The Future of Medical Education: ChatGPT-4 as a Tutor for the USMLE Step 1 Examination

• In Event: Saturday Roundtable Session 1:15 pm
  In Roundtable Session: GPT-Assisted Instruction (Table 17)

Abstract

In this study, we assessed the performance of OpenAI’s ChatGPT-4 on USMLE STEP 1 questions by systems and disciplines on the examination. AI answered 86% of the 1300 questions accurately, exceeding the estimated passing score of 60% with no significant difference in performance across clinical domains. Findings demonstrated an improvement over earlier models and consistent performance from biological processes to ethical considerations, raising question of potential implications as an interactive learning tool in pre-clinical medical student education. Authors provide examples and commentary on how students can leverage these features to receive real-time analogies and explanations tailored to their desired level of education. Appropriate application potentially enables us to enhance learning outcomes, especially in fields as critical and demanding as medicine.

Authors

• Brendan P. Mackey, Drexel University

• James Cross, Drexel University College of Medicine

• Razmig Garabet, Drexel University