JMIR Publications

ABSTRACT

Background:

Delivering high-quality problem-based learning (PBL) during internship is resource-intensive and difficult to scale without consistent facilitation. Although generative artificial intelligence (AI) is increasingly being explored in health professions education, many applications function primarily as on-demand answer tools and may not reproduce core process features of PBL.

Objective:

We developed MAPLE-CR, an AI-supported environment for clinical reasoning that positions generative AI not as an answer-delivery aid but as a process-oriented scaffold for PBL orchestration. The system was designed to support cognitive mechanisms through reasoning prompts, social-interactional mechanisms through simulated tutor and peer roles, and regulatory mechanisms through structured workflow and feedback loops. We evaluated its short-term learning outcomes, learner experience, and feasibility of repeated use.

Methods:

We conducted a two-stage educational evaluation. In a randomized controlled study (N=52; intervention: n=26; control: n=26), interns completed parallel pre- and post-tests around a standardized case. The intervention group engaged in asynchronous, scaffolded PBL in MAPLE-CR, whereas controls completed case-matched self-study using materials derived from the same case and learning objectives. Learner experience was assessed through a post-session questionnaire and open-ended feedback. In a voluntary follow-up, 18 participants completed 1 MAPLE-CR case per week for 4 additional weeks to examine repeated-use feasibility and longitudinal patterns across novel cases.

Results:

Baseline pretest scores were comparable between groups (P=.558). The intervention group achieved higher posttest scores than controls (between-group median difference 8.25 points; P=.001) and greater score gains (median difference 6.60 points; P=.013). Post-session questionnaire responses and open-ended feedback indicated high satisfaction, involvement, perceived support, and self-efficacy. In the follow-up phase, repeated practice across weekly novel cases was feasible. The largest gains were observed early, after which performance stabilized over subsequent attempts.

Conclusions:

MAPLE-CR was feasible and was associated with improved short-term clinical reasoning knowledge outcomes and high learner acceptability compared with case-matched self-study. These findings suggest that AI-supported, process-oriented PBL may provide a scalable supplement for clinical reasoning practice when facilitator capacity and small-group scheduling are constrained. Further research should examine its effects using stronger active comparators and longer-term outcome measures. Clinical Trial: Chinese Clinical Trial Registry ChiCTR2500115799; https://www.chictr.org.cn/showprojEN.html?proj=297863