Let’s Talk Systems - What Do We Really Mean by a Systems Approach in AI-Enhanced Medical Education?Surapaneni, Krishna Mohan
doi: 10.1007/s40670-026-02783-8pmid: N/A
Artificial intelligence is no longer peripheral to medical education; it is steadily reshaping how knowledge is accessed, interpreted, and applied. Yet its presence has outpaced conceptual clarity. AI is used to generate assessments, simulate clinical encounters, provide adaptive feedback, and support decision-making, but these applications often exist as isolated innovations rather than components of an integrated educational design. It is high time we engage with AI through a systems thinking approach. This perspective article first interrogates current patterns of use through an utility–intent-outcome analysis, examining not only what AI does, but what it changes. It then adapts Bronfenbrenner’s bioecological model to situate AI within interconnected learner, institutional, cultural, and temporal systems. The final section reflects on the structural, ethical, and implementation science and challenges that must be addressed to move from episodic adoption toward sustainable, coherent transformation.
Collaborative Testing in Medical Education: Insights from a Hybrid Individual–Group Examination ApproachKubick, Becca N.; Thompson, Andrew R.; Lowrie, Donald J.
doi: 10.1007/s40670-026-02773-wpmid: N/A
Team-based learning is a widely accepted instructional approach in undergraduate medical education, but the application of its principles outside of its traditional framework has not been fully explored. Here, we describe the implementation of a hybrid individual-group examination format during a two-week Multisystems course at the University of Cincinnati College of Medicine. Students completed two separate in-house examinations during the course, each with individual and group components, as well as a cumulative NMBE final examination that was taken individually. A portion of the second individual examination retested concepts from the first examination to explore the impact that collaborative test-taking had on student learning. Student perceptions of the experience were evaluated using a survey after the final examination. We found a small but statistically significant decline in average in-house and NBME individual examination scores following the transition from traditional individual testing to the hybrid individual-group format. However, measured improvement in performance on retested concepts suggests that collaborative testing had a positive impact on student learning, especially among lower performing students. This testing model was overwhelmingly supported by students, who reported that group testing enhanced their learning experience and decreased the stress surrounding preparation for the examinations. Opinions on whether this approach to assessment should be utilized other courses were divided and heavily influenced by respondee class rank, highlighting a potential barrier to implementation. These findings shed light on the benefits and drawbacks of collaborative testing in the medical school context.
PeruMedQA: A Stress Evaluation Using Ten Large Language Models to Answer Medical ExamsCarrillo-Larco, Rodrigo M.
doi: 10.1007/s40670-026-02780-xpmid: N/A
LLMs have demonstrated remarkable ability in answering medical examinations. However, whether their performance remains stable under stress evaluations is unknown. We used PeruMedQA (n=8,380), a multiple-choice question-answering dataset, and ten medical LLMs. The stress test consisted of randomly shuffling the multiple-choice answers. Using paired t-tests and Wilcoxon tests, we compared LLM accuracy on the original versus the shuffled exams. MedGemma 27B, OctoMed-7B, and Meditron 7B did not exhibit statistically significant differences, overall and stratified by year/specialty. The largest non-significant differences for these LLMs were −2.82, −3.15, and −5.96 percentage points, respectively. These three LLMs may represent robust options for AI applications in Spanish-speaking Latin America.
Piloting an AI Introduction Program for Incoming Medical Students: A Novel Approach to Medical EducationTa, Kenny T. L.; Patel, Ayush; Keller, Nathan; Pandey, Shivansh R; Kc, Abhinab; Arries, Cade
doi: 10.1007/s40670-026-02782-9pmid: N/A
BackgroundArtificial intelligence (AI) is increasingly recognized for its potential to enhance personalized learning and clinical decision-making in medical education. However, limited integration of AI into medical curricula may leave students feeling underprepared for AI-driven healthcare. This study piloted an AI education program for first-year medical students, assessing post-session perceptions of knowledge, confidence, and perspectives regarding AI in medicine.MethodsAn interactive AI session was conducted for 241 first-year medical students at University of Minnesota Medical School. The session included components of lecture, live demonstrations, and exercises with AI tools such as ChatGPT, Copilot, and Gemini, where students applied AI to various educational and clinical scenarios. Discussions included potential benefits, harms, and limitations including ethical implications, bias and privacy concerns. A post-session survey captured students’ understanding, perceived utility, and confidence regarding AI use in education and in medicine.ResultsOf 241 medical students, 180 attended the session, and 92 of attendees (51%) completed the post-session survey. Findings showed that a majority recognized AI’s value for enhancing study efficiency and diagnostic support, with over 50% indicating increased confidence in incorporating AI tools into their education and possible future practice. Concerns emerged around AI reliability, potential over-reliance, and ethical issues, especially regarding data privacy and bias.ConclusionsThis pilot study provides preliminary evidence supporting the early introduction of AI concepts in medical education. Students’ ethical concerns highlight the need for balanced curricula that combine foundational AI literacy with critical ethical understanding. The study’s descriptive, post-session design limits causal inference; future work should incorporate pre- and post-assessments and longitudinal follow-up to evaluate the sustained impact of AI education on clinical competencies and ethical awareness.
