Novo Nordisk Foundation Center for University Science Education

Learning And Teaching In The Age Of AI

AI is everywhere in education, offering students new ways to learn and do more. At the same time, it can undermine learning, encourage superficial engagement, and foster a classroom culture of cheating and dishonesty. This material is designed for those who want tools to better distinguish between the different influences of AI in education.

A View On Learning

Learning is a weak concept: we cannot see it directly, only infer it from changes in behavior, it cannot be unanimously defined, and no single theory explains it fully. For practical purposes, let us see learning in three overlapping ways:

Participation — becoming increasingly competent in shared practices and discourses.
Acquisition — developing skills and knowledge through anticipation, experimentation, rehearsal, and reflection.
Transformation — shifts in perspectives, beliefs, or values, often linked to emancipation.

These perspectives show how learning is sometimes about doing, sometimes about knowing, and sometimes about becoming. Calling learning a weak concept does not mean it is vague or unimportant; rather, its complexity and multiplicity demand a broad understanding. And we must also remember that affect, safety, and motivation are essential to any sound view of learning.

Four Perspectives On Learning To Help Navigate AI

Technology evolves rapidly, but the fundamental way that learning happens stays the same. In this spotlight we will highlight four perspectives that can make it easier for you to make specific decisions about how to regulate and use these technologies with your students.

1: Hours and effort equals learning

What you do, do a lot, do it with dedication and a commitment, and always do as well as you can – that is what you will learn. Cognitive tools change learning by taking over some activities. And that raises a difficult question: Is this activity still worth learning? Or is it a good thing that machines take care of it? If it is important to master a skill that your tools can make very easy, then think about it like being in a gym. You want to train your muscles, so you do exercises – you don’t ask your robot to do them for you. Learning is a bit like that. It takes repetition and focused attention.

Lab work is the beating heart of chemistry education. But how do students truly learn in the lab?

2: Reflection and theoretical knowledge matters

Just training doesn’t always do the trick. To learn something well, you often need to relate your training and experience to a larger image of the world. But these images are often incomplete or outdated – and updating them can hurt and it requires dedication. It’s not just about practice. You need to relate what you learn to the way you see things and think about the world, and that kind of learning is relatively costly. It requires reflection and often a shift in your worldview.

3: Learning happens in relation to tools and infrastructures

As infrastructures change, so do learning processes. Language itself is a “knowledge infrastructure,” shaping how we think; the same goes for algebra or programming. When such infrastructures are automated—whether solving algebra or generating text by prompt—our thinking is inevitably influenced, though the extent is not fully known.
At the same time, learning new tools and using artifacts creates a bidirectional influence. Tools shape how we think and act by making some things easier and others harder, yet our agency is not surrendered. We adapt, stretch, and repurpose technologies based on our goals. In this way, learning is shaped by a dynamic interplay between person and tool—an influence that develops over time.

4: Learning is a relational and affective phenomenon

It matters who you are with, whether you feel seen and safe, and whether you share goals and purposes with peers in the learning process. Strong human-to-human relations with peers and mentors are critical for building a supportive and efficient learning environment. For teachers, this makes it important to recognize the complete ecology of their classroom: which activities are mediated by technologies, which discussions unfold online, and which take place face-to-face. Understanding this ecology is central to supporting meaningful and empowering learning experiences.

Examples: Writing And Coding With AI

Below are two cases where AI really matters for student work and how the four perspectives might help us understand the consequences for student learning. The first case is writing. The second is coding.

Writing and AI

Writing is everywhere in higher education. Many of us tend to treat well-written work as evidence of high-quality thinking. This is unfair to poor writers who will struggle unnecessarily to prove their skills. But it is also partly true.

Writing is not just a way to present thinking. In some sense, writing is thinking. When students write, they externalize their ideas, test arguments and refine concepts. They build identity as academic thinkers.

Remember, that what you do in detail matters for what you learn. If you spend time shaping sentences and arguments, that is what you learn. If you fully outsource writing to AI, for example by asking for a finished report, much of that learning disappears.

So, the details matter. How do students write with AI? Do they dictate and revise? Do they prompt at a high level? Do they edit sentence by sentence? How much are they engaged in the process? Those choices strongly affect learning.

Coding and AI

AI will change how programming is taught. AI makes it possible to spend less time on syntax. Syntax is hard and machines are getting very good at it. But struggling with syntax also forces deep engagement. You have to understand what the code actually does. If that struggle disappears completely, important learning may disappear with it.

This leads to a key point: what kind of knowledge do students engage with when they prompt AI? Prompting is often short and abstract. A student might prompt like this:

“Make a program that asks how the user feels and changes the background color based on mood.”

That prompt focuses on interaction, not on programming concepts. If the goal is to learn programming, a different prompt might be better. For example:

“Create a function that takes a numerical input and maps it to a color scale.”

Now the focus is on concepts, variables, and functions.

These different ways of working with AI matter. As is the case for both coding and writing, students can avoid conceptual engagement and still get a working solution. And that makes the learning loss hard to see. That is why AI does not just change practice. It changes what students learn and how they learn it. This shows the first two principles – what you do in details matter and knowledge and reflection matter. But it also shows the third principle — that learning happens in relation to an infrastructure. AI is an infrastructural change.