Engineering faculty are weighing how to embrace new tools while preserving the experiences students need to become engineers

---

Leading up to Cal Poly AI Day 
This story is part of a series exploring how the Cal Poly community is putting artificial intelligence into practice while thinking critically about how it should be used. Cal Poly AI Day takes place Thursday, Oct. 22.

---

There is a point in tightening a screw when something simply feels wrong, a subtle change in resistance that tells an experienced engineer to stop before damaging the threads.

Mechanical engineering lecturer Lawrence Domingo uses that example from the machine shop to show how students learn to read the feel and sound of the task itself. In his design courses, that kind of knowledge comes through experience, including the times a task does not go as planned.

At Cal Poly, that kind of embodied learning sits at the heart of Learn by Doing. But as artificial intelligence becomes increasingly capable of writing code, debugging problems and handling technical tasks more quickly, faculty are confronting a new question:

How much of that struggle still needs to belong to the student?

Domingo and three other engineering faculty members in a mentoring quad recently explored that question together in a paper about AI in engineering education. Representing three departments, they bring different perspectives on AI in the classroom but share a concern with what students gain — and potentially lose — as the technology lowers barriers and bypasses some of the mistakes and persistence that shape learning.

Learning From the Wrong Turn

Domingo’s own approach starts with a distinction between courses where students are learning the basics and design courses where they are expected to apply what they know. In those project-based courses, he has encouraged students to use AI to help write code or create early versions of software when coding is only one piece of a larger project.

But he does not want AI to remove every challenge. His goal, he said, is for students to build enough foundational knowledge to become “experts using a tool instead of novices needing a crutch.”

“AI can help us do what we know how to do more quickly, but often the whole point of a design process is to go through it,” Domingo wrote.

Small, reversible failures can help students gain confidence in their ability to solve problems and become more comfortable trying again after something goes wrong. Sketching ideas and building prototypes also give students opportunities to test assumptions, adjust their designs and communicate ideas with teammates — experiences that can be difficult to replicate through AI.

Computer engineering professor John Oliver has seen a similar tension play out in his own classrooms.

His students tend to have strong backgrounds in programming and digital design, giving them the foundation to scrutinize AI-generated code. In circuits and electronics, where their preparation is often weaker, they are more likely to accept AI-generated designs, sometimes leading to flawed implementations.

At the same time, Oliver sees potential for AI to ease some of the frustrations that come with learning engineering. Debugging can mean hours spent tracking down subtle errors, and that grind can discourage some students from continuing in computing.

If AI lightens that burden, Oliver wonders whether computer engineering could become more accessible to students with different strengths. Creative or artistic thinkers, for example, may not see themselves in the traditional image of engineers as people who are simply “good at hard things.”

Yet that same difficulty can also teach.

“The long, sometimes painful process of debugging is not just an obstacle; it is also a formative experience,” Oliver wrote. “It builds resilience, attention to detail, and a deep understanding of systems.”

What Still Belongs to the Student

As AI gets better at producing finished work, faculty are reconsidering how students can show what they understand.

That shift has already pushed Oliver to rethink assessment.

“AI has rendered my weakest pedagogical techniques — such as multiple choice Canvas quizzes — as ineffective,” he wrote, adding that it has pushed him toward more authentic ways for students to show what they have learned.

Andrea Schuman, a computer engineering faculty member who teaches first-year students, has encountered the same issue at the introductory level. By spring 2026, AI agents could fully complete the lab programming assignments in her introductory computer organization course.

“The question then becomes how to incentivize students to learn the fundamentals,” Schuman wrote.

Each student must present a project in person and explain the code to Schuman or a teaching assistant. She also relies on in-class worksheets and paper-based assessments to gauge what students know without AI.

The finished assignment, in other words, is no longer enough evidence of learning.

Materials engineering professor Nicole Johnson-Glauch sees both promise and cost in AI. AI could offer tutoring outside normal hours, including for students balancing classes with jobs or caregiving responsibilities.

But access is only one consideration. “We know that engineering is a series of trade-offs,” she wrote. Johnson-Glauch points to risks that include inaccurate information and the environmental resources needed to support AI systems.

Those costs, she argues, belong alongside AI’s educational benefits when considering how the technology should be used.

The four faculty members do not land in the same place on AI, but their paper reflects a shared focus on what the technology can make easier without replacing experiences students still need for themselves.

At Cal Poly, that conversation reaches directly into the meaning of Learn by Doing. As AI takes a larger role in classrooms and the careers students will enter, faculty are making those choices deliberately, using the technology where it expands opportunity while preserving the experiences that build engineering judgment.

Learn more about Cal Poly AI Day on Oct. 22.