Evidence Traceability and Adaptive Judgment Regulation in Medical Education: Research Data, Materials, and Analysis Code

Overview This research dataset provides de-identified numerical data, clinical educational materials, scoring records, and analysis code for a multisite investigation of evidence traceability and adaptive judgment regulation in artificial intelligence-supported medical education. The investigation comprises two independent 2 × 2 randomized factorial experiments conducted across three educational institutions. Experiment A involved 1,080 medical students and examined diagnostic accuracy when students encountered correct or misleading AI recommendations. Experiment B involved 600 students and evaluated decisions to accept or reject AI recommendations that conflicted with students' initial judgments. An unaided assessment approximately 14 days after Experiment A evaluated diagnostic performance on new cardiovascular cases and evidence-classification tasks. Dataset Contents The archive includes: De-identified participant-level and trial-level numerical analysis data. Clinical vignettes, diagnostic answer keys, AI advice versions, and case-linked evidence passages. Experimental allocation, outcome, and process-measure records. Numerical written-reasoning ratings and scoring analyses. Qwen3.5-27B engineering evaluation outputs and expert rating data. Independent case–evidence measurement-development data. Clinical case-risk assessment and review-routing analyses. R and Python code for statistical estimation, sensitivity analyses, and computational verification. Model outputs, bootstrap inference records, data documentation, and file-integrity manifests. The statistical materials support mixed-effects modeling, standardized treatment-effect estimation, bootstrap-t inference, multiple-testing adjustment, missing-outcome sensitivity analyses, measurement evaluation, and examination of intervention processes. Data Access and Reproducibility The archive provides structured numerical data, study materials, documented statistical outputs, and executable analysis resources. A README and accompanying file indexes identify the relevant datasets, analysis programs, and computational procedures. Student free-text responses and signed consent documents are maintained separately from the public dataset. Access to restricted research records is subject to applicable ethical, privacy, and institutional requirements.