Mary Steps Out: Capturing Patient Experience through Qualitative and AI Methods

AbstractFrank Jackson's 1982 thought experiment, "Mary's Room," illustrates the philosophical divide between propositional and experiential knowledge. We present a compelling case for the incorporation of lived experience into biomedical research and advocate the integration of AI - particularly large language models (LLMs) such as GPT-4 - to bridge this epistemological gap. When paired with sophisticated natural language processing techniques, LLMs could systematically analyze qualitative data from disconnected electronic health record data. We explore methodologic use cases - including grounded theory and thematic analysis - while addressing the challenges of analytical fidelity and bias reduction with continuous human oversight. We suggest that AI-augmented qualitative research can uncover hidden insights from a multitude of disparate datasets, revealing patient experiences that would otherwise remain inaccessible. This integrated approach could enrich the understanding of health and disease, while ensuring it is as inclusive and reflective of human complexity as the lives it seeks to understand and improve.