2026 ASEE Annual Conference & Exposition

Traditional education largely relies on a standardized, one-size-fits-all approach, which often does not meet the diverse learning preferences of students. Personalization of education has been a constant challenge, and recent advances in large language models (LLMs) have opened new avenues for personalizing instruction at scale. Existing literature emphasizes prompt engineering, domain-specific fine-tuning, retrieval-augmented generation, knowledge-graph integration, and educational theory-driven model adaptation. Other approaches emphasize tailored problem generation or automated tagging of knowledge components. Despite the abundant literature on LLMs, there are very few studies that focus on utilizing LLMs to tailor education across different settings, such as project-based learning, labs, and large lecture settings. To address this gap, our scoping review aims to explore the different adaptive learning strategies, their advantages, limitations, and implications for future research or practice in educational contexts, including applications to project-based learning, labs, and large lecture settings.

An initial set of 366 journal articles was retrieved using a Python script that accessed the Semantic Scholar API. The corpus was then screened using predefined inclusion and exclusion criteria. Studies were included if they were published in peer-reviewed journals or conferences from 2021 onward and contained search strings such as “personalized learning” or “adaptive learning.” This process resulted in a final dataset of 26 empirical studies that met the criteria for analysis.

The 26 empirical studies illustrate a variety of LLM applications in engineering education, including personalized tutoring, curriculum development, and instructional adaptation. For instance, in efforts to mimic learner behaviors, LLMs were used to construct learner profiles based on educational theories or information organization frameworks, thereby generating responses based on the individual's learning preference. Moreover, approaches to determine correct prerequisites helped outline personalized paths and methods to identify knowledge gaps from assessment items. In short-term professional training programs, LLMs have shown to significantly expedite curriculum formulation. Similarly, on the instructor side, LLMs were integrated to generate tailored problems to target common student misconceptions. Furthermore, LLMs enabled ongoing adaptation in visual question-answering frameworks in specialized fields such as robotic surgical education, addressing differences in procedures and data distributions to improve training in clinical simulation environments.

This review provides important insights for both researchers and educators. Faculty can utilize personalized LLMs in large lecture settings to enhance student engagement and accommodate diverse learning needs at scale. Similarly, customization of curricula can be done by adjusting assessments and content to fit individual learning paths. Additionally, in situations where real-world learning is difficult, like robotic surgical simulations, instructors can use LLMs to create immersive, safe training experiences that connect theory to practice. In order to guide the creation of more rigorous and fair LLM frameworks in engineering education, this review identifies areas for future study that will look into biases, ethical considerations, and integration constraints.