The need for health AI ethics in medical school education

Readying Medical Students for Medical AI: The Need to Embed AI Ethics Education

“…In coming years, it may become necessary to give medical students more substantial technical training in AI, machine learning, and/or data science. Indeed, Katznelson and Gerke note that the AMA already recommends including data scientists and engineers in medical schools [9]. However, we submit that the need to learn about medical AI in greater technical detail as part of the basic curriculum (i.e., from those with technical expertise in AI) will depend partly on the nature and extent of medical AI uptake in clinical practice.…”

“…• informed consent: It remains an open question as to whether there is something distinctive and essential about medical AI that requires informed consent about its use, and, if so, what aspects of medical AI need to be communicated as part of a sound informed consent process [9,15].…”

“…• bias: Awareness is growing about the existence of sometimes obvious but also subtle biases within AI, and questions remain as to whether and how these biases can or should be mitigated [9].…”

“…• safety: New information technologies like AI can bring foreseen and unforeseen errors and risks to health delivery and care [16]. For example, medical AI devices are trained to minimize various mathematical "loss functions" (i.e., numerically defined objectives, such as average prediction error or classification error), and may not, through their explicit design and default use, necessarily minimize harm to patients [9].…”

“…Fewer papers still examine teaching medical AI ethics to undergraduates and graduates. In a notable exception to this rule, Katznelson and Gerke recently identified and sought to justify key ethical ideas that medical students need to learn, as contextualized through real-life case studies involving AI [9]. However, their recommendations mainly address the "What?"…”