Large Language Models, argumentation, and healthcare

Abstract There are various ways in which Large Language Models (LLMs), the latest breakthrough in Artificial Intelligence, are relevant for medicine: this paper focuses on their potential for supporting and improving argumentation in healthcare, both for patients and for practitioners. The message is mostly positive, suggesting adoption of such systems, but with specific cautions: most notably, the need to leverage them for enhancing human communicative and epistemic performance, rather than replacing it, and the importance of training users on few key principles to guide their deployment of LLMs in healthcare. The paper is accompanied by four concrete use cases, included in the supplementary materials, that constitute an integral and crucial part of this contribution.