“The patient is a 59-year-old man who presents with fatigue on exercise, paleness, as well as slight numbness and tingling in both feet. No family history of anemia or diabetes, but I still want to rule them out with a blood test. Does this seem reasonable? »
This is the 6th or 7th question I ask in today's session, but the repetitiveness doesn't bother my tutor.
“Yes, performing blood tests is a reasonable approach to help rule out conditions such as diabetes and anemia, as well as other potential causes of the patient's symptoms.”
“Given the progression of the tingling, I think it is prudent to rule out injury at this stage. Is there something I'm missing?
My tutor doesn't mind that I'm still unsure, even though I've spent weeks on these types of exercises together.
“Although localized injury may be less likely due to the bilateral nature and progression of tingling symptoms, it is nevertheless important not to rule it out prematurely. A complete clinical evaluation should include a thorough neurological examination, blood tests, and possibly neuroimaging studies such as MRI or CT scan, to identify any structural abnormalities or lesions.
For several weeks, I have been calling on my new tutor – Dr ChatGPT – to help me better develop and refine my clinical reasoning. As a preclinical medical student, I have a lot of work to do: my tutor doesn't mind if I ask a question or thirty. My tutor is always available, infinitely patient and never condescending. Like billions of other people, I thought ChatGPT was a new and fun technology, but personally, I didn't really have a clue about the best way to use it. After seeing GPT-4 used at a Clinicopathology Conference (CPC) last year, I sat down to see if it could help me improve my reasoning in a clinical setting. It's not a perfect tool, but it can offer a diagnosis and give clear, logical reasoning as to why that diagnosis is the best. I knew I had to find a way to incorporate it into my own education.
Of course, I had some apprehension. Many concerns have already been raised about the phenomenon now called hallucination, the propensity of large language models (LLMs) like GPT to confidently invent information. I experienced this when I asked ChatGPT to help me with a literature review. The bibliography looked good; it was in APA format and had authors and dates, but they weren't all real articles or reviews. Additionally, concerns have already been raised that LLMs cannot replace human reasoning. However, in reality, research shows that LLMs perform as well or better than humans on many reasoning tasks.
There was no doubt that I was learning, but the more I worked with my tutor, the more questions I asked myself: Is it ethical to use AI to organize course material? How about asking AI to predict test questions based on these documents? Even what I do with my clinical vignettes walks a fine line; it would be easy to just forward the entire case to GPT-4 and request the diagnosis. Can AI be used to cheat, or will over-reliance on it weaken rather than strengthen my clinical reasoning?
I'm certainly not alone in trying to find ways to use AI in my medical training; many of my classmates do the exact same thing. In many ways, AI has forced me and my fellow students to have important conversations about the purpose of medical education. No physician can reasonably be expected to possess even a small fraction of all medical knowledge. The existence of products like UpToDate and Micromedex presupposes an accepted limit to a doctor's intelligence. We cannot know everything at all times or keep up with all the new scientific knowledge.
While medical students will always need to rely on our intelligence, we are already seeing the need for intelligence, like UpToDate, to keep the knowledge available to us until we are faced with a situation in which we can apply them. How much will AI reasoning abilities play into the discussion about what is expected of a student? We want to have strong reasoning skills, but is using AI to augment these skills acceptable, or even beneficial? These are the debates we are just beginning to have as we consider our future in medicine, conversations that are currently happening without faculty.
I do not have the audacity to suggest answers to these questions, I only point them out as part of the zeitgeist of modern medicine, debates with which I and my comrades will have to grapple throughout our career. We are already dealing with them. Sooner or later, our teachers will have to do this too. This technology is still in its infancy, but I will be part of the last generation of medical students who remember medicine before AI. It is critical not to long for the “good old days,” but rather to figure out how AI will improve patient outcomes and our practice of medicine. I want to be part of a generation that embraces AI, not as a shortcut to education but as a tool to improve it.
Right now, GPT-4 is my tutor, highlighting my weaknesses, suggesting questions I should consider, and helping me strengthen my clinical reasoning. And my story is not unique. I know a composition teacher who has adopted ChatGPT and has his students compete to improve their rhetorical skills. My 7-year-old son is using AI to learn math this summer, receiving feedback on his calculation process, instead of just correcting his answers.
Like any tool, it depends on how we use it. The time spent with my machine tutor helped me immensely. This is already paying off through simulated patient interviews and early clinical exposures: my knowledge and reasoning have improved significantly over the past few weeks. The conversation I opened this article with was about a case in my nephrology and vitamin unit. Even though the actual diagnosis was pernicious anemia, the GPT allowed me to not categorize my reasoning too early in the process. This helped me broaden my differential beyond the unit I was studying, instead allowing me to focus on the patient and their symptoms. Asking all of my questions to GPT helped me ask better questions of patients in the clinic and helped me consider factors that I otherwise would not have done.
Ironically, my tutor is very aware of his own abilities and limitations: “My answers should not be used as a substitute for professional medical advice, diagnosis or treatment. For any health-related concerns, it is important to consult a qualified healthcare professional.
AI is here and it will change medicine and medical education. If we are involved in these conversations, we can guarantee that change will be for the better.
Jeffrey D. Goddard is a medical student.