Medical schools should teach empathy and AI judgment, guest column argues

Medical schools should put less emphasis on memorization and more on patient communication, empathy and evaluating artificial intelligence, James W. Schroeder Jr. argues in a guest column. Schroeder, chief academic officer at Hospital for Special Surgery, says doctors will need to use AI tools while retaining the human judgment and compassion patients rely on.

The change matters as AI becomes part of clinical work. Schroeder writes that eight in 10 doctors already use AI for tasks including note-taking and reviewing medical research. He argues that training must prepare future physicians both to benefit from these tools and to recognize their limits, rather than accept AI outputs without question.

What skills should schools emphasize?

Schroeder calls for stronger training in bedside manner and conversations with patients and their families, including during anxious or life-changing moments. He cites a figure that more than half of patients want doctors to show qualities such as compassion and a good bedside manner. He also points to studies associating physician empathy and bedside manner with better outcomes for some conditions.

He says AI systems can reproduce blind spots in the human-generated data used to train them. The column cites one study that found significant performance disparities by race, gender and age in nearly 30% of cancer diagnostic tasks. Schroeder argues medical students should learn to spot potential bias and use AI to support, not replace, their clinical judgment.

How could clinical training change?

Medical students traditionally learn through clinical rotations and residency, shadowing experienced doctors and building knowledge from observed cases. Schroeder argues that approach should evolve as AI tools make it possible to search clinical databases and synthesize research quickly. He says trainees should watch senior physicians incorporate AI into their work, assess its results and communicate difficult news to patients.

The column says ChatGPT began outscoring medical students on the United States Medical Licensing Examination three years ago, and that AI models have since become more capable. Schroeder’s view is that students should not try to outmemorize machines. The column does not specify which curricula or training programs should change first, or how schools should measure students’ ability to identify biased AI results.

This report draws on information from The News Herald.