What Kind of School Are We Building? Beyond Access: Toward Advanced Competence

In 2012, 0.9% of Kazakhstani fifteen-year-olds reached the top PISA levels in mathematics[1]. In 2025, 1% did. Over those thirteen years almost everything around the school changed. State built new schools, bought equipment, rewrote curricula, trained teachers, brought in digital platforms. Yet one percent is still one percent.

Talk about Kazakhstan’s schools has long run in one groove: what is missing. Not enough teachers. Not enough materials, discipline, parental attention, motivation in children. The PISA-2025 results for Kazakhstan show something else[2]. Most teenagers report their teacher’s support. School discipline here is better than the OECD average. There is less bullying. Shortages of staff and teaching materials have fallen noticeably. The gap in results between children from richer and poorer families is small.

All of this is very good news. So why does a school in which so much already works convert what it has so weakly into a child’s ability to read, to reason, to solve unfamiliar problems?

In Kazakhstan, 22,680 fifteen-year-olds from 643 schools took part in PISA-2025. PISA is not an examination on the school curriculum[3]. The student is given an unfamiliar text, task, or situation. The test asks whether he can use what he has been taught: grasp the meaning, compare information, choose a method, apply familiar knowledge in new conditions. The value of PISA therefore lies in how far a student can use knowledge without being told how, without a ready-made instruction.

Results change less than the school itself

Seen in retrospect, the numbers for Kazakhstan are quite harsh. In 2012 our fifteen-year-olds averaged 432 points in mathematics, 393 in reading, and 425 in science. In 2025 the figures were 414, 385, and 420. Mathematics stands out. In 2012, 45.2% of Kazakhstani schoolchildren were below the baseline second level. Today there are more of them, about 56%. Other results matter even more.

About 1% of students in Kazakhstan reach the top PISA levels in mathematics, against roughly 8% on OECD average. In science it is about 0.5% against more than 7%. In reading the upper layer is also extremely thin. For the country this is no longer simply a question of school scores. Kazakhstan needs a diversified economy, needs to develop science, technological production, artificial intelligence. But a small group of Olympiad winners and the graduates of a few special schools will not be enough for such an economy. It needs a broad layer of people who can work with complex knowledge, retrain, recognize an unfamiliar problem, and look for a solution where no ready algorithm exists.

Where will that group come from in ten or fifteen years?

PISA-2025 also brings very good news. The gap in science between students from the top and bottom quarters of the socio-economic scale is 41 points in Kazakhstan. On OECD average it is about 85. Socio-economic standing explains about 5% of the variation in our students’ results, against roughly 12% in the OECD. For a society this is a serious advantage, and it has to be preserved. Yes, our schools’ separate children by social origin comparatively little. At the same time, they very rarely bring them to the highest levels of competence. This is equality with a low ceiling.

A thoughtful reader will object that the picture is overstated. That PISA gives an averaged view across 643 schools, and the strong layer simply dissolves in it. That the ceiling is not low at all, the strong students are there, gathered in Nazarbayev intellectual schools, the Bilim-Innovation lyceums, the physics-and-mathematics schools, and in the sample these schools are too small a share to move the average. That judging a country’s potential by its mean score is like judging elite sport by fitness-badge standards. And, finally, that concentrating strong students in a few schools is not a flaw at all but a working model. A middle-income country has learned to lift a few students above the ceiling; the question now is how to raise it for everyone? Singapore[4], China[5], and Vietnam[6] did exactly this in their early stages. This objection can be accepted almost in full.

Averaging does smooth things over. The strong layer really exists, and graduates of these schools enter strong universities around the world. But then the question should be put more precisely. Why has the ceiling not risen in thirteen years? A selective schools reproduce itself; that is how they are built. Since, they  select a child who is already strong and brings him through. It does not produce strong students where none were selected. So however, many such schools there are, the ceiling of the mass result does not move.

It is worth remembering that PISA is not hunting for a few geniuses in each country. Levels 5 and 6, the highest ones, measure the ability to work with a complex unfamiliar task, to choose a strategy independently, to carry knowledge across situations, to understand difficult texts, to judge information critically. Abilities of this kind are gradually becoming necessary not only to a researcher or an IT engineer.  In an economy where technologies and professions change faster than before, the ability to retrain yourself becomes a mass competence. So, the question “why do we have so few strong students?” matters no less than the familiar question “why do we have so many weak ones?”

What happens in class?

