From Burden to Guidance: Leveraging Large Language Models to Build Feasible, Learner-Centered Remediation Plans

Individualized Learning Plans (ILPs) are widely used to remediate Emergency Medicine (EM) residents who do not meet performance benchmarks, yet resident engagement and perceived educational value vary. Residents often struggle with goal setting, prioritizing gaps, selecting resources, and integrating studying into demanding schedules. Emerging large language models (LLMs) offer opportunities to improve personalization, structure, and feasibility for remediation frameworks. These tools may reduce cognitive burden and improve the structure, feasibility, and acceptability of remediation frameworks.

This innovation aimed to redesign an existing ILP remediation process using an LLM to generate structured ILPs that integrate learner preferences, clinical schedules, available resources, and national curricular content outlines. Resident perceptions of the existing ILP were first assessed to identify barriers and inform the design of an LLM-supported framework.

A mixed-methods, anonymous survey was distributed to EM residents participating in an ILP program between 2023 and 2025. Quantitative data was analyzed descriptively, and free-text responses were thematically coded by two independent reviewers. Identified barriers informed the development of an LLM-supported ILP incorporating learner-defined goals, learning styles, rotation schedules, available study hours, and identified content gaps. Feasibility was explored through simulation using ten de-identified resident schedules across three residency blocks.

Of 93 residents contacted, 19 responded (20.4%). Most respondents (95.7%) reported that the existing ILP was not useful in addressing knowledge gaps. Thematic analysis identified barriers, including limited relevance to clinical practice or exam preparation, difficulty meeting learning goals, and feelings of guilt with unmet goals. Simulated LLM-supported ILPs projected a mean of 9.85 study hours per four-week block, with adaptive variation by learning style and rotation schedule while maintaining alignment with national content domains.

ILP-based remediation is commonly used in EM training programs, yet its effectiveness is often limited by inconsistent engagement and substantial cognitive burden placed on residents to prioritize content, select resources, and manage time. This innovation used resident feedback to redesign an existing ILP process to improve structure and feasibility while enhancing individualized learning. Although most respondents reported dissatisfaction with the current ILP, findings must be interpreted cautiously due to the low response rate. Similarly, our survey was intended to inform design, rather than test outcomes. Thematic analysis identified several barriers: limited relevance to clinical practice or exam preparation, insufficient effectiveness in meeting learning goals, and feelings of guilt with unmet goals. These findings suggest that a primary limitation of the existing ILP was not a lack of motivation, but insufficient structure to support learning within real-world constraints. In response, we designed a new LLM-supported longitudinal ILP that shifts the burden of organization and prioritization from the learner to the system. The framework integrates learner-defined goals, learning preferences, clinical rotation schedules, available study time, and identified knowledge gaps to generate individualized study plans aligned with national EM curricular standards. The framework also supports structured feedback from faculty advisors and offers optional individualized study scheduling to promote accountability and adaptability across varied clinical rotations. Simulation of the LLM-supported ILPs using de-identified resident schedules demonstrated feasibility, with a projected study time of 9.85 study hours per four-week block and adaptive variation based on learning style, learner level, and rotation schedule. This innovation is currently being piloted, incorporating self-determined resident goals and automated LLM-supported reminders for residents and advisors. Future work from this study group will investigate feasibility and adaptability while maintaining curricular alignment with ABEM content domains. The framework is designed to be adaptable across training programs and may offer a scalable approach to improving remediation processes in graduate medical education.