Dynamic and Dyadic Relationships between Facial Behavior, Working Alliance, and Treatment Outcomes during Depression Therapy

Objective: Previous work has yielded mixed results regarding the relationships of interpersonal synchrony with working alliance and treatment outcomes. We sought to clarify these relationships by applying more sophisticated dynamic and dyadic methods.Method: Adult outpatients with depression (N=65, Age 18--64, 65\% Female, 68\% White) participated in up to eight sessions of cognitive-behavioral or interpersonal psychotherapy. Sessions occurred either in-person or via teletherapy and were video-recorded. Facial computing tools estimated the momentary intensity of patients' and therapists' scowling and smiling, and the interpersonal synchrony of these measures during each session was quantified using windowed cross-correlation analyses. Hypotheses about within-dyad processes (i.e., session-to-session changes) and between-dyad associations (i.e., of average tendencies) were tested using hierarchical Bayesian mediation models.Results: Numerous and nuanced associations were found between facial behavior, working alliance, and treatment outcomes (e.g., session-to-session increases in patient-rated working alliance were predicted by higher-than-average therapist smiling). Interpersonal synchrony did not predict within-dyad changes in working alliance or depression symptoms; however, on the between-dyad level, higher average patient-rated working alliance was predicted by both higher average smile synchrony and lower average scowl synchrony.Conclusions: The facial behavior of patients and therapists provides a valuable window into the therapeutic process, especially with regard to the alliance. Contributions come from the behaviors of each person as well as their level of interpersonal synchrony. However, contrary to extant theories, the associations of synchrony with working alliance and outcomes depend on the behavior being synchronized, the informant providing ratings, and the level of analysis being conducted.Public Health Significance: These results help refine clinical theories about psychotherapy and unlock new possibilities for computationally assessing its progress.