What Happens When AI Becomes a Second Opinion to the Doctor's Mind?

A few weeks ago, I started to notice a new consensus: doctors are no longer just debating whether generative AI can think like a physician. Now, they’re also thinking with it. So, I can’t help wondering: what happens when AI becomes a readily available second opinion to a doctor’s mind?

The ability to evaluate your own thinking is metacognition, helping you to catch mistakes, handle the unknown, and create a game plan to reach a goal. Atrophy of metacognition lies a step beyond deskilling. With deskilling, technology reduces your ability to do a task. For example, a doctor who always asks a large language model (LLM) to give them the treatment plan for depression might find that task harder to do on their own over time. But with loss of metacognition, technology reduces your ability to monitor and evaluate your abilities. You’re not just outsourcing your thinking but also the act of judging your own thinking. Over time, the doctor treating depression might grow less sure they trust themselves to do it without digital confirmation. That’s the risk of metacognitive atrophy: the potential to lose clinical intuition.

To care for patients, metacognition is critical. The inner, expert clinician monologue you need, particularly when the stakes are high or cases are ambiguous, emerges from exposure to a breadth of patient care over years of medical training and practice. In navigating these cases, you iterate a precise process of identifying gaps in your knowledge base: you wrestle with ambiguity, debate the evidence that supports different diagnoses, and revise your assumptions about treatments. One of the most rewarding aspects of my role as teaching faculty in Psychiatry is helping trainees to refine this reflective process in real-time. It’s what builds the metacognition that guides them in deciding what they know, when to ask for help, and which questions to ask. Sitting with uncertainty isn’t a vulnerability; it’s part of a productive struggle that drives learning.

Consultation with a colleague, complex case discussions among teams, and review of peer-reviewed textbooks and literature are examples of how you develop collaborative reasoning in medicine. But AI doesn’t just retrieve information or give you answers, it generates outputs that mimic the structure of reasoning.

When used as a reflective tool with attention to how to think through and solve problems, AI has shown value in deepening learning by acting as a coach that can reduce cognitive blind spots. Thoughtful assessment of prompts and output accuracy are part of the critical thinking needed to implement reflective use.

However, the same digital tool that may strengthen reflection is also externalizing it. Dependence on outsourced validation without forethought on how one uses AI could weaken confidence in one’s reasoning. The concern is the potential to shift the growth of professional intuition and cognitive trust in any industry that requires nuanced judgement. Who really holds authority: the human expert or an always-on LLM sidekick?

As I see how AI is shaping the medical landscape, I’m struck by the significance of how doctors come to trust their own minds. The healthiest relationship with AI in medicine might be one where we establish clear processes to deliberately preserve the first go at thinking, originate the ideas that AI expands, and maintain defined checkpoints for internal reasoning. With this, monitoring for overreliance is key. There is no doubt that AI is here to stay and to think with us. What I wonder about is how we know where a machine’s intuition ends, and the doctor’s begins.