At the Association of Health Care Journalists Fall Summit in Los Angeles, medical educators pushed back against the idea that using AI in medical school amounts to cheating. Graciela Gonzalez-Hernandez, PhD, vice chair for research and education in computational biomedicine at Cedars-Sinai, said plainly that "AI is not plagiarism." She described an AI-friendly approach to assignments: asking students to show how they constructed prompts and then working through the output together. "If students spend their time, their energy, and their intellectual capacity on how I'm going to cheat my professor," she said, "we're just wasting our time."
Still, panelists differed on how much clinical reasoning should be preserved. Moderator Eli Cahan, MD, MSc, a Stanford neonatology fellow and investigative journalist, compared the skill to navigation before GPS, asking whether some forms of clinical reasoning can be safely ceded to AI. Tina Hernandez-Boussard, PhD, a Stanford biomedical informatics professor, offered a cautionary example: medical residents in Uganda, after losing electricity and AI support, were unable to confidently synthesize EKG information to make patient decisions. That, she said, is "a real issue to think about."
Gonzalez-Hernandez agreed that de-skilling is a potential problem and suggested training scenarios run "with paper and pencil" to ensure physicians can interpret data directly. But she also argued AI can add a layer of objectivity to clinical notes, which often carry subtle signs of physician doubt or bias. "AI can push back: 'Have you asked about this?'" she said, proposing that missed questions could be added to rebalance decision-making from a neutral point of view. The goal, she said, is to shape AI use so it helps physicians notice things they might otherwise miss.