Studying requires friction, but artificial intelligence is removing this friction. That's why the future arena of learning in an AI environment must be reimagined in two distinct places: a 'starship' where AI is omnipresent, and a 'monastery' where AI remains strictly outside.

Universities survived the internet revolution relatively unscathed, despite many ominous reports. Now the AI revolution looms for higher education. As Dean of a business school, I evidently worry about the future of learning. Artificial intelligence promises learning without resistance. Answers appear automatically, texts write themselves, and statistical models analyse complex problems in seconds.

Learning thus seems to become an efficient, automatic and predictably efficient process. But it's a trap. Real learning doesn't happen through passive consumption of information; learning requires productive struggle. Beware the illusion of superficial understanding. Without friction, there can be no understanding; without sweat, there is no mastery.

British historian Niall Ferguson coined the useful metaphor in his essay, The Cloister and the Starship. He suggests that students must move alternately between two zones. In the 'starship', AI is present everywhere. The assignment is to master the newest technology, simulate the world and develop expertise with AI. In the 'Cloister', by contrast, AI is excluded. The cloister sounds like a medieval study, but it's a metaphor for a learning environment where head, hand and conversation are central. There is no AI temptation or distraction.

This separation can occur in space (institutes or buildings with or without AI) or in time (learning periods in which AI is or isn't permitted). Learning encompasses both the duty to learn how to work with AI and the necessity to think critically without AI. Students cannot learn to devise formal proofs by contradiction or learn to multiply matrices if an AI model constantly spoon-feeds them the answers. Falling back on pen and paper is necessary when required. Deep learning always demands a bit of friction. Overcoming that friction – that's where the joy and motivation for learning come from.

No guidelines

Business schools are experimenting extensively with AI to maintain that friction. At Northwestern Kellogg School of Management, the classic case study is being turned on its head. Students no longer receive a ready-made bundle of facts; they must extract information from AI characters and distil a strategy from it. The designers 'fight fire with fire': not to accelerate, but to slow students down and force them into active reasoning. Students conduct more than 50 conversations per character before a solution appears

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

Another example is explicitly separating learning with AI and learning about AI. Many schools, including UNSW Business School, have 'Generative AI in Practice' workshops: being trained to generate new business models, whilst other assignments remain completely AI-free. Compared to students who generate ideas, AI ideas appear more convergent and similar to each other, and human ideas developed with AI assistance also tend to converge. The precision and speed of thinking increase, but variation shrinks. That's the warning for learning: AI smooths the paths, but may make thought-tracks slick. We must move away from slick, back to friction.

If we want to deploy AI without losing 'deep learning', we must redesign the arena. Plan “space periods” for AI exploration, prototypes and data work, and plan “monastery blocks” for core skills: reasoning from a blank page, writing, calculating, modelling, negotiating. The skill is only acquired when you can use it readily without AI help.

Smartphones out, laptops closed

The monastery is not some dream of nostalgia, but a temporary technological necessity for learning: smartphones out, laptops closed, if necessary in an actual monastery hall – for the symbolism. Sometimes friction is not a bug, but a design feature. Which students have both the discipline and perseverance to endure the monastery and the imagination to utilise the starship? Those who only want to sit in the cockpit without ever exploring the terrain on foot don't learn deeply.

Learn more: When AI gets smarter, do humans get dumber?

The counterargument will be predictable: fully banning AI doesn't work, and the separation will be difficult and complex to implement. That's correct, but if we want to teach essential skills, I see no alternative. Deploy AI when the technology can purposefully raise the bar (much) higher: more interactive scenarios, stricter feedback, faster and parallel iterations. Switch off AI when it makes human learning lazy and superficial. The future of learning is not frictionless. On the contrary, AI forces us to deliberately build in friction and learning effort.

Frederik Anseel is Professor of Management and Dean of UNSW Business School. He studies how people and organisations learn and adapt to change, and his research has been published in leading journals such as Journal of Applied Psychology, Journal of Management, American Psychologist, and Psychological Science. A version of this post was first published in De Tijd.