AI in the Engineering Classroom: A Hands-On Experiment in Trust and Accuracy

Mei Sun

The Lightbulb Moment

With generative AI tools like ChatGPT on the rise, I assumed students of the younger generation would embrace AI quickly and extensively, adopting it readily while overlooking its flaws.

Like many educators, I worried they might rely too much on AI, risking misinformation, while those
without AI skills could not be competitive to advance their career. Simply telling them about AI’s strengths and weaknesses wouldn’t be enough—they needed firsthand experience. Then came my lightbulb moment: let them explore AI’s capabilities and limitations in engineering
studies themselves.

In Fall 2024, I introduced an AI-based reflection assignment in my Environmental Engineering Lab course (CEGR3155). As an optional extra-credit task, students submitted a previous assignment to an AI tool of their choice, assessing its ability to calculate averages and standard
deviations, identify outliers, and compare accuracy and precision. Since students had already received feedback on their original work, they could critically evaluate AI’s correctness, completeness, clarity, and relevance.

The AI Experiment: Then and Now

Then: Student reflections revealed an unexpected trend—most were initially skeptical of AI, contrasting my assumption that they had fully adopted it. Many had never used AI tools before and hesitated to rely on them.

Now: After the assignment, students—and I—gained valuable insights:

Trust in AI: Initially doubtful, students were impressed by how often AI answered correctly, showing its potential for assisting in engineering tasks.
Clarity of Explanations: Nearly all students found AI’s responses well-structured and easy to follow, proving AI’s ability to communicate complex concepts effectively.
Accuracy Issues: AI frequently made simple math errors, undermining trust in its numerical reliability.
Handling of Unanswerable Questions: Some students found that AI fabricated data or used flawed logic rather than admitting uncertainty.
Inconsistency: The same AI tool provided varying levels of detail and different answers to different students, highlighting its unpredictability.

The Golden Nugget: A New Approach to AI in Engineering

The biggest success of this assignment was the shift in mindset. Instead of seeing AI as an all-knowing solution or an unreliable tool, students learned to use it critically—as a supplement, not a replacement, for their own problem-solving skills.

They became more adept at verifying calculations, questioning logic, and identifying AI’s limitations. This transformation from passive acceptance to active engagement is precisely the AI literacy needed in engineering education.

This experience has also reshaped my teaching. I now explicitly incorporate AI literacy, encouraging students to challenge AI, test its limits, and cross-check its outputs. Future engineers need to not only use AI but also understand when and how to trust it.

Final Thoughts

AI is neither infallible nor something to fear—it is a tool to be questioned, verified, and refined, much like the problem-solving mindset engineers must cultivate. This journey from skepticism to informed trust has been enlightening for both my students and me.

By engaging with AI critically, we can harness its power while remaining aware of its flaws, ensuring it enhances rather than replaces human expertise in engineering education.