Constructive AI-lignment

In education, constructive alignment is a well-known concept. Every instructor knows that learning objectives, teaching methods, and assessment methods need to be aligned.

What do you want students to learn? How are you going to teach them? And how will you assess whether they have learned it? These are the three key questions instructors continuously ask themselves.

Every course, specialization, and degree program has learning objectives. Some are very specific, while others are more overarching. But one of TU/e’s core objectives is to teach students to become critical thinkers.

Being able to critically reflect on a complex issue, make a well-founded proposal for a possible solution, and afterward critically assess whether a solution actually solves the problem: these are the core competencies of an engineer.

For years, we had a pretty good idea of how to teach students to become critical thinkers. Through challenge-based education, for example. At every stage of the degree program, we try to present students with increasingly complex issues that they can (and must) examine critically.

Where things are currently going wrong is in assessment. Critical reflection is often assessed by requiring students to submit reports. Unfortunately, students no longer seem to want to write reports. In their view, AI can do that too.

They give the issues we present them with to Copilot, Claude, or ChatGPT and copy the answer. So they have learned nothing, but if we cannot prove that the answer came from AI, they will still pass.

There is work to be done here for instructors. We will have to make our assessments AI-proof. I’ll call it constructive AI-lignment.

But students have work to do here too: Stop using AI to “scrape together points.” You learn absolutely nothing if you keep asking AI for help.

And even if you manage to bluff your way through your studies with AI, you will eventually be found out, because employers expect engineers who can think critically. They can use AI themselves; they don’t need you for that!

Boudewijn van Dongen is a full professor of Process Analytics at TU/e. The views expressed in this column are his own.

This column was translated using AI-assisted tools and reviewed by the author and an editor.