Artificial Intelligence in Medical Education

Abstract

To the Editor: I greatly appreciated the commentary by Dr. Carin 1 for calling attention to the frontier of artificial intelligence (AI) within medical education. I agree with his argument for the inclusion of AI in medical education and would like to propose a direction for future curriculum development. An AI medical curriculum should provide trainees with the skills to critically evaluate AI applications akin to critiquing a research article introducing a new medication, procedure, or surgical technique. Machine learning and AI applications will become more common in clinical medicine, and physicians must have the expertise to evaluate whether to incorporate these algorithms into their practice. The assessment of AI applications requires a unique curriculum because current evidence-based medicine (EBM) guidelines are incomplete for the evaluation of AI medical algorithms. The issues of “black box” interpretability, data security, and decision liability create problems not addressed by traditional biostatistics. 2 Unfortunately, there is not yet a consensus methodology for critiquing AI algorithms for medical use. However, recent publications in Nature3 and JAMIA Open4 have begun to offer promising schemas for systematic evaluation. The schema proposed by Park and colleagues 4 is the most complete and was published only in October 2020. Ultimately, I believe medical educators should allow for the field of medical AI to develop a consensus best practice method for algorithm evaluation. In many ways, the current AI environment mirrors the EBM paradigm of the 1990s, where a central work, such as the JAMA article “Users’ guides to the medical literature,” 5 must be established before widespread adoption into medical education. Once a similar consensus is formed within medical AI, educators will have a clear outline to develop a curriculum that trains students to critique AI products for inclusion into their clinical practice.

Ähnliche Arbeiten