There’s a version of the AI argument that comes up everywhere now. We handed our memory to our phones. We’re now handing over our thinking. Whatever we stop doing for ourselves, we’ll lose the ability to do.
Some of that is worth taking seriously.
The cognitive offloading research is real. There’s a well-known finding that simply having your phone in the room reduces the working memory available to you, even when you don’t touch it.
And there’s the MIT Media Lab study everyone quotes, where people wrote essays either with an AI assistant, with a search engine, or unaided, while wearing EEG caps. The assisted group showed the weakest brain connectivity, remembered less of what they’d written, and felt less ownership of it.
That gets reported as proof that AI makes you stupid. So it’s worth looking at what was actually tested.
Fifty-four people, mostly students, given twenty minutes to produce an SAT-style essay on a topic they had no stake in. Nobody writing about something they knew, with a reason to care whether it came out right. The researchers say plainly that their findings are specific to essay writing in an educational setting and may not generalise to other tasks. They also published a FAQ asking people to stop using words like brain rot and damage about their work.
Then there’s the part almost nobody quotes.
In a later session the researchers swapped the groups over. The people who had written unaided first and were only then given the AI showed higher memory recall and stronger engagement than the group who’d had it from the start.
The order mattered. Bringing a tool to thinking you have already done is not the same act as handing the thinking over at the beginning.
Which is the distinction the whole argument turns on, sitting inside the study most often used to shut it down.
We Have Had This Argument Before
Socrates thought writing would ruin memory. People would stop holding things in their heads because the marks on the page would hold them instead.
He wasn’t entirely wrong. We did stop memorising at the scale people once did. He was wrong about what it cost us, because what writing bought was worth incomparably more.
The same argument arrived with print, with the calculator, with spellcheck. Each time, the objection was that a skill would be lost. Each time, a skill was lost. And each time, the thing that replaced it turned out to be a bigger deal than what went.
That doesn’t settle the current argument. This might be the time it goes the other way. But anyone making the case has to explain why this instance is different, and noticing that people write badly online isn’t that explanation.
Two People Doing Completely Different Things
Here’s where most of these arguments go wrong.
They treat all use of AI as one act.
Someone asks a chatbot for a paragraph, doesn’t read it, and posts it. They’ve done nothing. They don’t know whether it’s true, whether it makes sense, or whether it says what they think. If it’s wrong they won’t notice, because they never engaged with it in the first place. Whatever that is, it isn’t thinking, and I’m not going to defend it.
Now take someone else. They ask for a draft. They read it, disagree with half of it, cut the parts that are flattering rather than true, spot a claim that doesn’t hold, rewrite the ending because it lands in the wrong place, and put back the thing that was theirs and got smoothed away.
That person has done a great deal of thinking. Arguably more than they’d have done alone, because they’ve had to articulate why each thing is wrong rather than just avoiding it.
Both of those people used the same tool. They are not doing the same thing, and an argument that can’t tell them apart isn’t describing anything real.
Execution Is Not Judgement
The difference between them is where the judgement sits.
Maths isn’t my thing. I can add up on paper, I’m just slower at it than I’d like and I don’t enjoy it. But I can look at a tax total, an invoice, or a set of expenses and tell you when something’s wrong, because I know roughly what those numbers should look like. The calculator does the arithmetic. It doesn’t do the noticing.
I write about what I know about. That isn’t an accident. I know when a piece of writing sounds right and when it doesn’t, and I can only know that in the areas where I know the subject. Put me on something I know nothing about and I’d have no way of telling whether what came back was any good.
In both cases the tool supplies the execution and I supply the judgement.
That’s the part nobody can hand over. And it’s the part being ignored when this gets called cheating, because you cannot correct something unless you already know what right looks like.
What Photography Did To Painting
Photography is the sharpest historical case, because the objection was made loudly and the answer took decades to become obvious.
Painters said it would be the end of art. What it ended was painting as the only way to obtain an accurate likeness of a person or a place. That was a real loss to portrait painters and no loss at all to painting, which went off and did other things once it was released from the job of recording.
The skill that disappeared was a technical one. The judgement about what was worth painting, and how, stayed exactly where it was.
Who The Old Requirement Kept Out
There’s a second question underneath this and it gets asked far less often.
