Some thoughts on AI in education

I have been advocating for AI usage in education for a while, mainly because of how much I have personally benefited from AI in my own study.

However, after teaching this term, I am not so sure about this belief anymore.

This term, I have been tutoring a self-directed language learning course and a data science for chemical engineering course. Towards the end of the term, students gave presentations, and I noticed that many were using AI to generate not only the content, but also the slides and probably the scripts.

Because students in each course were presenting on similar themes, the pattern became obvious after watching presentation after presentation. A lot of the presentations started to look very similar.

Of course, AI use in student work is not new. In previous terms, we had already seen it in essay writing, and there is now at least some guidance around that. What felt different this term was seeing GenAI move much more visibly into presentations.

Seeing an AI-generated presentation delivered in real time also feels different from reading an AI-assisted essay. When the slides, wording, and possibly even the script all seem AI-generated, it can feel much more concerning.

At first glance, AI-generated slides can look very polished. But on closer examination, they are often quite generic.

There are lots of text boxes, shapes, and visual elements without any obvious reason for them to be there. The wording can feel overly dramatic, with patterns such as “It is not …, but …” or “This changed everything …” appearing again and again.

And then there is the colour coding. ChatGPT really seems to like pastel colour palettes.

(The slide below is what I asked ChatGPT to generate from the content above, and I think the style captures these patterns pretty well!)

Of course, none of these things is necessarily a problem on its own. There is nothing wrong with pastel colours or dramatic language. What concerns me more is when big conclusions appear without the personalised details underneath them.

And this becomes even more visible during the Q&A, when students sometimes struggle to answer questions about what they have just presented.

If something really changed your thinking, what happened? What did you struggle with? What surprised you? What exactly changed your mind?

Those details matter. They are often where the learning actually is.

A spectrum of AI use

At the same time, I noticed something else that was quite interesting.

There seems to be a spectrum in terms of how students use GenAI and how much effort they put into using it. Interestingly, the work that felt most genuine often came from students at the two ends of this spectrum.

At one end were students who used AI extensively, but clearly put a lot of effort into interacting with it. You could tell that they had thought about the content. Their slides contained personalised details, and when we asked questions, they could explain their thinking beyond what was written on the slides.

At the other end were students who seemed not to use AI at all. Their work might not have been as polished, but you could also feel that it came from them.

What concerns me most is the group somewhere in the middle: students who use GenAI, but put very little effort into engaging with it. They may get AI to generate the content, slides, or script without really working through the ideas themselves. AI seems to become more of a crutch than a tool for learning.

So the part that concerns me most is not simply that students are using AI. It is whether they are using it mindfully, and whether they understand how the way they use it might affect their own learning.

So how should we guide students in using AI?

To be completely honest …I’m not sure.

I haven’t yet found an approach that feels immediately useful in the classroom. Many of the resources I have come across, including university teaching resources, are useful at the level of principles and policy, but I have found fewer hands-on examples of what we can actually do with students.

Different courses have also taken very different approaches to AI, depending on the subject and learning objectives. Some restrict AI use quite heavily. But I don’t think a blanket restriction is necessarily the right solution for the courses I teach.

In fact, some students have told us that they really appreciate being allowed to use AI. It has helped them overcome barriers and do things they might otherwise have struggled to move forward with. Importantly, some of these students also engage deeply with AI: they ask questions, iterate, check the answers, and use it to develop their own thinking.

I think this is exactly the kind of AI use we want to encourage.

From my understanding, I believe the key is mindfulness of using AI. I hope to give students not only the freedom to use AI, but also help them understand what that freedom means.

Something I want to try next term

Letting students experience the same kind of activity in three different ways.

First, they could complete it without AI at all, perhaps using only pen and paper.

Then, they could use AI while trusting its output quite blindly—give it the task, take the answer, and use what it produces.

Finally, they could use AI more as a co-pilot: asking questions, checking its reasoning, challenging its answers, asking follow-up questions, and making their own decisions.

The point would not be to show that one approach is always better than the others. Instead, I would like students to compare the experiences for themselves.

I don’t expect an activity like this to suddenly stop students from using AI to complete their work. Some students may still choose to trust AI blindly, and ultimately that is their choice.

But at least the choice becomes a more informed one.