The Crisis of School Reality: Why AI May Amplify School Problems Rather Than Solve Them

So far we have spoken about the future: global trends, scenarios, and how the roles of the teacher and assessment are changing. But every discussion of a bright digital future for schools has an inconvenient reverse side, and it would be dishonest to ignore it. Before bringing artificial intelligence into the classroom, we need to look honestly at the classroom itself — at school as it exists today, with all its unresolved problems.

This is where a trap lies, one that is easy to fall into on a wave of enthusiasm. We are accustomed to thinking of AI as a tool that solves problems. The logic seems impeccable: school is stalling — give it a smart technology, and it will move. But international experience and our own research suggest something else, much less pleasant. Under certain conditions, AI does not cure the illnesses of school; it amplifies them. To understand why, we must first make an honest diagnosis.

Let us begin with what usually remains outside enthusiastic presentations. In reality, decisions that change school life every day are increasingly made not by schools themselves, and not even at the level of educational strategy. They are made by providers of digital solutions and regional education authorities — quickly, without prior reflection, without pilots, and without evaluation of consequences. At the same time, such decisions immediately change the practice of millions of participants.

A paradox emerges. On one side are attractive futurological concepts of educational transformation. On the other is everyday reality, where the logic of change is set not by pedagogical meaning, but by an external managerial or commercial impulse. In this picture, school becomes not the subject of change but its object. It is forced to adapt to what has been imposed from above or sold from the side, without the ability to manage its own transformation meaningfully.

This is the first worrying symptom. An institution that does not manage its own changes will not be able to manage the arrival of AI meaningfully either. It will simply accept it the way it accepts everything else: as another external given to which it must adjust.

Now to the root problem, without which the whole conversation about AI hangs in the air. This problem is functional illiteracy.

This does not mean that children cannot read or count. It means something else: a weak ability to navigate complex, unfamiliar situations, apply knowledge in practice, reason, and make independent decisions. A student may know a rule but be unable to use it when the conditions of the task do not match the template from the textbook. They can reproduce an answer but become lost when they need to understand rather than remember.

This should be emphasized: it is not the misfortune of a few “weak” schools or negligent teachers. It is a systemic effect, and its causes are structural. Modern school is still largely solving tasks formed in a previous era. Content is overloaded with formal procedures. The entire logic of learning and assessment is built around reproducing the correct answer, not around understanding and reasoning. The system is tuned to reward what a child has memorized and repeated accurately — and it develops this ability best of all.

Remember this diagnosis. A few paragraphs later, it will become clear why it is critical precisely in the discussion of AI.

Participants in our foresight described the same thing in their own words, without academic terminology. School is boring; it has few projects, games, and experiments — little that makes children think and try. Dependence on ready-made answers is growing, while critical thinking is weakening. Teachers are burning out, and there is a lack of what are called soft skills. The gap between city and village is not narrowing but widening. These are not abstract expert anxieties; they are what people inside the system see every day.

The market says more clearly than any survey that trust in school has weakened. Alongside the formal education system, an entire parallel world has grown: tutoring, exam preparation, online schools, foreign platforms, and informal educational services. This world lives by its own rules, outside the institutional framework of school, and is barely taken into account by education governance.

The very fact of its rapid growth is a diagnosis. Families vote with their wallets: if they massively pay extra for education on the side, then what school provides is either insufficient or not trusted. The exam preparation market is especially revealing. From an auxiliary mechanism it has become an independent factor of inequality: those who can afford a good tutor gain an advantage unrelated to the child’s abilities.

There is another bitter irony here. Training to fit the format of an exam imitates an educational result — the score rises — but does not form stable knowledge, let alone thinking skills. The child learns to pass, not to understand. The main burden, both financial and organizational, is shifted to families. School, in this arrangement, quietly loses its status as the basic institution of education, remaining a place one attends “for the record,” while real learning happens somewhere else.

This parallel world is also geographically uneven. An expensive tutor and a good online school are available in a large city and almost unreachable in a village. Formal inequality among schools is layered onto inequality of opportunities outside them, and the gap described by participants in our study becomes entrenched. Any technology introduced into this system risks strengthening precisely this stratification rather than smoothing it out.

Before anyone decides that “bad teachers” are to blame, it must be said clearly and firmly: this is not about them. Weak school results are systemic, and they have three structural causes.