According to PISA-2025, more than eighty percent of teenagers say their teacher is interested in their learning, keeps explaining until the class understands, and is ready to give extra help. All three indicators are above OECD averages. Shortages of teachers, materials, and infrastructure have indeed fallen in recent years. This does not mean that all is well with the quality of teaching. PISA measures perceived support, not the depth of pedagogical work. So let us put the question more precisely again. If the teacher is there, the child comes to school, the material environment improves, and social origin shapes the result relatively weakly, then the next question has to be looked for not only around the lesson but inside it. It is one thing for a student to memorize a rule and solve ten problems of a familiar type. It is another to see the eleventh, worded differently, and work out for himself what to do. It is one thing to retell a paragraph. It is another to take an unfamiliar text and find the main point, spot a contradiction, compare several sources. It is one thing to know a formula, another to understand where and why it should be applied. The point is that the lesson is built around the right answer. The answer can be checked, it takes less time, and it is what goes into the reporting. The ability to handle a task with no ready answer is hard to check, and it does not enter the reporting at all. A school does what is asked of it. One of the weak links in our school may lie between knowledge received and knowledge used independently. If so, what we need to count is not resources, but what they turn into in class.

AI makes this question urgent

PISA-2025 lets us see the same gap from another side. 47% of Kazakhstani fifteen-year-olds use artificial intelligence to help with schoolwork at least once a week. 38% use AI to get a summary of a text they are supposed to read. Only 36% reach the baseline level of reading literacy. PISA does not prove that using AI worsens reading. But the proximity of these figures’ matters. A tool has appeared in school that can perform an intellectual operation before that operation has become a stable skill in the child. A weak student once still had to finish reading the page. Today an AI can retell it. And what matters here is that the child is not being lazy. He does not see the difference between reading a text and finding out what the text is about. For him these are one and the same action with the same result. Building that difference was exactly the school’s job. So, the question of AI in education is far more complicated than “permit or forbid”.

Which intellectual work can be handed to the AI, and which must a child first learn to do himself? The more powerful the technology becomes, the more it matters that a person can set the task, check the answer, notice the error, and ask the next question.

For many years we measured school education, reasonably enough, by the number of schools built, teachers trained, equipment purchased, programs and digital platforms introduced. All of these indicators are matter. But PISA-2025 suggests looking at the other end of the process.

What does the invested resource turn into?

Not how many texts a student has read, but whether he can work with an unfamiliar one on his own. Not how many problems he has solved by a model, but whether he can solve one where the conditions have changed. Not whether the class has covered the program, but whether the child can use what he studied without a ready-made instruction. We have learned to count inputs. Now we need to learn to measure the conversion. In this sense the next stage of reform in school education is not only the school, the teacher, or the textbook. It is the cognitive work of the student.

The Kazakhstani school already solves several crucial tasks. It has widened access considerably. By international data, it ties a child’s educational fate to the family’s material position comparatively weakly. It has a supportive teacher. Most teenagers do not reject school. So, what stands before us is not a story about an institution that fails to work. The opposite. The school works better socially than cognitively. And this may be exactly what makes the next stage of modernization different from the previous one. PISA-2025 puts a question to Kazakhstan far more important than our place in an international ranking.

What kind of school are we building?

Thirteen years ago, one percent reached the top levels. Today it is one percent again. But all this time the conversation has been about them, about the strong, about the Olympiad winners, about the ones who will be picked for a specialized school. The other ninety-nine percent started first grade in 2012. They are around twenty now. They are already working. And those who are fifteen today will enter the labour market in the mid-thirties. By then professions will be changing faster than generations do, and it will not be one percent who have to retrain, but all ninety-nine. The ability to make sense of an unfamiliar problem will stop being a mark of talent. It will become a condition of employment.

A school of access we have built, the child has a place, a teacher, and resources. A school of social integration too, for the most part, a child’s result depends little on family income. What remains is the third, where a child learns to read, to reason, to doubt, and to carry knowledge into a new situation. What kind of school are we building for those ninety-nine?

---

[1] https://www.oecd.org/en/publications/2026/09/pisa-2025-results-volume-i_5265bfb1.html

[2] OECD (2026), PISA 2025 Results (Volume I): Future-Ready Students, PISA, OECD Publishing, Paris, https://doi.org/10.1787/73451bc5-en.

[3] Run by the Organisation for Economic Co-operation and Development (OECD) every three years among fifteen-year-olds.

[4] https://en.wikipedia.org/wiki/Gifted_Education_Programme_(Singapore)

[5] https://www.britannica.com/place/China/Education

[6] https://en.wikipedia.org/wiki/List_of_high_schools_for_the_gifted_in_Vietnam