2026 ASEE Annual Conference & Exposition

Can Large Language Models Recognize Neurodiverse Student Challenges and Strengths? A Complementary Study with ChatGPT and Machine Learning

Presented at DSAI-Session 13: Equity, Inclusion, and Diverse Learner Needs in Engineering Education

Large language models (LLMs) such as ChatGPT are increasingly used by college students as on-demand learning companions, producing student-authored language that can surface how learners describe difficulties and successes. For neurodiverse students (e.g., autism, ADHD, specific learning differences), recognizing both challenges and strengths is essential for designing inclusive, strength-based educational technologies. This paper examines whether LLM-based methods can classify short qualitative excerpts related to neurodiverse students' experiences as expressing a challenge, a strength, or both.
As a pilot study, we curate a dataset of 284 literature-grounded excerpts describing neurodivergent experiences in academic contexts and label each excerpt through team-based human coding. We then evaluate (1) ChatGPT as a structured-output classifier in an agreement setting against human annotations, and (2) a supervised machine learning approach (contrastive learning) in a generalization setting using an 80/20 train–test split. ChatGPT achieves 0.71 accuracy (F1 = 0.77) with high mean self-reported confidence (0.91), demonstrating the strongest agreement for challenge-oriented excerpts and lower reliability for strength and mixed-category expressions. The contrastive learning model achieves 0.88 test accuracy, with strong separation between Challenge (0.97 class-wise accuracy) and Strength (0.81), but limited performance on the sparsely represented categories.
Overall, the results indicate that supervised representation learning can more reliably distinguish challenge- and strength-oriented expressions in small qualitative datasets, while LLM-based classification offers a useful but limited reference point for interpretive alignment with human judgment. These findings support the cautious use of AI as an assistive analytic tool for surfacing student-expressed challenges and assets, informing more inclusive and strength-based approaches to engineering education without inferring diagnoses or replacing human judgment.

Authors

Dr. Zhuwei Qin San Francisco State University [biography]

Dr. Zhuwei Qin is currently an assistant professor in the School of Engineering at San Francisco State University (SFSU). His research focuses on designing efficient AI computing algorithms and human-centered AI systems for real-world deployment, with applications such as health technology, rehabilitation, and AI education. Dr. Qin serves as the director of the Mobile and Intelligent Computing Laboratory (MIC Lab) at SFSU. Dr. Qin's research endeavors are dedicated to addressing the inherent challenges related to efficiency and robustness in the practical application of deep learning within real-world environments. A central focus of MIC Lab is to achieve computational acceleration for deep learning, including deep neural networks and large language models, on low-power hardware without reliance on heavy compute or storage resources. His research addresses the efficiency, reliability, and security challenges in the algorithm design and system optimization, as well as application development to create intelligent mobile edge computing systems.
Yiyi Wang San Francisco State University [biography]

Yiyi Wang is an assistant professor of civil engineering at San Francisco State University. In addition to engineering education, her research also focuses on the nexus between mapping, information technology, and transportation and has published in Accident Analysis & Prevention, Journal of Transportation Geography, and Annuals of Regional Science. She served on the Transportation Research Board (TRB) ABJ80 Statistical Analysis committee and the National Cooperative Highway Research Program (NCHRP) panel. She advises the student chapter of the Society of Women Engineers (SWE) at SFSU.
MARY K REQUA http://orcid.org/https://0000-0002-1936-386X San Francisco State University [biography]

Mary K. Requa received her PhD from the UC Berkeley/San Francisco State Joint Doctoral Program in Special Education. She is an associate professor and program coordinator in the mild/moderate support needs program. Her primary teaching responsibilities include courses related to understanding, assessing, and instructing individuals with mild and moderate support needs including methods and materials to promote successful learning outcomes for typical and atypical learners of all ages. Dr. Requa’s primary research interests include pre-service teacher education and inclusive practices. She is currently co-principal investigator on an NSF Improving Undergraduate STEM Education (IUSE) grant. This project aims to develop a prototype of an AI-driven teamwork platform and explore its feasibility in (1) providing neurodiversity-informed content to students and (2) assisting professors triage or mentor students during teamwork and increasing neurodiversity awareness among faculty and students.
Zhenyu Lin San Francisco State University

Note

The full paper will be available to logged in and registered conference attendees once the conference starts on June 21, 2026, and to all visitors after the conference ends on June 24, 2026

« View session