2025 ASEE Annual Conference & Exposition

Personalized learning has been a long-standing goal of education in the United States. To date, numerous computer-based tools have been developed to personalize the learning process; however, these tools are often fundamentally limited in their design as decision-tree models. This limits the capabilities of computer-based learning platforms to the resources that can be created by human developers. Large language models (LLMs) hold a unique potential to advance personalized learning due to their ability to quickly generate a near-infinite number of texts. Specifically, LLMs like OpenAI’s Generative Pretrained Transformer (GPT) models have the potential to adapt their outputs in response to a user’s requests and learning preferences. However, the performance of these models is highly dependent on the quality of the system and user prompts developed by the user which, together, define how the model will handle the presented task. As minor changes in their structure can lead to wide differences in performance, these prompts must be carefully designed. To that end, this paper presents a framework through which educational STEM-based textual learning materials can be personalized. Specifically in the framework defined here, GPT-4 first analyzes student choices to identify their learning preferences according to the Felder-Silverman learning style model, and then uses these preferences to personalize STEM-related educational texts. The results of this work are expected to revolutionize the application of large language models in K-12 and post-secondary science education. In many instances the use of these models in education is discouraged, often due to concerns surrounding academic integrity. To address these concerns, we discuss student perceptions and the potential benefits and drawbacks of this personalization framework. The hope is that this work will demonstrate the ways in which LLMs can support education by improving the accessibility of educational texts for a diverse set of learners.