2026 ASEE Annual Conference & Exposition

This Work-in-Progress (WIP) paper examines the development and early implementation of an AI-based teaching assistant designed to support student learning in a thermodynamics course. Foundational engineering courses such as thermodynamics remain challenging for many students due to their abstract principles and mathematically intensive problem-solving. This project investigates how an interactive chatbot, powered by generative AI, can provide personalized, on-demand guidance that complements traditional instruction and fosters self-regulated learning [1].
The chatbot prototype, built using a large-language-model API and trained with course-specific materials, worked examples, and guided-question prompts, is designed to help students interpret thermodynamic properties, apply the first and second laws, and analyze real-world systems. Unlike hiring additional teaching assistants, which can be limited by staffing budgets and office-hour constraints, the AI assistant provides 24/7 support, immediate feedback, and scalable accessibility to all students. It adapts to different learning styles by adjusting its response type, offering step-by-step derivations for analytical learners, conceptual explanations for verbal learners, and guided hints or visual breakdowns for problem solvers [2].
Implementation will occur in Spring 2025 across two recitation sections enrolling students from multiple engineering majors. One section will use the AI assistant alongside standard instruction, while the control section will follow traditional methods. Quantitative data will include chatbot usage logs (frequency and query types), quiz and exam scores, and homework completion rates. Qualitative data will be collected via surveys distributed through Canvas and Qualtrics at three intervals (start, midpoint, and end of semester). Surveys will include 5-point Likert-scale items (e.g., “The chatbot improved my understanding of thermodynamic concepts”) and open-ended questions (e.g., “Describe a moment when the AI tool helped you overcome a learning challenge”). Statistical analyses will include independent t-tests and one-way ANOVA to identify significant performance differences, along with correlation analysis to link engagement metrics to learning outcomes [3]. Open-ended responses will be thematically analyzed using NVivo to identify trends in usability, satisfaction, and perceived learning gains.
By documenting the design process, implementation plan, and early data analysis framework, this WIP study aims to provide actionable insights into how AI-driven tools can complement human instruction, enhance engagement, and increase accessibility in engineering education. The findings will guide refinement of the chatbot and inform future large-scale studies on integrating adaptive AI assistants into STEM curricula.
References:
[1] A. C. Graesser and D. S. McNamara, “Self-regulated learning in environments with pedagogical agents that interact in natural language,” Educational Psychologist, 2010.
[2] R. Winkler and D. Soergel, “Conversational agents in education: Supporting active learning in STEM,” Computers & Education, 2016.
[3] D. E. Gonda and A. H. Kerly, “Application of chatbot for computer-science learning: A review of benefits and challenges,” Journal of Education and e-Learning Research, 2021.