By: Daniela Melis
As I progress through my clerkship rotations, I am becoming more and more familiar with one of the most dreaded elements of higher education: standardized exams. In a typical clerkship core rotation, students will work 5-6 days per week, and then spend evenings completing mandatory academic components of the rotation, reading around patient cases, and revising for their end of rotation exam. Personally, I like to throw in a little me time and will reserve time to either watch Love Island or cook myself a nourishing dinner. After preparing for the exam for weeks with disseminated study materials and other resources like UWORLD or Boards and beyond, students will sit the nearly 3 hour exam and think through questions like: “Your patient travelled to Cuba for 3 weeks and returned with a seemingly viral gastroenteritis. Why did their mother’s cat pass away?” Results are typically released in a timely manner, but for obvious reasons regarding academic integrity, students are never able to see their responses to the exam questions. Furthermore, even when redacted versions of the exam are released, students may not have a significant amount of time to dedicate to this review process, as we have started the cycle of studying all over again for the next rotation. I believe this is an area where AI could potentially be applied. What I am imagining, is an AI program that analyzes the results of these exams and picks out areas of content that the student repeatedly got wrong or that they understood well. I believe that NBME already categorizes questions in terms of type and topic – thus, all of this data is out there waiting to be analyzed and presented! An AI program can not only identify areas of improvement for the student, but it could be designed to deliver this feedback in a way that protects the integrity of the test content while enhancing the knowledge base of the learner. There are already AI programs that function to convert student notes into educational podcasts, trained to present the academic material in an engaging manner (Google AI Notebook LM). Could these programs be combined so that students receive an audio or video message explaining where they went wrong and how they could improve, with the study resources they need to fully understand concepts like the difference between pseudohypoparathyroidism vs. pseudopseudohypoparathyroidism? With AI, the capabilities can be endlessly improved upon, but the need for testing would be imperative, so that students are not lead astray due to poor data interpretation or concept explanation. This type of technology must also be trained to recognize and evaluate trusted, reliable sources of medical information before it can be taught to present this information accurately and effectively. We also have to ask, what could be lost if the review process is entrusted to artificial intelligence? Overall, there is much to think about regarding the implementation of AI in healthcare, not only as a tool that can streamline problem points of EMRs, but as a tool that can and likely will be used to shape the minds of our future physicians.
Come to TISLEP 2024 to learn more about the intersection of burgeoning AI technology with medical education!
Daniela Melis
McMaster University Medical Student
Comments