News thumbnail
Science / Thu, 19 Nov 2026 Nature

AI-assisted curriculum mapping for competency-based dental education: a comparative study

Curriculum mapping is essential for competency-based dental education but is often time-consuming and dependent on faculty interpretation. This study compared the accuracy, quality, efficiency, and faculty acceptance of ChatGPT-assisted versus manual curriculum mapping. A fixed-order, repeated-measures comparative study included 90 dental faculty members from Periodontology, Oral Medicine, and Oral Diagnosis. Participants completed manual mapping followed by ChatGPT-assisted mapping after a seven-day washout period, aligning course-level Intended Learning Outcomes with Program Learning Outcomes. ChatGPT-assisted mapping was associated with significantly higher mapping accuracy, higher rubric-based quality scores, and shorter completion times than manual mapping.

Curriculum mapping is essential for competency-based dental education but is often time-consuming and dependent on faculty interpretation. As dental schools undergo digital transformation, artificial intelligence (AI)-based tools may enhance curriculum quality assurance and alignment processes. Large language models (LLMs) may support curriculum alignment tasks; however, evidence in dental education remains limited. This study compared the accuracy, quality, efficiency, and faculty acceptance of ChatGPT-assisted versus manual curriculum mapping. A fixed-order, repeated-measures comparative study included 90 dental faculty members from Periodontology, Oral Medicine, and Oral Diagnosis. Participants completed manual mapping followed by ChatGPT-assisted mapping after a seven-day washout period, aligning course-level Intended Learning Outcomes with Program Learning Outcomes. Mapping accuracy against an independent, Delphi-derived reference standard was the primary outcome. Secondary outcomes included rubric-based mapping quality, task completion time, error patterns, and faculty perceptions assessed using the Technology Acceptance Model (TAM). The primary comparison used participant-level paired t-tests, with participant-clustered logistic generalized estimating equations used to assess incorrect mapping occurrence. Mapping accuracy was significantly higher with ChatGPT assistance than with manual mapping (78.8% ± 7.9% vs. 74.3% ± 8.2%; mean difference = 4.5% points, 95% CI 2.81–6.19; Cohen’s d_z = 0.56; p < 0.001). A secondary participant-clustered analysis supported this finding, showing lower odds of incorrect mapping (OR = 0.78, 95% CI 0.62–0.98; p = 0.035). ChatGPT-assisted mapping yielded higher rubric-based quality scores (25.67 ± 0.26 vs. 20.96 ± 0.36; p < 0.001) and shorter completion times (14.0 ± 1.5 vs. 27.8 ± 2.9 min; p < 0.001). The participant-clustered analysis found no statistically significant difference in the relative distribution of error types between methods (p = 0.053). Faculty reported favorable technology acceptance, with all three TAM constructs rated significantly above the neutral midpoint (all p < 0.001). ChatGPT-assisted mapping was associated with significantly higher mapping accuracy, higher rubric-based quality scores, and shorter completion times than manual mapping. However, the fixed-order, non-randomized design precludes attributing these differences exclusively to AI assistance, as practice and familiarity effects may have contributed to the observed results. These findings support further evaluation of LLMs as faculty-supervised decision-support tools for curriculum alignment, particularly through randomized, counterbalanced, multicenter studies.

© All Rights Reserved.