that depressive symptoms per se did not influence directly participants’ willingness to interact with virtual agents, whereas gender and emotional valence did. Indeed, participants both with and without depressive symptoms exhibited a preference for female agents and for agents expressing happiness. Finally, the fourth goal was to further investigate the social cognition of individuals with depressive symptoms in human-agent interaction, by exploring their ability to recognize emotions expressed by virtual agents. Results showed an impairment in synthetic emotional expressions recognition in participants with depressive symptoms, while the hypothesized negative response bias towards sadness within this group was not confirmed.