Education Research: Integrating AI-Enabled Interactive Case-Based Learning in a Preclinical Neurosciences Course

Case-based learning is central to neuroscience education, yet traditional approaches provide limited opportunities for iterative, interactive practice and individualized feedback among large preclinical cohorts. Large language models (LLMs) may support case-based learning by enabling interactive patient simulations and scaling formative feedback. Feasibility studies have shown high accuracy and minimal hallucinations; however, few have examined their real-world integration within medical school courses. We introduced AI-enabled cases as consolidation exercises in a preclinical neurosciences course. In this prospective pre-post observational study, we evaluated learner engagement, perceived educational value, and compared examination performance of students with access to AI-enabled cases compared with the prior year.

Seven cases were developed and customized to the preclinical neurosciences course at Harvard Medical School. Cases were delivered through an online AI-enabled learning platform (TEACHABLE). All second-year students enrolled in the course completed weekly interactive cases in which they obtained history, examination findings, and investigations by questioning an LLM-simulated patient. Students then answered short-answer questions focused on clinical reasoning to reinforce course concepts. Responses were automatically scored and students received AI-generated feedback. Evaluation consisted of user engagement data, postcourse survey responses, and a comparison of midterm and final examination performance in AY26 (intervention year) vs the prior year (AY25).

A total of 172 students completed 1,236 cases and posed 23,310 questions to the LLM. Among 123 survey respondents (71.5%, 123/172), 77% agreed or strongly agreed that the cases consolidated their understanding of neurologic conditions, and 78% favored implementation of similar cases in other courses. Student feedback identified that they valued case interactivity and real-time feedback. Mean midterm scores were greater in AY26 (n = 172) at 85% (SD 7.3%) compared with 82% (SD 7.8%) in AY25 (n = 168) (p < 0.001). Mean final examination scores were similar between AY25 (88%, SD 5.7%, n = 168) and AY26 (89%, SD 6.7%, n = 172), with no statistically significant difference (p = 0.13).

LLM-supported interactive case-based learning offered a feasible, scalable, and well-received method for enhancing clinical correlation in preclinical neuroscience education. The approach maintained educational value and high-quality formative feedback without increasing faculty workload. These findings may be applicable in settings beyond preclinical neurosciences education.