Conversational dynamics describe how rapport, emotion, and goals evolve during a dialogue. In therapeutic settings, these changes are not incidental—they can influence when and how a therapist chooses to intervene. The authors of a new arXiv preprint argue that such dynamics are therefore critical to capturing realistic therapeutic interactions.

To address this, the paper introduces CALM-IT, a method for enhancing the quality of synthetic motivational interviewing dialogues. The approach explicitly incorporates conversational dynamics into the generation process, aiming to produce dialogue that better reflects the ebb and flow of real therapeutic conversations.

By focusing on these dynamics, CALM-IT could help improve synthetic dialogue for motivational interviewing, a technique where the therapist's responsiveness to emotional and relational shifts is central to its effectiveness.