Generative AI as Third Agent: LLMs and the Transformation of the Clinician-Patient Relationship (Preprint)

UNSTRUCTURED Use of generative artificial intelligence (AI) in healthcare presents a complex and evolving landscape with significant implications for patient-clinician interactions. Recognizing the new practical and ethical challenges raised by what may be referred to as “relational AI”–large language models (LLMs) able to “relate” to clinicians, patients and caretakers by generating human language–this paper examines the potential of generative AI to serve as a facilitator, interruptor, or both in patient-clinician relationships. Drawing on work as advocates of patient empowerment, students of computer science, and physician informaticists working to increase capacity for data exchange and mobile health, we recognize the potential for generative AI to enhance patient engagement, triage care, and support clinical decision-making. These same perspectives give us concern surrounding generative AI use and data privacy, algorithmic bias, moral injury, and the preservation of human connection. Considering the transformative power of LLMs on patient-clinician dynamics–and the still open questions about which direction that transformation will take–this paper outlines an analytic framework to understand the role and implications of generative AI in the patient-clinician relationship, and proposes an agenda for future research. Maximizing the positive potential of generative AI will require a thorough examination of which aspects of the patient-clinician relationship must remain human–and why–even if LLMs could provide a substitute. This inquiry will need to draw on ethics and philosophy to complement more traditional informatics imperatives such as patient-centered design, transparency on the use and limits of AI models, and collaboration between technologists, healthcare providers, and patient communities to shape the responsible integration of relational AI into clinical care.