Seven in Ten Brazilian Students Already Use AI in School. Almost No One Explained How.

#education #cognition #technology-and-society

August 26, 2026

If you're lazy, click here for the TL;DR

In April 2023, freshly sworn in as Estonia’s Minister of Education and Research, Kristina Kallas had an adviser walk into her office. “Dear Minister, I think we need to do something about AI.” She asked what. He said he wasn’t sure.

AI wasn’t on her agenda. ChatGPT was five months old. Two months later the adviser came back with the same conversation and one new detail: students were already using the thing in schools. And he still didn’t know what to do.

Three years on, Estonia is the first country, and by Kallas’s account still the only one, to hand its high school students a Socratic-style AI tutor.

The story comes from a fourteen-minute talk at TEDxUniversity of Tartu, posted on YouTube by the TEDx Talks channel. Kallas opens by warning that she isn’t going to talk about AI: “I’m going to talk about humans.”

It’s the best defence I’ve heard of a decision nearly every education ministry on the planet is quietly postponing, and it rests on a 2017 report almost nobody cites. It also overreaches in two places.

And it sidesteps a problem she doesn’t have, because she governs 1.3 million people with some of the strongest school results in Europe. That problem is ours, and it doesn’t sink her argument: with the Brazilian numbers in the equation, the argument gets more urgent and much harder to carry out.

A whole country volunteered as the treatment group

Almost a year passed between the adviser’s second visit and any decision. Kallas says the ministry went looking for “smarter people”: round after round of meetings with neuroscientists, cognitive scientists, tech companies and IT people, with a single question. What does AI do to learning.

The conclusion she drew there organises the rest of the talk. “It’s not about technology,” she says. Technology belongs to another ministry; hers is what happens to the brain and to the human capacity to learn.

Out of that consultation came a national AI strategy for schools at the end of 2024, and the programme itself started in February 2025. It’s called AI Leap; in Estonian, TI-hüpe. The name is no accident: it echoes Tiigrihüpe, the “Tiger Leap” announced in 1996, five years after leaving the Soviet orbit, whose official goal was not to buy computers but to connect every school in the country to the internet. That goal was declared met in early 2000. Estonia treats schooling as national infrastructure.

On 1 September 2025, AI Leap gave access to AI tools to some 20,000 students in grades 10 and 11 and to their teachers. This September it expands to vocational education (the step Kallas announces in the talk) and to a fresh cohort of tenth graders: 38,000 more students.

Three implementation details say more than the announcement:

Half the money is private. Companies fund half the budget; the ministry funds the other half.
The ministry doesn’t run it. Operations were handed to a foundation, and Kallas justifies that with enviable bluntness: “there are various reasons why ministry is not the best place to do such innovative initiatives.”
The stated focus isn’t students, it’s teachers. She uses the word “absolute”: most of the effort goes into training staff to use the tool in a way that pushes students toward analysis rather than toward a finished answer.

The tool isn’t the ChatGPT you open on your phone, either. It is, in her words, “a Socratic version of GPT for schools,” built through a research partnership with OpenAI so that it hands back a question instead of a paragraph.

No small country spends half a budget and a ministry’s reputation on a fad. The justification sits in a report from nine years ago.

The 2017 report nobody read

Here is the best number in the talk, and the easiest one to repeat wrong.

In 2017 the OECD published Computers and the Future of Skill Demand. The method is clever: researchers took the questions from PIAAC, the international survey that measures literacy, numeracy and digital problem solving among adults, asked experts which of those questions the machines of the day could already answer, and compared that against how workers actually scored.

The figure the OECD itself puts up front:

Only 13% of workers use these skills on a daily basis with a proficiency that is clearly higher than computers.

In the talk, Kallas compresses this into “two thirds of the workforce” having literacy, numeracy and digital skills lower than computers. The two thirds do come from the report: the figure sits in the preface, signed by the OECD. What the paraphrase swaps is the verb. The text speaks of being at or below computer capability, and the number it puts in the shop window is the flip side of the ledger, the 13% clearly above. Tying with the machine and losing to the machine are different professional situations.

Fix the sentence and the shock survives intact, but the date it lives in is not the one it appears to be. Those 13% are not a snapshot of the machines of 2017. They are the OECD’s projection for what computers would be capable of in 2026. For the machines that actually existed in 2016, the same report says the opposite: 44% of workers were clearly above them, and 31% level with them.

Read that again, slowly. In 2017, five years before ChatGPT existed, the OECD pointed at the year you are reading this in and said that two out of three workers would be, in that year, at best tied with a machine. The report did not age well or badly. It set a date, and the date is now.

