Empathy for Everyone: Generative AI that Improves Patient-Provider Cultural Attunement in Real Time

Abstract It is estimated that more than one in five U.S. adults live with a mental illness (57.8 million), with many racial and ethnic minorities (REM) groups reporting higher risk of persistence and disability from mental illness than their White counterparts. Despite widespread demand, more than half of U.S. adults (roughly 28 million) needing mental health treatment do not receive care. Treatment-seeking is particularly low among REM. Even among REM treatment-utilizers, 30-57% terminate prematurely, yielding poor treatment outcomes. Primary drivers of drop-out for REM include diminished client involvement and weak patient-provider alliance. This poses a particular challenge for retention as REM frequently fear that providers will be ill-equipped to respond to their mental health needs in an empathic, affirming manner. Equipping providers with innovative technology that facilitates cultural attunement in real-time, particularly for REM patients, is essential for addressing mental health disparities in the U.S. and globally at scale. Attunement in mental health care involves the provider accurately understanding and responding to the client’s lived experience. Cultural attunement goes a step further to describe the providers’ awareness and responsiveness to the intersections of societal context, culture, and power in client experience. Cultural attunement has been shown to be a key element that retains REM in mental health services. DMH applications, including telehealth and chat-based treatments, have unprecedented access to the content of treatment and quality of conversations and may therefore allow researchers to understand the key conversational moments, predictors of success, and mechanisms of change in culturally-attuned treatment. To our knowledge, we–mpathic.ai–are the first company to build capacities to detect and correct for cultural attunement in real-time with natural language processing (NLP) and generative artificial intelligence (AI) technologies. The current SBIR Phase I proposal is from a highly skilled and interdisciplinary team with expertise in automated evaluation of health services. In this SBIR Phase I project we will employ AI and NLP processing on conversational data from 300 hundred, 30-minute health coaching sessions (60% REM) to improve cultural attunement in provider-patient interactions in real-time. The product to be developed for this SBIR will specifically: (1) detect behaviors vital to patient-provider alliance and cultural attunement, and (2) offer real-time correction to improve clinical outcomes . We will compare our AI model outputs to human labelers and patient-ratings of therapeutic alliance and cultural attunement. We will also pilot dissemination of our technology with 50% of providers from our collaborators at Wave Life Inc., a mental health platform, to gather key acceptability and feasibility data prior to launching a randomized control trial of our technology as part of a Phase II SBIR.