Artificial intelligence, real learning? Four questions for your use of AI

The concern is well founded. In a study from Microsoft Research and Carnegie Mellon University, 319 knowledge workers described how they used generative AI in their work. The more confidence they had in the tool, the less critical thinking they felt they put into the task. And in a Polish study of experienced doctors performing colonoscopies, their ability to detect precancerous growths in procedures without AI fell – from 28.4 to 22.4 per cent – after a period of working with AI support.

Offloading mental work onto our surroundings is nothing new. We have done it with notes, calculators and satnav. What is new is that we can now offload the entire chain of thought – not just the calculation, but the analysis, the judgement and the wording too.

But as with so much else in the conversation about artificial intelligence, there are two sides to the story. The same technology that can tempt us to skip the thinking can also help us think more, more deeply and more broadly.

Developmental AI use is a choice

For a long time I have been experimenting with developmental AI use on my own brain. I build small RAG models in Claude and NotebookLM on the material I want to learn – not to have it summarised, but to be questioned on it, to have it explained again from a new angle and to have it repeated until it sticks. Right now I spend half an hour a day understanding how the language models I teach about actually work. In that process, AI is my patient sparring partner, not my shortcut around it.

I also use AI to design the learning itself: interactions and patterns of use that force me to apply my knowledge in new ways, combine it with other material and remember it afterwards.

The experience has made one thing clear to me. It is not the tool that decides whether I become wiser. It is the question I ask it. If I ask AI for the answer, I get an answer. If I ask AI to challenge me until I can formulate the answer myself, I get a competence.

Bloom's taxonomy as a framework for AI and learning

Bloom's taxonomy is a good framework for seeing the difference. It was developed by the American educational researcher Benjamin Bloom and colleagues in 1956 and revised in 2001 by Lorin Anderson and David Krathwohl. In the revised version, the taxonomy describes six levels of learning: remember, understand, apply, analyse, evaluate and create.

The taxonomy was written for schools, but it rests on solid knowledge of how people learn – and, as I see it, it applies just as much to adults at work.

Held up against AI, the taxonomy becomes a mirror. Generative AI can now solve tasks at all six levels for us. So the question is not what AI can do, but what it does to our own movement up the steps. If AI takes the steps for us, we stand still. If AI helps us up them, we develop.

Four reflection questions for developmental AI use

I have put together four questions you can ask yourself to find out whether you have used AI in a way that develops you:

1. Has AI saved me routine work – and have I invested the time saved in something that made me wiser or more skilled?
Saving time is not development in itself. It only becomes development when the time goes to the upper steps of the taxonomy rather than to more of the same tasks. I have written about the difference between the friction that drains us and the friction we grow from in the article on the right friction.

2. Has AI given me access to new knowledge, broken knowledge down or processed it so that I can apply it better in my work?
Here AI works on the lower steps – remember, understand, apply – and this is often where AI does most for learning. A heavy field of research can become accessible. A subject you never got round to learning can become understandable.

3. Has automating workflows or parts of processes prompted me to reflect on what a good process looks like in my working life – and what was the answer?
Automation requires someone to describe the process. That is analysis and evaluation, and it is the work that makes us wiser about our own profession. Whoever automates without reflecting learns nothing. Whoever reflects while automating gets to know their profession anew.

4. Since working with AI, have I become curious and developed an appetite for new knowledge I had no eye for before – and if so, what is it?
Curiosity is the most overlooked sign of developmental AI use. When working with AI opens new questions rather than closing them, we are on our way to the top step of the taxonomy: creating something ourselves. I have written about the motivation behind that movement in Is AI making us lazy?.

Reflection questions rather than a checklist

I tried to phrase the questions as a checklist, but I think they work better as reflection questions. A checklist can be ticked off. A reflection question demands an answer you have to formulate yourself – and that in itself is an exercise in developmental use.

The questions can be used on your own, after a week of AI at work. They can also be taken to a team meeting, where the answers often become more interesting than they are individually: who has spent the time saved on becoming more skilled? Who has discovered something new about their own process? And who has become more curious – or less?

That places responsibility where it belongs. AI does not decide whether we lose our skills. Our choices about what we use the technology for do.

Three questions to work with in your own organisation:

• Where do your employees use AI today to skip steps – and where do they use it to climb them?
• Which skills are so central to your professional expertise that they must be actively trained, even if AI can solve the task?
• When did you last talk about what AI has taught you – and not just about what AI has saved you?

Perhaps this is where the conversation about AI and learning should move: from the fear of what we lose to deliberate choices about what we want to become better at.

Sources

• Anderson, L. W. & Krathwohl, D. R. (eds.) (2001). A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives. Longman.
• Bloom, B. S. (ed.) (1956). Taxonomy of Educational Objectives. Handbook I: The Cognitive Domain. David McKay.
• Budzyń, K. et al. (2025). Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study. The Lancet Gastroenterology & Hepatology.
• Lee, H.-P. et al. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. CHI 2025. Microsoft Research and Carnegie Mellon University.