From Dyad to Triad: Artificial Intelligence in the Patient–Physician Relationship

Abstract

Commentary In “A Case of Bromism Influenced by Use of Artificial Intelligence,” Eichenberger and colleagues describe the case of a previously healthy 60-year-old man who presented to the emergency department with bromism (1). Thorough inquiry revealed that the patient had replaced sodium chloride with sodium bromide in his food following a discussion about salt substitutes with ChatGPT, an online chatbot powered by a large-language model (LLM). Although the medical details of the patient’s presentation and hospital course are clear, there are still many unanswered questions in this case about the patient’s artificial intelligence (AI) use. As discussed by the authors, he presumably used either GPT 3.5 or 4o and, while the authors were able to recreate the recommendation with the same chatbots, we do not have the chat transcript to fully understand the patient’s full conversation with the AI. This is critical to understanding this case because emerging evidence suggests that LLMs may offer both new opportunities and challenges for the physician–patient relationship that are meaningfully different from previous generations of internet tools.