If expressing an idea in writing requires a skill that not everyone has had the chance to develop, then plenty of people with something to say have been kept out of the conversation by the execution, not by having nothing to contribute. I’ve met a lot of them. Articulate in person, clear about what they mean, and unable to get it onto a page in a form anyone would take seriously.
Remove that barrier and what you see depends on where you were standing.
If you had the skill, and it was part of what made your contribution count, this looks like standards falling.
If you didn’t, it looks like a door opening.
Both are honest reactions. Only one of them tends to get written down, and it’s usually written by the people who had the skill.
The Objection That Holds
There is one version of the worry I think is right.
All of this depends on the person having judgement in the first place.
If you can’t tell whether the answer is wrong, the tool isn’t extending you, it’s replacing you, and you have no way of knowing when it has failed. Someone who never learned to estimate can’t sanity check a calculator. Someone who has never read anything decent can’t tell whether a draft is any good.
That’s a real problem, and it matters most for people who are learning rather than working.
But the answer to it is teaching judgement, not removing the tool. We faced the same choice with calculators and mostly got it right. You still learn how numbers behave. The calculator handles the labour.
How Judgement Gets Built
Judgement isn’t a fixed quantity you either have or you don’t.
It’s built by making a call, finding out how it went, and adjusting. Most people rarely complete that loop, because the outcome turns up weeks later tangled with everything else, so you seldom find out whether your reasoning was sound or you were lucky.
Anything that helps a person say what they think, look at it, and notice afterwards whether it held up is doing something useful. Not by supplying the answer. By making the reasoning visible enough to check.
That’s the thinking behind Otto, and all of our work, and it’s why what someone gets out of it depends on what they bring. Someone who engages, pushes back and questions what comes back gets a lot. Someone who reads passively and accepts whatever appears gets less. That isn’t a flaw in the design. It’s deliberate. The aim was never to hand someone an answer, it was to help them build the judgement to reach their own, and that only happens if they take part.
The question that carries across all of it is a small one.
Is there another way I could look at this?
Ask it of a calculator’s total, a draft, or a decision you’ve already made, and you’re doing the part that can’t be handed over.
Who Decided Unassisted Was The Only Real Thing?
Which brings me to the question I’d put back to anybody making this argument.
Where did the rule come from that a thing only counts if you did all of it in your own head?
Because delegating the production of words has been perfectly respectable for a very long time, as long as you paid a person to do it. Executives dictated to secretaries. Politicians and chief executives have used speechwriters for a century and nobody suggests the speech isn’t theirs. Ghostwritten memoirs sit in every bookshop under the name of the person who lived the life rather than the person who wrote the sentences. Your solicitor drafts your contract. Your accountant writes the numbers. None of those people are thought to have stopped thinking, and none of them are accused of cheating.
In each case we accepted a simple division. Somebody else does the execution. The person whose name is on it supplies the judgement, the substance and the decisions about what it should say.
That is the arrangement being objected to now. What changed is not the division of labour. What changed is that it stopped being expensive.
For as long as this kind of help had to be hired, it was a marker of having arrived. It signalled that your time was worth more than the typing. Now the same help costs a few pounds a month and is available to a woman writing about her own life at eleven at night after work, and suddenly there’s a purity test that never existed when the assistance came with a salary attached.
I’m not saying everyone objecting is defending their own position. Plenty of people are worried about something real, and the objection that holds is the one about judgement, which I take seriously.
But it’s worth asking of any rule you’ve inherited: who set it, and what were they protecting when they did?
Because a standard that says work only counts when it’s unassisted has never applied evenly. It has always been applied hardest to the people who couldn’t previously afford the help.
About the Author
Karen Ferguson is the founder of MindMotive AI, where she develops Digital Partners designed for specific purposes. Otto, built to help people make sense of situations, relationships and decisions at the point they need to, is one of them.
She has worked as a therapist for 26 years, as a hypnotherapist since 2000 and later as a counsellor, and for nearly 20 of those years as a trainer and supervisor. She has trained more than 1,000 therapists.
MindMotive AI was created in response to a simple question: what happens to the people who never access traditional forms of support? Karen’s work combines decades of professional experience with the deliberate development of Digital Partners designed to give people another way of accessing support when and how they choose.
Karen is a Fellow of the ACCPH (Accredited Counsellors, Coaches, Psychotherapists and Hypnotherapists).
Reference: Kosmyna, N. et al. (2025) Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab. arXiv preprint arXiv:2506.08872.