Today, we’ll be looking at the opportunities and challenges of AI in education. Let’s start by taking a step back and thinking about reading and writing.
Building blocks of learning
Putting aside the best thing about reading – the sheer joy of it – it’s one of the fundamental building blocks of learning.
It builds vocabulary – essential for fluency and comprehension.
It builds background knowledge – essential for learning.
It builds language skills – both written and oral.
Crucially, it builds intellectual stamina – the very opposite of the attentional havoc wreaked by watching 25-second videos for hours every day. And attention is fundamental to learning.
What about writing?
Occasionally, writing merely records thoughts that emerge fully formed.
But we all know that writing is a tool not just for recording but also for thinking.
Committing things to paper extends our working memory and allows us to develop more complicated ideas.
The writer of anything of substance must read widely, produce a conceptual framework, find the right words, wrestle with syntax and organise the logic of their arguments.
The precision and clarity required to commit this to paper quickly exposes fuzzy thinking – and sometimes forces us to abandon our first thoughts.
As the saying goes: “How do I know what I think until I see what I write?”
Writing is thinking.
Both reading and writing give us the tools to think better. And mastery of the basics is the gateway to higher thinking.
We all recognise that there’s a difference between the quality of understanding and analysis of beginners and experts. But there’s no path to expertise that doesn’t pass through the basics first. It doesn’t matter what the field is – from English literature to physics: the highest thinking skills never come without mastery of the basics.
AI as a shortcut
So: reading, writing and foundational knowledge are all tools to think better. What does that mean for AI?
Until about five minutes ago, reading, writing and thinking at a basic level were inescapable. Finding and processing information required us to read widely. The only way to produce writing was to sit down and wrestle with the language and the logic. You had to think your way into any problem.
But no more.
Do you want an overview of the main arguments in a 300-page brief? Do you need an analysis of the impact of the French revolution or wealth inequality in Western democracies? No reading or writing or thinking required – just feed in the prompt.
Is that good or bad? That depends on whether your goal is process or product.
Process vs Product
In many contexts, the goal is to generate a suitable product. A civil servant needs a proposal addressing child poverty. A lawyer needs litigation options for a client. A marketer needs data analysis for a brand campaign. The objective is a product; the person behind it is irrelevant.
But in education, it’s the opposite.
Pupils aren’t asked to solve maths problems because the teacher doesn’t know the answer. They aren’t asked to write essays on Macbeth because the world needs more 1500-word analyses by teenagers. The product is irrelevant. We ask them to write because the process develops their minds.
And here we encounter the problem of cognitive off-loading.
Cognitive Off-loading
Thinking is such hard work that it’s something we try to off-load where we can.
There are different tools:
Using pen and paper for a complicated multiplication means I don’t have to hold all the steps in my mind at once. This form of cognitive off-loading sidesteps the tight limits on working memory and makes complex calculations much easier.
Trying to memorise my shopping list is difficult, so I write a list.
Navigating to a new place is hard – so I use a satnav.
Cognitive off-loading helps us do more complex things.
But the effect will also be that I absorb less:
I write a list and make no effort to hold the items in my mind.
I use the satnav, then have no idea how to get back.
No loss, surely. And it’s tempting to assume that off-loading the basic thinking will free up our minds for more complex analysis. The key is that I’m not trying to build knowledge based on my shopping list or my satnav destination. It’s a simple task quickly forgotten.
But for critical thinking there is no workaround: basic skills are the pathway to higher skills.
If you don’t know the French revolution in great depth, you’re not going to generate interesting analysis of how it affected the politics of Europe and America. If your mathematical skills aren’t robust, you’re not going to be able to make creative connections which help you solve a problem in rocket science.
Effects of AI
AI has enormous benefits:
It gives access to vast stores of data.
It can assemble and arrange that data instantaneously.
It saves us from reading and writing and thinking for ourselves.
And AI has enormous costs.
Saved from the effort of reading, we lose fluency and knowledge and attention span.
Saved from the effort of writing, we are left with fuzzy thinking and superficial understanding.
Saved from the effort of rehearsing basic skills, we lose them.
AI thus undermines the essential building blocks for critical analysis; sophisticated communication; and originality of thought.
Schools and workplaces
Is this a problem? We come back to the purpose.
In education, it’s a big problem. Children need to read and write and master the basics.
Most of the obvious uses of AI interfere with this so we are creating defined spaces in our pupils’ daily lives where AI cannot undermine their learning. But there are use cases where it can help: for instance, to identify errors which the pupil can self-correct. Or to generate opposing arguments which stress-test their thesis. We are developing some interesting applications over time.
Is AI a problem in the workplace? I think that’s more ambiguous – because workplaces tend to have a short-term need (the product, where AI can really help); and a long-term need (developing the skills of their people – which are compromised by over-reliance on AI). Senior leaders will need a clear long-term vision for their organisations in order to strike the appropriate balance.
Over-reliance on AI
The biggest challenge, in my view, is the temptation to overuse AI. One of my favourite education thinkers, Daisy Christodoulou, compares the problem to the challenge of obesity. We have always been wired to seek calorie-dense food, but it wasn’t until we created an obesogenic environment – with unlimited access to high-calorie foods and low physical demands – that obesity emerged as a problem at scale.
In the same way, we are wired to avoid the hard thinking which makes us smarter but our laziness has been curbed by circumstances, in which serious reading and writing and thinking were unavoidable. AI is removing those built-in challenges and making our daily environment, in Daisy’s words, ‘stupidogenic’.
Evidence base
It’s early days, so the evidence is still rather thin. But the first studies are coming through, generally with exactly the results one would predict.
Early studies are showing, for instance, that working with AI diminishes people’s understanding of their own work. The authors suggest that excessive use of AI will result in cognitive atrophy.
Another study found that less AI meant more critical thinking, and greater creativity. They also found that the automation of routine tasks removed the opportunity for workers to practise using their judgment, leaving them unprepared to deal with exceptional cases. This is something of a blow to the dream that AI can do the grunt work, leaving the humans free to focus on the hard cases.
And studies are repeatedly finding that AI produces the greatest improvement where it is used by experts with a deep skills base honed over many years without the use of AI. In our context, that means the teachers. We are already using it in interesting ways and developing more behind the scenes. But that’s for another day.
Guard rails
I think of AI like those running shoes with carbon fibre springs in the sole. If someone like me trains in them daily, my legs will adjust to the extra power and become weaker. But if someone like Mo Farah pounds out the miles in normal shoes for years, and then pulls on the carbon fibre shoes on race day, he’ll get a world-beating boost.
How we learn is determined by the fundamental structures of the brain, so whilst AI is a new technology, the guard rails on its educational uses are clear. The most reliable way to use it is to introduce it later (on race day) – when the foundations of knowledge are solid and thinking skills are fluent.
There’s no shortcut: learning deeply requires friction and effort. Schools must not copy workplaces by using AI to produce a better product with as little effort as possible.
That effort you just eliminated?
That’s the learning.
