“The eyes only see what the brain knows.” A patient said this to me recently, and it stopped me in my tracks. He was reflecting on his own experience in the medical system: the many physicians he had seen for the same problems, and the various different diagnoses he had received. Although each of those physicians likely trained at LCME-accredited institutions, the path beyond medical school, including residency, fellowship, and the accumulation of lived clinical experience, changed the lens through which he was seen.

This is not a new concept. Novelists and philosophers such as Robertson Davies, Goethe, and Carlyle have all been circling this concept for centuries. I have been witness to this problem my entire medical career. However, I did not recognize until recently that I had the power to create change in a way that went far beyond my own classroom, into those across the country and perhaps even further.

The toolbox of clinical reasoning

When we discuss clinical reasoning and use it to walk through clinical cases, I tell my students to think of themselves as carrying a toolbox. One that contains everything they know how to use. But they do not need to use every tool. Not every situation calls for a flathead screwdriver, and using the wrong one on an IKEA piece will not just fail to help, it might cause more damage. And what if something just is not fitting right? Do you have the experience to know which tool to reach for, and how to use it in an unexpected situation?

When my students encounter a case of chest pain during their cardiac block in the pre-clerkship curriculum, the first thing they think of is a cardiac cause. That is, until I prompt them to think about what is most likely, what could kill their patient, and of course the zebras they always seem to gravitate toward. Who does not love learning an unusual diagnosis and getting to show they know it?

The day I taught them that chest pain could be early herpes zoster, before the rash appears, backfired on me. They anchored so hard on that fact that they always asked whether the patient had chickenpox as a child, whether they had the zoster vaccine, whether they were immunosuppressed. They preferentially considered this diagnosis over far more likely ones. In teaching them something they should consider, however unlikely, I had biased them. They did not yet have the real-world experience to triage likelihood. That would come later.

How context shapes the clinical lens

Once they enter residency and fellowship, the context of their training shapes their differential, their approach to the history and physical, their workup. And for the things that are not definitively resolved by labs, biopsies, or imaging, they rely on training combined with anecdotal experience to guide them. And there it is: The brain only knows what the eyes see. If they only ask certain questions, only perform certain exam maneuvers, only look for certain findings, they will miss the clues that might have led somewhere else entirely.

Just to be clear, I am not talking about diagnoses such as myocardial infarction or pulmonary embolus, which have distinctive illness scripts with relatively classic risk factors, history, exam, labs, and imaging findings. I am talking about the things that are not clear.

• Dyspnea with normal spirometry: Is it asthma, vocal cord dysfunction, anxiety, all three, or something else?
• The rash that changes in morphology over time might be called one name by one physician and another by someone else depending on when it presents to clinic.
• And the pathologist who reads the biopsy does not have the entire clinical context and is not examining the patient. Their read could potentially be impacted by these factors.

The same with the radiologist. If they have the full clinical context, not just “Sudden dyspnea and chest pain. Evaluate for PE” on the imaging order, might they find something that leads to an unusual diagnosis, rather than ruling out PE and noting some non-specific findings?

Using artificial intelligence to reveal reasoning

I have been contemplating whether artificial intelligence has the opportunity to help us be better clinicians, not just help us work more efficiently. And since I am a medical educator, I began thinking about how to help medical students and health professions students approach a case with as much of an unbiased framework as possible.

We observe them practicing histories and physical exams in small groups and at the patient’s bedside. But we cannot hear their thought process in real time. If we could ask them to share their initial thoughts and be challenged Socratically to share their thought process, might we identify weaknesses we could strengthen and biases we could reflect upon?

I learned how to vibe code, building a program with AI as my collaborator, and created an AI-powered clinical reasoning platform. It meets students at the exact moment the filter is forming. Before clinical rotations have hardened their pattern recognition into specialty-shaped grooves, the platform asks them to reason out loud, to externalize their thought process in a structured way. It does not just ask what they think. It asks why they think it, and what they have not considered yet. The Socratic challenge is not correcting them toward a right answer, it is revealing the shape of their reasoning so they can see it themselves.

And because the platform now spans more than ten health professions, it offers something unusual: a space where a medical student, a pharmacy student, and a physical therapy student can reason through the same case and arrive at different but equally valid clinical questions. It illuminates dimensions of the presentation that no single lens would catch. Putting those perspectives in dialogue might be one of the most underutilized tools we have.

Lauren Fine is an allergist-immunologist.