Empathic chatbot replies made people see it as more able to feel, an experiment of 716 finds

In the social isolation scenario, empathic expression increased the chatbot's perceived capacity to feel, and that perception was positively associated with attitudes toward its recommendations. The pattern differed in the social anxiety scenario, where the effect of empathy depended on how narrative the chatbot's message was.

The authors say the results offer insights into how AI systems can communicate more effectively in mental health care, though the measured outcomes are perceptions and attitudes rather than health results.

The study examined an AI chatbot in a mental health care setting. Researchers varied whether the chatbot displayed empathic cues and how narrative its messages were, then measured how participants perceived its mental capacities and how they viewed its recommendations.

In the social isolation condition, the pathway was straightforward: empathy increased perceived capacity to feel, and that perception was positively associated with attitudes toward the recommendations.

In the social anxiety condition, results were conditional. Empathy increased perceived capacity to feel only for low-narrativity messages, while narrativity itself increased perceived capacity to feel when the chatbot did not express empathy.