The first is an outdated and overloaded assessment infrastructure. Existing assessment mechanisms are designed to reproduce knowledge and formally meet requirements, rather than to develop thinking and the ability to apply knowledge in practice. We measure the wrong thing — and therefore get the wrong result.

The second is methodological deficits. Pedagogical approaches often do not correspond to the goals of forming functional literacy. Add to this the distortions that arise when international educational instruments are transferred and translated without considering local context — and the result is a methodology that stalls in ordinary conditions.

The third is long-term underinvestment in teacher training and professional development, multiplied by the effect of negative selection. Here the key conclusion of our study is especially important: teachers are not the source of the problem; they are themselves the result of the existing system of management and preparation. Blaming a teacher for the fact that the system has not invested in their development for years is like blaming a plant for not being watered. Responsibility lies with the institutional architecture, not with individuals inside it.

There is one more symptom, perhaps the most treacherous because it disguises itself as success. It is the formal character of progress.

On paper, everything is moving in the right direction: educational programs and regulations are oriented toward developing functional literacy, use all the right words, and set all the right goals. But this is weakly reflected in real results. Formal progress does not turn into sustained improvement and does not reduce structural deficits.

Worse still, monitoring and assessment mechanisms are increasingly used not for honest diagnosis and policy correction, but as instruments of self-presentation — to demonstrate reporting rather than to see the truth. Without honest feedback, reforms reproduce the appearance of change without touching its foundations. The system reports success, problems remain in place, and the cycle repeats.

Now let us put everything together, and it becomes clear why AI in today’s school is dangerous not in itself, but in combination with its unresolved problems.

This is not a theoretical fear. International experience shows directly that in systems oriented mainly toward memorization and reproduction, rather than the development of thinking, AI contributes to the outsourcing of cognitive functions and to the decline of students’ independence. This is an observed pattern, not an invented one — and Kazakhstan, with the diagnosis described above, falls exactly into the risk zone.

Recall the diagnosis: the system is tuned to memorization and reproduction of ready-made answers, not to the development of thinking. Now imagine that powerful generative AI enters such a system — an ideal machine for producing ready-made answers. What happens? The technology does not come into conflict with the school. On the contrary, it fits perfectly into its logic and pushes that logic to the limit. Why think if one can generate? Why understand if the answer appears instantly in finished form?

This is what outsourcing thinking means: the student begins to shift cognitive work onto an algorithm. This happens most readily precisely where school has not taught children to think, but has trained them to reproduce. Instead of strengthening the developmental function of school, AI in such an environment undermines it. The technology begins to work as a compensator for systemic deficits — plugging holes and creating the illusion that everything is fine — rather than as a tool for development.

The trust between student and teacher, one of the few things still holding school together, also suffers. If a child receives answers from an algorithm and the teacher cannot tell where the student’s work ends and the machine’s work begins, the very fabric of pedagogical relations breaks down. Assessment stops meaning anything. Effort is devalued. A school already losing authority loses it even faster.

This is why an honest diagnosis was so important. AI is not a neutral tool that is equally good for any school. It is an amplifier. In a healthy system oriented toward thinking, it will amplify thinking. In a system oriented toward reproduction, it will amplify reproduction and dependence on ready-made answers. Technology multiplies what already exists.

From all this follows a conclusion that changes the whole frame of the discussion about AI in education.

The crisis of school reality is not a background that can be ignored while rushing toward a beautiful digital future. It is the context that determines what AI will become for school: medicine or poison. Introducing AI on top of unresolved basic problems risks turning technology into something that does not cure the illness, but preserves and aggravates it, translating today’s managerial deformations into long-term social effects.

This is not an argument against AI. It is an argument for sequence. First comes an honest discussion of what is broken in school: functional illiteracy, lost trust, assessment that measures the wrong thing, and teachers whom the system has failed to support for years. Only then — or, more precisely, simultaneously with this — comes AI, built in so that it heals rather than finishes off.

That is why the conversation about artificial intelligence in school is in fact a conversation about something else: trust, equality, and the quality of educational policy itself. AI has merely held up a mirror to what was already there. If in that mirror we see a system that does not manage its own changes and measures the wrong things, then the starting point is not the purchase of technologies.

But even an honest diagnosis is not enough. Behind all these problems stands another, the hardest institutional knot — the question of data and digital platforms on which everything rests. Who owns the educational history of a child? Who is responsible for an algorithm’s decisions? Where is the boundary between the state and private platforms? This is the subject of the next material in the series.

The material was prepared based on the results of foresight sessions held in autumn 2025 and winter 2026.