Is anybody there? Clinician embodiment, artificial intelligence, and the clinical encounter

Abstract: Debate about the role, ethics, and performance of artificial intelligence (AI) in medicine continues at a furious pace, and no less so in clinical medicine. In a series of recent empirical studies, large language model (LLM) and clinician performance has been compared side-by-side, evaluating various aspects of the clinical encounter including diagnostics, therapeutics, and empathy. While these studies have uncovered some technical errors, the implicit assumption is that if the errors are corrected and performance is superior, it may be unethical not to employ AI tools in these tasks.

In this presentation, drawing on the philosophical tradition of phenomenology and the phenomena of the clinical encounter, I will examine the implications of embodiment, especially embodiment of clinicians, on what Dorothee Legrand calls the “intercorporeality” of medicine. It is by their own corporeality that human clinicians inhabit time, place, and morbidity with patients, among other features of their encounter. These features are necessary for medical practice and, because they are pre-intentional, cannot be simulated by language models.

I argue that their incorporeality will remain a permanent challenge for “artificial clinicians” despite astonishing gains in simulating some aspects of clinical encounter. The role of AI tools for clinical medicine is briefly considered in this light. Finally, these conclusions may be fruitfully applied to other proposed roles for AI in supplementing or replacing certain human relationships.