Actionable companion · Feedback & scaffolding

Learning with agency: Generative AI for Personalized Learning Platforms in Otolaryngology–Head and Neck Surgery Education: Scoping Review

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paper Search paper agent CC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Age / stage not established. Verify the participant population and task before transferring the findings.

Original authors’ abstract, reproduced under the linked source license:

Abstract Background Generative AI, including large language models (LLMs), may enable personalized learning in medical education. In otolaryngology–head and neck surgery (ear, nose, and throat [ENT]) education, AI may support knowledge retrieval, adaptive assessment, feedback, and scalable learning, but evidence remains fragmented. Objective This scoping review aimed to map generative AI research in otolaryngology education; summarize applications, performance, benefits, and risks; and identify directions for AI-enabled personalized learning. Methods We included English-language empirical studies, preprints, and technical or developmental evaluations published from 2017 through 2026 with direct relevance to professional ENT education; nonempirical publications and studies without a professional educational focus were excluded. PubMed, Embase, Web of Science, Scopus, ERIC, IEEE Xplore, and CINAHL were searched, supplemented by citation searching. The final search was conducted on July 6, 2026. Data were charted using a standardized form by 2 reviewers and synthesized through a 3-stage inductive thematic analysis, with third reviewer adjudication. No formal critical appraisal or meta-analysis was performed because of study heterogeneity. Results A total of 42 studies published between 2023 and 2026 were included. Most were performance studies (n=33, 78.6%), followed by system development (n=3, 7.1%), educational intervention studies (n=3, 7.1%), surveys (n=2, 4.8%), and guideline development (n=1, 2.4%). Most studies evaluated question sets or model outputs rather than participants. Applications included knowledge retrieval, revision, board-style question practice, adaptive assessment, AI avatar–delivered instruction, and operative-planning support. Seven major thematic domains were identified: (1) educational support and learning enhancement; (2) learning outcomes, educational effectiveness, and pedagogical value; (3) model performance, accuracy, and variability; (4) assessment, examination readiness, and question-type sensitivity; (5) reliability, safety, and risks; (6) adoption, usability, and implementation barriers; and (7) advanced AI architectures and future educational design. Newer models performed strongly on selected structured knowledge tasks, but performance varied by task, difficulty, modality, and reasoning demands. Evidence remained heterogeneous and benchmarking focused, with only 3 educational interventions and limited evidence for sustained learning, clinical reasoning, or real-world outcomes. Hallucinations, unreliable sourcing, confidently incorrect outputs, and contextual and multimodal reasoning limitations supported supervised rather than autonomous use. Conclusions By moving beyond reviews centered on AI applications or model performance, this review provides an education-focused synthesis linking generative AI capabilities, learning outcomes, assessment, safety, implementation, and emerging architectures to the design of personalized learning in otolaryngology. Its contribution is a conceptual shift from evaluating standalone chatbots toward developing supervised, pedagogically governed, knowledge-grounded, and adaptive learning systems that preserve learner reasoning and expert oversight. These findings inform educators, curriculum designers, and developers when integrating generative AI into formative assessment, feedback, remediation, and personalized learning activities, while avoiding unsupported autonomous or high-stakes use. Future research should prioritize longitudinal educational outcomes, curriculum integration, higher-order reasoning, equitable implementation, and governance.

What kind of evidence is this?

Conversational & pedagogical agents · Evidence synthesis

Stage: Age / stage not established. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

This scoping review was conducted and reported in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist and the methodology described by Tricco et al [18]. A review protocol was developed before the review and is provided in Multimedia Appendix 1. The protocol was not prospectively registered.

Limitations and uncertainty · source passage

The evidence base has several limitations. First, studies were heterogeneous and predominantly focused on benchmarking and short-term outcomes, with limited longitudinal evidence regarding knowledge retention, transfer to clinical reasoning, or impact on real-world clinical performance. The current literature also provides limited evidence regarding whether repeated use of generative AI influences learners’ independent reasoning, metacognitive skills, or longer-term reliance on external cognitive support. Second, many evaluations rely on artificial test environments or curated question banks, which may not fully capture the cognitive and contextual complexity of authentic clinical learning. This limitation is particularly relevant because the reviewed evidence indicates substantially stronger model performance on constrained, theoretical, or guideline-based tasks than on clinical reasoning, contextual decision-making, and multimodal interpretation. Benchmark accuracy may therefore overestimate educational or clinical competence in real-world settings. Third, the rapid evolution of LLMs means that findings can become partially obsolete within months, limiting the temporal generalizability of performance comparisons between specific model versions. Moreover, differences in access to newer models and supporting infrastructure may limit the generalizability of implementation studies across institutions and may introduce inequities that have received comparatively little empirical investigation.

Design the system before the pilot

Define whether the agent asks questions, gives hints, explains or assesses. Test hallucinated feedback, excessive praise, misleading confidence and responses when the learner is confused. A conversational interface alone does not establish autonomous agency.

Give feedback on the learner’s reasoning, then fade hints. Record whether the learner can complete the next task without that scaffold.
Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.

A 90-day district pilot

Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
Days 16–30 · Design. Treat the paper as background evidence until its population is verified in the full text. Do not assign an age-specific intervention or extend adult results to children. Name an accountable owner and set the baseline, success threshold and budget before implementation.
Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
Days 61–90 · Decide. Extract the participant ages or education stage, recruitment, task and comparator before selecting a local pilot. If these cannot be established, retain the paper for design discussion rather than age-band efficacy claims. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Jiang-Tao Zhong; Zai-Zai Cao; Shui-Hong Zhou. Generative AI for Personalized Learning Platforms in Otolaryngology–Head and Neck Surgery Education: Scoping Review. JMIR Medical Education (2026-10-19). https://doi.org/10.2196/91785

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.