Dynamic harmony: Unveiling therapeutic attunement in emotionally focused couples therapy via machine learning

Abstract Objective The goals of the study were to examine therapists' and clients' emotional states and expressions in an emotionally focused therapy (EFT) couple session, to assess therapeutic attunement between the clients and the therapist, and to explore its alignment with EFT techniques. Background Therapeutic attunement is crucial for fostering a therapeutic alliance in couples therapy, yet examining triadic relationships between therapist and partners is methodologically challenging. This case study introduces a novel computational social science approach to capture attunement in an EFT session. Method A full‐length, publicly available EFT session video was analyzed. We generated text, audio, and image data for computerized tracking and conducted a multimodal analysis of emotions using mixture of experts machine learning models. Results Seven emotion states were analyzed: anger, fear, surprise, disgust, joy, sadness, and neutral. The results indicated a close alignment between the couple and the therapist's emotions, suggesting high attunement. Three types of attunement by timing were identified: on time, therapist initiated, and delayed. Attunement peaks aligned with EFT techniques. Conclusion High levels of therapeutic attunement, facilitated by EFT techniques, can be effectively captured and analyzed using machine learning. Implications This study highlights the feasibility of using machine learning to track attunement dynamics and aids therapists in exploring therapeutic ruptures.