OpenEvidence and I can have an iterative conversation grounded in a case\u2019s peculiarities, a difference in kind from when I used to look up general information and try to apply it to an individual situation. The app lets me input my patient\u2019s labs, or upload an imaging report photo, and then offers suggestions about what it sees. Not infrequently, it suggests something I did not consider.<\/p>\n
I\u2019m surprised by how much I learn from, and enjoy, these interactions. More important, my patient care has improved as conversations with the agent push my thinking beyond its usual limits.<\/p>\n
For the residents I supervise, however, the app\u2019s speed and anticipatory helpfulness are a hindrance as much as an enhancement. When I skim a chatbot\u2019s response to a question about back pain, I\u2019m intuitively testing its suggestions against my hard-won mental model for how to approach back pain. Could I have developed that if I\u2019d had a robot leading my thinking all along?<\/p>\n
I asked Adam Rodman, a physician and A.I. researcher at Harvard Medical School, how to think about these risks. He said he shared my concern that trainees aren\u2019t benefiting from \u201cproductive struggles\u201d as much as their predecessors. \u201cIn experienced hands, decision support might cause a little de-skilling, but a little de-skilling probably doesn\u2019t matter much to you,\u201d Dr. Rodman told me. \u201cBut if you do that to somebody who has much less skill, the cognitive offloading can cause never-skilling.\u201d<\/p>\n
I don\u2019t think it\u2019s critical that my residents memorize every diagnostic protocol, just as it\u2019s no longer critical that I recall my friends\u2019 phone numbers. But I do want them to learn to hear problems the way a doctor is supposed to, as stories that contain clues that set off a cascade of cognitive associations. And I want them to develop the cardinal instincts that amount to good judgment in medicine: the ability to distinguish sick from not sick and to prioritize competing patient concerns, as well as the tact to effectively seek information beyond what a patient voluntarily reveals.<\/p>\n
A.I. is only the latest tool to provoke a reckoning around which skills are truly essential to doctoring. As a student, I struggled valiantly to identify heart murmurs in my patients, even though I had access to imaging techniques that made my stethoscope somewhat redundant. Similarly, students today need opportunities to build cognitive muscles through study, even if they will seldom rely on their own memories over their machines.<\/p>\n
Dr. Rodman shared ideas for how to help learners experience worthwhile friction despite the availability of A.I. \u2014 proposals like A.I.-free clinical reasoning sessions and A.I. literacy training. We don\u2019t yet know, however, what caring for patients will even look like once artificial intelligence is fully integrated into medicine, nor what its capabilities will be.<\/p>\n
Current doctors-to-be may be the most adversely affected generation when it comes to A.I. They\u2019re training before we know enough about how to teach them well in this new paradigm. I want their patients to benefit, as I think mine are, from doctors supported by these tools. But I also want them to develop a talent for reasoning their way through complex problems full of distractions and ambiguities, and rich with emotional stakes.<\/p>\n
The hardest parts of doctoring will stay hard, I think, even if some other tasks are made easier. Mastering those challenges will still be where the utility and the pleasure of medicine will be found. The profession has adapted to waves of paradigm-shifting technological change before, yet the privilege of taking care of people remains.<\/p>\n
![\"\"\/](\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/09\/cropped-_zWoAgAAAAB6ObQXEgDpVg-override.png\")
<\/p>\n