The question Kallas draws from it is uncomfortable even for her, and she asks it without hedging: if there are computers that can do this, what is education for? “Should we actually give up learning altogether?”

The tree she draws and doesn’t name

The answer shows up on a slide with six rungs: remember, understand, apply, analyse, evaluate, create. Kallas calls it a “cognitive tree” and never gives it a name. It’s the revised Bloom’s taxonomy, the most widely used lesson-planning scheme in the world.

Where that ladder comes from

The slide reproduces the 2001 revision by Lorin Anderson and David Krathwohl of the classification Benjamin Bloom and colleagues proposed in 1956. It swapped nouns for verbs (“knowledge” became “remember”) and flipped the top two rungs: in the original, synthesis came before evaluation; in the revision, creating became the last step.

Before the ladder, she splits the brain into two modes. The lower cognitive mode is memory and automatism: you learn to ride a bicycle once and don’t relearn it every time you get on.

The example she picks to show what that costs is the best moment of the talk. A stereotype, Kallas says, is the lower mode working exactly as designed: “we learned that Italians are loud, and then we expect every Italian to be loud every time we meet them, because that’s how our brain operates.” The higher mode is the opposite: relearning in every new situation, using what you know to face what you don’t.

The argument runs straight from there. The bottom three rungs are where school systems have been stopping for two centuries: you memorise that the Second World War began in 1939, you understand what that means, you repeat it on the test when asked. The top three get reserved for the PhD: analysing why it began in 1939, who the actors were, what motives they had; evaluating ethically; creating something new.

And here Kallas is more interesting than the simplified version doing the rounds. She doesn’t say AI is only good at the bottom rungs. She says it can handle almost the whole ladder, “with a few exceptions”: in her view the models are still weak at ethical evaluation and at creation, because they only create “out of statistical probability cases.” In that sense, she argues, “human brain is still unique.”

It’s a strong and debatable claim. But the practical conclusion doesn’t depend on it: a school that stops at “apply” is training people to compete in the range where the machine always wins. Kallas wants analysis, evaluation and creation starting in grade 7, with twelve- and thirteen-year-olds.

The mechanism she offers for why that is hard, though, doesn’t hold. She says the brain prefers the lower mode because it burns less energy, that it is “our biggest organ”, and that anyone who crams for a hard exam “will be dead next day physically”, more spent than after a marathon. The brain is not the biggest organ, and the additional cost of hard thinking is small: it already burns about a fifth of the body’s resting metabolic rate all the time, and a stretch of mental effort takes a little more brainpower than usual, but not much more.

The behaviour she describes is real: the brain is a cognitive miser. But the price isn’t calories. It’s attention, motivation and tolerance for the discomfort of not knowing. That changes what you do about it: you can’t feed anyone up into higher-order thinking, you can only design tasks that don’t work without it.

Which leaves the second place where she overreaches. That one is bigger.

The printing press never taught anyone to read

Kallas uses a specific word, repeatedly: evolution. AI puts us under “evolutionary pressure”, she says, the way Gutenberg’s press did, and back then “all the humans actually evolved in a capacity to learn reading.” Today’s pressure, she insists, isn’t physical but mental, and the organ under pressure is the brain: “we have to evolve.”

That isn’t what happened. Gutenberg printed his Bible around 1450. Universal literacy only appeared in the West four centuries later, and not because books existed: because compulsory schooling existed, along with paid teachers and a state willing to enforce attendance.

Brazil hasn’t finished the job. IBGE, the national statistics bureau, counted 8.4 million illiterate people aged 15 or over in 2025, the first time the rate fell below 5% in the series that begins in 2016.

Learning to read does change the brain, and that part is well documented: comparing literate with illiterate adults shows that reading reorganises the visual pathway of the cortex, hijacking for letters a region that used to respond more to faces. But it happens within a single lifetime, recycling circuitry that already existed. Five centuries of printing didn’t move a single base pair.

Calling that evolution is generous to the process and cruel to whoever has to carry it out. Evolution is automatic; public policy is not. If the comparison with the press holds, it says the opposite of what the metaphor implies: the cognitive leap after the last great technological shock took centuries, cost a fortune and depended on institutions.

Which brings us to the part where Estonia plays the cards it has, and we play ours.

Translating this for 7.4 million

The entire AI Leap pilot has 20,000 students. Brazil has 7.4 million in-person high school enrolments, according to the 2025 School Census run by Inep, the federal education statistics institute: 368 times more. What in Estonia is a cohort, here is a country.