Comparison of Error Rates in Technical Skills and Contextual Skills in a Large-Scale Peer-Assessed Clinical Skills ExaminationMesserer, David Alexander Christian; Keis, Oliver; Horneffer, Astrid
doi: 10.1007/s40670-026-02764-xpmid: N/A
Training of clinical skills often prioritizes technical performance, while contextual skills such as hygienic practice, communication, and documentation receive less systematic attention. In a large-scale peer-assessed examination, third-year medical students completed examination encounters (n = 2,076) using a standardized 20-item binary checklist with predefined allocations to technical and contextual skills. Error rates were low but variable in both categories (technical: 10% (5%; 15%); contextual: 0% (0%; 17%), median with interquartile range). These findings support integrating contextual competence into structured skills education and suggest that brief, checklist-based assessments may reveal actionable deficits in hygiene and documentation that would otherwise remain implicit.
Enhancing eXcellence Through Community, Empowerment, and Learning (EXCEL): aPilot Study of aSelf-Determination Theory-Informed Peer-FacilitatedPsychoeducationalIntervention for Health Sciences TraineesMladen, S. N.; Zhai, Y.; Gorman, C. A.; Rogers, D. A.; Rampe, R.; McCarthy, S.; Watson, D.; Boitet, L. M.
doi: 10.1007/s40670-026-02770-zpmid: N/A
Health sciences trainees experience substantial distress from factors derived from the learning environment. Yet traditional training programs rarely provide structured opportunities to build autonomy-supportive nontechnical skills needed to navigate these challenges. To address this gap, we developed and evaluated a novel peer-led psychoeducational group intervention designed for health sciences trainees grounded in Self-Determination Theory (SDT). This program aimed to strengthen trainees’ access to resources supporting competence, relatedness, and autonomy through three structured components: psychoeducation, guided discussion, and targeted skill-development. A pre-post design was used to collect Work-Related Basic Psychological Need Satisfaction Scale (BPNS-W) and perceived stress data, and the post-session survey included items from the Session Wants and Needs Outcome Measure (SWAN-OM) and two general evaluation questions. Mean-level paired analyses demonstrated a significant increase in resource acquisition related to autonomy, and perceived stress decreased significantly from pre- to post-session. Importantly, latent change score modeling revealed substantial variability in individual trajectories, and indicated trainees with lower baseline basic psychological need satisfaction showed the largest gains, which were statistically significant across all three SDT domains. Trainees positively evaluated the program, reporting that they felt listened to, and were able to identify self-support strategies and explore their feelings. Participants also indicated that they found the sessions helpful and were likely to attend future sessions. Together, this study supports the utility of an SDT-informed, peer-facilitated intervention model for reducing trainee perceived stress and strengthening resources in support of basic psychological need satisfaction.
Near-Peer Teaching to Integrate Anatomical Imaging into the Medical School CurriculumWarfield, Andrew E.; Plante, Curtis A.; Walsh, Ryan; Hielscher, Abigail
doi: 10.1007/s40670-026-02760-1pmid: N/A
PurposeAdditional instruction in imaging can solidify medical students’ anatomical knowledge, which in turn, supports better clinical reasoning skills. Imaging education is often constrained due to insufficient time in the curriculum as well as instructor availability. Near-peer teaching can address some of these constraints.MethodsImaging at our school is integrated across seven anatomy blocks in the first-year curriculum. One-hour optional review sessions, led by second-year medical students, were incorporated into each block with a focus on high yield exam concepts. Attendance was collected at each session. Pre- and post-session quizzes, post-session surveys, and an end of course questionnaire measured student knowledge and perceptions. Quiz performance was compared before and after each session. Survey responses were recorded on a Likert scale and quantitatively analyzed. Open ended questions were analyzed using thematic analysis.ResultsAn average of 50.7 ± 6.9 students attended each session and students performed better on the post-session quizzes for 6/7 sessions. Students felt more knowledgeable about anatomy and imaging topics and were more prepared to answer test questions following these sessions. Longer-term knowledge retention, assessed at the end of the course using multiple choice questions from each pre/post-session quiz, was also evident. Students felt satisfied with near-peer facilitators and with session content and delivery. Comments focused on desiring more time to answer practice questions, acquiring learning tips from facilitators, and preferring longer sessions.ConclusionsThis study demonstrates that near-peer-led imaging sessions are feasible, engaging, and have positive effects on student knowledge and test preparedness.