ISE assistant professor earns NSF funding to examine AI’s impact on engineering problem-formulation skills
Published: Oct 26, 2026 1:15 PM
By Dustin Duncan
Casey Eaton, assistant professor of industrial and systems engineering, works with students. Eaton received $200,000 in National Science Foundation funding to study how generative AI affects engineering students’ learning.
When engineering students perform well on take-home assignments but struggle to replicate that performance on exams, what explains the difference?
Casey Eaton, assistant professor in the Department of Industrial and Systems Engineering (ISE), has received $200,000 from the National Science Foundation for a project to find out.
Eaton began asking that question last fall while co-teaching ISE’s introductory systems engineering course, after noticing a disconnect between some students’ take-home work and their exam performance.
“We started to ask this question of what could be explaining this big gap between these abilities,” Eaton said. “One of the hypotheses we came up with is maybe some students are offloading some of those tasks to a generative AI (artificial intelligence) tool, and then they’re not having the same learning experience as if they developed that requirement set themselves.”
Eaton’s project will examine whether and how generative AI affects the way students learn problem formulation, a skill she describes as fundamental to engineering.
“Everybody formulates problems as engineers. It’s setting up what you’re going to solve, what you’re going to create and design and code,” Eaton said. “You can’t engineer or solve a problem if you don’t know what it is. We have to set it up to solve it.”
Part of Eaton’s research will examine cognitive offloading, or shifting some of the mental effort required to complete a task to a tool.
“If I’m having to get in there and figure out what a system should do, think through it, and figure out how all of the subsystems need to work together, that’s really a ‘heavy’ cognitive task,” Eaton said. “What we mean by cognitive offloading here is having a tool, like generative AI, that eases that effort and does some of those difficult, heavy thinking tasks.”
Cognitive offloading can be useful in practice, but learning presents a different question.
“Learning is uncomfortable and really challenging,” Eaton said. “Sometimes it’s the high demand or difficult challenges that gets you the thinking and fundamental skills.”
The project is not intended to determine whether generative AI is good or bad for engineering education. Eaton is studying how the work students do changes when the technology is introduced.
Eaton will examine how changes in students’ problem-formulation processes affect learning and the transfer of those skills. She is also interested in whether the finished product itself remains a reliable way for instructors to evaluate what a student understands.
“We used to view the things students created as a surrogate for their understanding,” Eaton said. “Now I don’t think the product itself is always a good representation of how well we have mastered those skills, at least when we have these other tools.”
Chih-hsuan Wang, professor in the College of Education, will serve as the program evaluator for the project, while Andrew Olwinik, director of Experiential Learning Programs and assistant professor in the School of Engineering and Applied Sciences at the University at Buffalo, will serve as the project mentor.
Eaton is also examining how accurately people evaluate AI-assisted work. Her research will explore whether generative AI can produce material that appears polished and convincing even when the underlying quality is no better — or potentially worse — than something produced without it.
That question extends beyond the classroom. Eaton said some engineering organizations are using generative AI to produce requirements and other engineering artifacts, shifting engineers from creating those materials to reviewing and verifying what the tool produces.
At the same time, colleges are now graduating engineers who have had access to publicly available generative AI tools for a significant portion of their undergraduate education.
“We don’t really know the effects yet,” Eaton said. “I think it’ll be interesting to see both how education changes and then how practice changes once we have engineers who are educated in a different environment.”
Eaton said systems engineering faculty are already responding to generative AI in different ways. Some have eliminated take-home work, while others require students to use the tools as part of their coursework.
“I think we’re still learning and gathering information about generative AI,” Eaton said. “What is the goal of learning for these classes with or without the tool? Does how we teach and evaluate need to change?”
Media Contact: , dzd0065@auburn.edu, 334-844-2336