Scale is the smaller problem. The starting point is the real one.

Indicator (PISA 2022)	Brazil	Estonia	OECD average
Mathematics	379	510	472
Reading	410	511	476
Creative thinking (0 to 60)	23	36	33

In PISA 2022, 73% of Brazilian 15-year-olds scored below Level 2 in mathematics, the floor the OECD treats as the minimum for adult life. Across the organisation’s member countries the figure was 31%. In reading, half of ours fell below the floor.

The most uncomfortable line in the table is the last one. In 2022 PISA measured creative thinking for the first time, the top rung of Bloom’s ladder, precisely the territory Kallas points to as still human. Brazil scored 23 out of 60 on a test taken in 64 countries. The OECD does not publish rankings, it publishes scores with a margin of error; within that margin, Brazil falls into a cluster of countries between 44th and 53rd place. Estonia scored 36.

The adult side is no better: Inaf 2024, Brazil’s functional literacy indicator, found 29% of Brazilians aged 15 to 64 to be functionally illiterate, the same level as 2018. Among young people aged 15 to 29 it got worse: from 14% to 16%.

Now the number that closes the argument. TIC Educação 2024, the annual survey of technology in schools by Cetic.br, the research arm of Brazil’s internet governance body, interviewed more than ten thousand people in a thousand schools: seven in ten high school students who use the internet already turn to generative AI for schoolwork, and only 32% received any guidance from their school on how to do it. And the share of teachers who had taken part in continuing training on digital technologies in the twelve months before the survey fell from 65% in 2021 to 54% in 2024; in municipal school systems, from 62% to 43%. The steepest part of that drop had already happened by 2022.

While Estonia was deciding what students should do with AI, we decided where they may not hold a phone: Law 15,100, from January 2025, restricted personal devices across all of basic education. It’s a defensible law and probably a good one. It just doesn’t answer the same question. Brazil’s education ministry released in April of this year a guidance document, Artificial Intelligence in Basic Education: a serious document, three years after that Estonian adviser knocked on the minister’s door.

So the conclusion is that Brazil has no foundation for this bet? It’s almost the opposite.

Where I might be wrong

The best experiment on the subject was run with around a thousand high school students at a school in Turkey and published in PNAS in 2025. Three groups studied mathematics: no AI, a plain GPT-4, and a “GPT Tutor” configured to give hints written by teachers instead of answers.

During the exercises, everyone with AI did better: 48% above control with plain GPT, 127% with the tutor. Then came the exam, with no AI in hand. The students who had used plain GPT scored 17% worse than the control group. The ones who had used the guardrailed tutor lost nothing.

Read that result the way it hurts. The version without guardrails, the one that degrades learning, is exactly the one seven in ten Brazilian teenagers already have open on their phones, free, with no guidance at all. The version that did no harm is the one Estonia went and built with OpenAI and shipped with teacher training attached.

Brazil isn’t sitting out the experiment. It’s inside it, in the wrong arm, without informed consent and with nobody measuring.

None of which proves the Estonian bet works. And it’s only fair to record that the fear driving Kallas comes in the very vocabulary I’m about to qualify: what the ministry was most afraid of, she says, was that doing nothing would bring “brain rot and cognitive offloading”, with the education system coasting on inertia through the same old mental processes.

The trouble is that the evidence behind that vocabulary is thin. The most cited study on cognitive offloading, the idea that outsourcing thinking atrophies the capacity to think, is correlational and based on self-reporting: it shows association, not cause. The MIT work that made headlines under the name “cognitive debt” has 54 participants, is a preprint without peer review, and has already drawn a technical commentary disputing its analysis. “Brain rot” is an internet phrase, not a diagnosis. And one experiment at one Turkish school is not the world.

Kallas admits the hole herself, and it’s the most trustworthy sentence in the fourteen minutes: “we also don’t know much what is going on. We are really observing and learning as we go along because there are no clear answers or fixed answers yet.” She says the country runs scientific monitoring of the programme, and in the next breath that there is “very little science” on the subject. Which is why, she adds, Estonia isn’t asking for applause: it’s asking everyone else to come along and learn with them.

What’s left

Three years ago, an adviser walked into a minister’s office and twice said he didn’t know what to do. Today that doubt has become an entire country volunteering itself as its own treatment group.

Seven in ten students already use AI, no plan, no guardrails, no trained teachers — that has already happened. Our problem was never not knowing. It was pretending we’re still deciding.

When AI Knows Everything, What Should Humans Learn? | Kristina Kallas | TEDxUniversity of Tartu — TEDx Talks