>


>

What AI Won't Be Able to Deliver for You

What AI Won't Be Able to Deliver for You

MIT measured what happens in the brain when writing with AI: 83% could not cite a single sentence from their own essay. Why judgment cannot be delegated, and what universities should demand instead.

Gabriel Serrano

In 2025, a team from the MIT Media Lab did something no one had done before: they looked inside people's brains while they were writing with artificial intelligence.

The experiment was straightforward to describe. Fifty-four participants, divided into three groups, wrote essays over the course of four months. The first group wrote with ChatGPT. The second, with a web search engine and nothing else. The third, unassisted: using only their minds. All wore a thirty-two-electrode cap measuring their brain activity in real time, accompanied by interviews after each session.

The findings that emerged have become one of the most cited studies of the year, and they warrant a careful reading, because they do not say what the headlines claimed they did.

The Debt

The first finding is almost geometric. Brain connectivity—the degree to which different brain regions work together during a task—systematically decreased the more external assistance was present. Those writing unaided displayed the strongest and most extensive networks. Those using a search engine showed intermediate engagement. Those using AI exhibited the weakest coupling of the three.

[CHART 1: g1-conectividad-mit.png]

So far, nothing scandalous: if someone else does the heavy lifting, you exert less effort. The compelling part came later. In the fourth session, the researchers turned the tables: those who had spent months writing with AI were required to write on their own. And their brains did not return to the level of those who had always written unaided. They remained underactivated, like a muscle that has spent the entire season on the bench. The authors coined a term for this phenomenon: cognitive debt. Today's discount on effort is paid back with interest tomorrow.

The 83%

Yet the metric that truly matters to anyone dedicated to teaching is quite different, and it carries a statistical cruelty that is hard to surpass.

Upon finishing their writing, the researchers asked each participant to quote a sentence from their own essay. The very one they had turned in just minutes earlier. Among those who had written unaided, nearly everyone did so without difficulty. Among those who had written with AI, 83% were unable to quote a single sentence.

[CHART 2: g2-memoria-83.png]

Consider this for a moment. It is not that they forgot details: it is that the text bearing their name had never truly passed through them. And the participants themselves noticed it. When asked whether the essay was theirs, those who wrote unaided claimed authorship almost unanimously; those who used AI hesitated, qualified their answers, and shared the credit with the machine. The study refers to this as a fragmented sense of ownership. Put plainly: they signed something that was not entirely theirs, and they knew it.

Here lies the crux of the matter, and it ought to be framed with precision. The study does not prove that AI is inherently bad. It proves that writing is the very mechanism through which ideas become our own. Memory is consolidated by choosing each word, by grappling with every sentence. Whoever delegates writing does not simply skip a chore: they forfeit ownership.

This Is Not About a Hypothetical Future

It is worth stating plainly that we are not discussing an approaching threat. We are describing today's classroom.

In Spain, 89% of university students admit to using artificial intelligence in their academic work, and nine out of ten use it for studying, according to surveys by the CYD Foundation. Around four out of ten use it daily. According to Eurostat, Spain is the European Union country where AI is used most for educational purposes: twenty percentage points above the average.

[CHART 3: g3-uso-ia-universitarios.png]

And here, one must resist the temptation of moral panic. That 89% is not a generation of cheaters. It is a generation with an extraordinary tool in its pocket and, all too often, facing assignments that demand nothing the tool cannot already provide. When a machine can do the work, the implicit message is that the work never required the person. Students have understood that message better than anyone.

Why Surveillance Is Not the Answer

The immediate reaction across half the academic world was technological: detectors. Software promising to distinguish human text from generated text. And here, the data is unforgiving.

A team from Stanford evaluated seven detectors against exam essays written by non-native English speakers: the average false-positive rate was 61.3%. More than nine out of ten essays were flagged as suspicious by at least one detector. With native writers, there was virtually no error. The reason is both technical and perverse: detectors penalize simple vocabulary and straightforward syntax—which is to say, they punish precisely those writing in their second language.

