Advanced AI to capture the multiscale nature of motor attunement within the patient-psychiatrist relationship

Abstract:

Interventions in medicine are influenced by the quality of clinical rapport. This is particularly true in psychiatry where such relationship strongly impacts patient engagement, adherence to medications, symptom relief, hospitalizations and healthcare costs. Too little systematic attention has been devoted to understanding the implicit, early aspects of the therapeutic relationship: they are often dismissed as a “non-specific” factor, implying they are scarcely susceptible to change and represent unavoidable bias (1).
This project focuses on the embodied interpersonal and intersubjective aspects of the therapeutic relationship, including not only cognitive/content communication but, first and foremost, emotional, motor and overall nonverbal and sensorimotor communication. These concepts can be merged into the notion of “embodied attunement” (2,3). Achieving attunement in psychiatry is complicated by impairment of illness awareness, communication and motor impairments associated with mental disorders (4–7). Our working hypothesis is that the development of optimal therapeutic attunement is signalled by, and require an adequate level of motor synchronization and coordination between the patient and the clinician. These dimensions modulate interpersonal feelings, promote better socio-affective affiliation and, ultimately, patient recovery (8,9).
However, obtaining a reliable, objective assessment of non-verbal attunement within a therapeutic relationship is challenging. Human movements are hierarchically nested, across multiple temporal scales, and every layer influence and is influenced by the others (10,11). Nonetheless, decoding such hierarchical organization allows to encompass the expressive and emotional movement qualities (e.g., hesitation, aggressiveness, fragility) that are needed to resonate with an interlocutor when the interaction is psychologically pregnant. Thus, a multiscale motor quantification of attunement may be necessary to understand if, how, when and why the dyad has developed an effective therapeutic alliance.
This project will exploit the notable advances of technologies to track human motion as well as state of the art AI-based methods for human movement analysis (12). We will merge psychiatric expertise with a solid neuroscientific background and advanced computational models of motor coordination to derive indices of attunement, and will use them to improve the prediction of clinical outcomes using machine learning methods. In sum, the project will seek to identify the motor “fingerprint” of the emergent patient-doctor alliance. This study shall enhance our scientific understanding, as well as provide actionable evidence to develop didactic, clinical and technological initiatives to improve mental healthcare.