The rise of AI in medicine is both exhilarating and deeply unsettling, especially when you consider its impact on the next generation of doctors. Personally, I think we’re standing at a crossroads where technology could either elevate medical training or inadvertently hollow it out. What makes this particularly fascinating is how AI tools like OpenEvidence are reshaping the very foundation of clinical reasoning—a skill that’s supposed to be honed through years of struggle, failure, and reflection.
The Illusion of Competence
One thing that immediately stands out is how AI can create the illusion of competence. Trainees today can lean on tools like OpenEvidence to generate near-perfect diagnoses in seconds, bypassing the messy, often painful process of trial and error. From my perspective, this isn’t just about deskilling—it’s about never-skilling. What many people don’t realize is that medical training is as much about learning to think as it is about learning facts. The struggle to piece together a diagnosis, to grapple with uncertainty, is where true clinical judgment is born. AI risks short-circuiting this process, leaving us with doctors who look competent on paper but lack the bedrock of independent reasoning.
The Apprenticeship Paradox
If you take a step back and think about it, medicine is one of the last great apprenticeships. Residents and fellows learn by doing, by making mistakes, and by being corrected. AI, however, inserts itself into this process in a way that’s fundamentally different from previous technologies. It’s not just a tool; it’s a cognitive crutch. What this really suggests is that we’re not just augmenting human judgment—we’re outsourcing it. And that raises a deeper question: What happens when the next generation of doctors becomes so reliant on AI that they can’t function without it?
The Trap of the Arms Race
Here’s a detail that I find especially interesting: many trainees know they’re becoming over-reliant on AI, but they feel trapped. It’s an arms race where opting out feels like falling behind. This isn’t just about individual choice; it’s a systemic issue. In my opinion, the solution can’t rely on self-restraint. Medical schools and residency programs need to step in and redefine the rules of engagement. A simple but powerful shift could be to enforce a reason-first, consult-AI-second approach. Trainees should be required to make their unaided assessments visible, to commit to a diagnosis before turning to the machine.
The Role of Friction in Learning
What makes this particularly intriguing is the concept of desirable difficulties—a term from learning science that emphasizes the value of struggle in skill acquisition. AI, when used after an independent attempt, could actually enhance learning by highlighting gaps and biases. But the sequencing matters. If AI is the first port of call, it becomes a substitute for thinking, not a supplement. A detail that I find especially interesting is how aviation handles this: pilots are required to periodically disengage autopilot to maintain manual flying skills. Medicine needs a similar discipline—a deliberate, structured approach to ensure trainees don’t lose their unaided reasoning abilities.
Interrogating the Machine
This raises a deeper question: Can we teach trainees to interrogate AI, not just use it? I think the answer lies in simulation-based training. Imagine drills where trainees are presented with AI-generated assessments that contain subtle flaws. The goal wouldn’t just be to get the right answer but to understand why the AI might be wrong. This isn’t about fostering skepticism for its own sake but about cultivating disciplined judgment. What this really suggests is that AI literacy should be as much a part of medical training as anatomy or pharmacology.
The Bigger Picture
If you take a step back and think about it, this isn’t just about medicine—it’s about how we as a society integrate AI into professions that require human judgment. The stakes are particularly high in healthcare because lives are on the line. Patients need doctors who can stand apart from the machine, who can question its reasoning, and who can recognize when it’s wrong. In my opinion, the goal shouldn’t be to make medical training harder but to make it smarter. AI should augment human reasoning, not replace it.
Final Thoughts
As someone who’s watched this debate unfold, I’m struck by how much of it comes down to sequencing and intention. AI is a tool, not a teacher. It can provide answers, but it can’t teach the art of questioning. The challenge for medical education is to ensure that trainees develop the kind of judgment that AI can’t replicate—the kind that comes from years of struggle, uncertainty, and reflection. Personally, I think we’re up to the task, but it’ll require a deliberate, thoughtful approach. The future of medicine depends on it.