Vanderbilt University disabled the detector in its anti-plagiarism platform in 2023 following a simple calculation anyone could make: even accepting a 1% false-positive rate, when applied to the 75,000 papers it grades annually, it yields roughly 750 wrongful accusations per year. Seven hundred and fifty innocent students per academic year.

The detection arms race is lost. And lost races are best abandoned early, before the collateral damage falls once again on the most vulnerable.

What Serious Universities Are Doing

If surveillance does not work, what remains is to reconsider what we ask for. And that is precisely what is happening.

In the United States, sales of handwritten exam booklets—the classic blue books—have surged by 30% at Texas A&M, 50% at the University of Florida, and 80% at Berkeley over the course of two academic years. A professor in Wisconsin explained why he returned to paper: in a lecture of four hundred students, he received about fifty final exams showing clear evidence of ChatGPT; a few even had the logo pasted in. But what is telling is what he now looks for in handwritten exams: not the most polished phrasing, but the ability to connect the course concepts with one another and articulate why they matter.

[CHART 4: g4-blue-books.png]

The handwritten exam is neither the sole answer nor the best solution for everything: timed writing sacrifices research and revision, which are also forms of thinking. Yet the direction of this shift is unmistakable, and it is the right one: in the presence of an executing machine, rigorous assessment migrates toward that which cannot be delegated. The first-person perspective. The oral defense. Discretion applied to one's own life. That which is built alongside others.

Writing Is Retaining

That is why a final humanities capstone should be written without artificial intelligence. Not out of nostalgia, nor as a punishment, nor out of distrust toward a technology that is here to stay and that we must learn to use wisely.

It is written without AI because the final essay is not merely a deliverable product: it is the vehicle through which the journey becomes yours. What is evaluated is not the elegance of the prose. It is the presence of a mind behind it: independent judgment, intellectual integrity, and a person who can quote their own sentences because they chose them one by one.

Artificial intelligence will only write better and better. That battle is settled, and that is fine. The question that endures is another, and it is the only one that ever truly mattered: when your education is complete, will there be someone there? Someone with something genuinely their own to say, and the words to say it?

That is something no one else can hand to you. Not even the machine.

Cover image generated with artificial intelligence.


About Us


Discover the school you
never had.

Get Nomba Corporate →


What if this were your next year?


What if this were your next year?


What if this were your next year?


A journey of ideas, conversations, and twenty minutes a day to keep growing.


Continue reading


Continue reading


Continue reading

The Humanities Today

Tolstoy in a Business Meeting: Why Literature Develops Better Executives

Tolstoy in a Business Meeting: Why Literature Develops Better Executives

Tolstoy in a Business Meeting: Why Literature Develops Better Executives

The Russian novel teaches about committees what no MBA can ever explain to you.

Applied Philosophy

The Adult on Autopilot

The Adult on Autopilot

The Adult on Autopilot

The week unfolds, everything functions, yet you were not entirely there. The ancients named that way of living millennia ago, and one does not escape it with a checklist of habits.

The Era

Your degree expires; your judgment does not

Your degree expires; your judgment does not

Your degree expires; your judgment does not

Everyone repeats that skills expire in five years. We went looking for the source, and it is nowhere to be found. What does endure, however, is sound judgment: why the university should be a community, not merely an event.

The Era

Employability in the Age of AI: Hiring for Judgment

Employability in the Age of AI: Hiring for Judgment

Employability in the Age of AI: Hiring for Judgment

Youth employment drops by 13% in positions most exposed to AI, while that of seasoned professionals rises. What protects the latter, and what companies are already demanding: judgment.

Cultural Leadership

The Leader Companies Can No Longer Find: Profile and Why It Is in Short Supply

The Leader Companies Can No Longer Find: Profile and Why It Is in Short Supply

The Leader Companies Can No Longer Find: Profile and Why It Is in Short Supply

Technical talent is abundant. A profile with cultural judgment is the scarce asset.

This site uses cookies
We use cookies to enhance your experience, measure performance, and personalize content. Cookie Policy