Hermeneutics as impediment to AI in medicine

Area of Science:

• Philosophy of Mind
• Medical Ethics
• Artificial Intelligence

Background:

• Current AI, like generative models, processes information based on probability from vast datasets.
• This approach assumes human knowledge can be captured propositionally, similar to AI's text generation.
• This assumption is challenged by the nature of human understanding in complex fields like medicine.

Purpose of the Study:

• To challenge the notion that artificial intelligence (AI) can fully replace human roles in medicine.
• To explore the limitations of AI in replicating human understanding and its implications for medical practice.
• To highlight the importance of non-textual, embodied, and dialogical aspects of human cognition in medicine.

Main Methods:

• Drawing on Hans-Georg Gadamer's hermeneutic philosophy.
• Analyzing the nature of human understanding as an interpretive, situated, and dialogical process.
• Contrasting human cognition with the operational principles of generative AI.

Main Results:

• Human understanding is interpretive, relying on historical situatedness and pre-existing judgments, unlike AI's rule-based approach.
• Human understanding is dialogical, involving a 'fusion of horizons' and questioning of prejudices, which AI cannot replicate.
• AI lacks embodiment, a crucial factor in human perception and understanding, and thus cannot grasp non-textual sources of meaning.

Conclusions:

• AI faces inherent limitations in core medical tasks such as history-taking, physical examination, diagnostic reasoning, and treatment negotiation.
• Embodied, dialogical, and historically situated aspects of human understanding are vital for ethical medical practice.
• While AI can assist physicians, it cannot replace the nuanced, human-centric elements of the physician-patient relationship.