Pacing the Cure: A Psychophysiological Analysis of User Responses to GAI Conversational Patterns

While the preliminary exploration of generative AI in aiding psychotherapy has already commenced, there has been no research on the intrinsic selection mechanism on the user side during the process of GAI-assisted psychological healing. To address this gap, we experimented with 60 participants using the fine-tuned large language model, assessing emotional responses through facial expressions, heart rate variability, skin conductance response, and SAM scales. Results showed that after communicating with GAI therapists, the participants’ emotional valence and dominance increased. Participants engaging with GAI therapists who provided slow response times and rich content experienced higher happiness and were more relaxed, while those interacting with faster, brief-content GAI therapists showed greater surprise. Additionally, despite the stress levels initially slightly rising during conversations with GAI therapists, they decrease over time and ultimately lead to an improved emotional state. These findings validate the effectiveness of GAI therapists and provide a basis for applying GAI in different psychotherapy scenarios.