In Reply to Webster.

Abstract

We would like to thank Webster for his insight on our article, “Artificial intelligence in undergraduate medical education: A scoping review.” Our article highlights the differing views on the impact of artificial intelligence (AI) in medicine. As noted by Webster, studies, including the one by Masters, 1 argue for the disruptive potential of AI tools, postulating the (almost) complete replacement of physicians by AI systems. This perspective contrasts with most studies that discuss the diagnostic and predictive role of AI tools and their capacity to process large data to aid physicians’ medical decision making. Therefore, while we agree that the complete replacement of physicians by AI is unlikely as of now, the expanding role of AI in health care is evident by the growing literature on the collaboration between clinician experts and AI. 2 Furthermore, we agree that the transparency of AI tools is critical in becoming mindful users of AI in the health care setting. However, we cannot completely understand AI’s decision-making process. This has led to pitfalls, including those highlighted by Webster, where an AI software assessed patients with pneumonia and asthma to be at a lower risk of complications than patients with pneumonia alone, as it failed to account for the variable of intensive care unit admission. 3 However, this limitation of AI tools is not a reason to discount AI. Instead, it highlights the importance of AI training in medicine and the involvement of physicians during the development of AI tools. As users of AI, physicians must understand its strengths and limitations, identifying the variables involved in its decision making to ensure the validity of its algorithms. Hence, we believe that the transparency of AI tools is not a binary outcome but a goal that we must continuously strive to achieve. This effort is critical in avoiding the phenomenon of black box AI, especially as the impact of AI on health care is continued and imminent.