Effects of Nonsensical Responses in Virtual Human Simulations on Clinicians’ Empathic Communication and Emotional Responses

In this manuscript, we report on research that explores the application of virtual human patients to train clinicians on empathic communication skills. During training, clinicians received empathy scores as they interacted with two virtual humans portraying suicidal ideation, who at times provided nonsensical responses. We video-recorded clinicians' interactions with virtual humans and analyzed their facial expressions, as well as their verbal responses. In phase I of our study, we analyzed clinicians' facial expressions during three key moments: after a sensical response from a virtual human (baseline), following the last nonsensical response of the interaction, and after a sensical response that followed the last nonsensical response. In phase I, facial expressions were grouped into Negative (anger, disgust, sadness, and fear) and Positive (happiness, neutral, and surprise) facial affective behaviors. We observed that nonsensical responses from virtual humans can negatively affect clinicians' positive and negative facial affective behaviors. We found a significant increase in the percentage of time clinicians express negative facial affective behaviors immediately following nonsensical responses. In phase II, we recruited additional clinician-participants and investigated how different proportions of nonsensical responses affect clinicians' facial expressions of individual basic emotions (instead of groups of positive and negative facial expressions), as well as whether nonsensical responses moderate the association between expressions of basic emotions and empathy scores obtained by clinicians during training. We observed a statistically significant positive interaction between proportions of nonsensical responses and angry facial expressions in predicting average empathy scores. That is, the relationship between anger and empathy scores was significant at low and mean levels of nonsensical responses, but not at high levels. These results suggest that at low and mean levels nonsensical responses negatively impact clinicians' performance, hindering their ability to acquire empathy skills. We discuss the impacts of technological limitations during virtual human interactions, particularly nonsensical responses, and the importance of controlling for such issues.