Intelligently integrating AI into education

Making AI at school a lever for learning, without compromising critical thinking or educational rigour

Artificial intelligence is now an integral part of primary, secondary and higher education in the Wallonia-Brussels Federation, to the point that it is legitimate to speak of a genuine before and after in our education system. It offers considerable opportunities: significant time savings for teachers in their administrative tasks, lesson preparation, documentation, assessment, and continuous educational support. For pupils and students, AI provides a form of omnipresent tutoring, available 24/7, capable of re-explaining material, suggesting exercises and supporting individual progress. But this power also comes with serious dangers: the temptation of laziness, the risk of cognitive decline if overused, confabulation, and addiction to screens and instant answers. Hence the need for stronger pedagogical, ethical and critical oversight so that these tools support learning without replacing it.

In this context, the Wallonia-Brussels Federation already has digital resources and AI-related tools, but the overall picture remains fragmented and lagging behind other regions, notably Flanders and certain foreign countries. There is a certain scattering of initiatives to be regretted, along with a tendency to multiply reflections and commissions on AI ethics rather than quickly implementing concrete solutions in the classroom. At the same time, the private offering is developing rapidly, increasing the risk of widening inequalities in access and quality if schools do not structure their own framework. In this context, it is becoming urgent for teachers to actively train in the use of AI and to teach young people to use these tools appropriately, critically and responsibly, rather than simply enduring their presence.

Finally, the issue of AI forces us to revisit the respective purposes of primary, secondary and higher education. In primary school, the challenge is to consolidate fundamental learning and preserve effort, attention and reading, while gradually introducing minimal digital literacy. In secondary school, it is more about developing critical thinking, the ability to structure one's own thought and to use AI as an analytical tool, not a substitute. In higher education, the focus is on how to integrate AI into curricula, teaching methods and assessment, so that it strengthens academic rigour rather than diluting it. All this translates into a series of concrete proposals for developing AI in education, within a coherent framework combining innovation, pupil protection and raised educational ambitions.

Challenges, opportunities and risks of AI in education

Artificial intelligence is now establishing itself as a major transformation of education in the Wallonia-Brussels Federation. Whether we like it or not, it is already present in the practices of pupils, students and teachers. There will undeniably be a before and after this revolution. Like writing, printing and the internet before it, AI constitutes a civilisational break that transforms the content, methods and purposes of education.

This transformation is all the more striking because it is happening in an extremely short timeframe. Whereas previous educational revolutions unfolded over decades, or even centuries, AI has spread in just a few years. The episode of academic work massively written using AI acted as a brutal revelation: traditional forms of assessment were suddenly undermined, even rendered obsolete, exposing an education system caught off guard.

Faced with this disruption, the opportunities are considerable. AI offers unprecedented potential to improve pedagogical efficiency. It allows teachers to save valuable time in lesson preparation, creating exercises, marking and monitoring pupils. This time saving can be reinvested in what lies at the heart of the profession: the pedagogical relationship, individualised support and the transmission of meaning.

For learners, AI opens the way to genuinely personalised education. Every pupil can benefit from support tailored to their own pace, weaknesses and strengths, with permanent access to explanations, exercises and resources. Learning is no longer confined to the space and time of the classroom, but becomes a continuous process, accessible at any time.

AI is also a powerful lever for inclusion. It can compensate for certain difficulties, particularly for pupils with disabilities, pupils with dyslexia or those who do not fully master the language of instruction. It facilitates access to knowledge and allows teaching materials to be adapted to a diversity of profiles, thereby helping to reduce certain structural inequalities.

It also offers a major opportunity to rethink teaching methods. Far from being limited to a tool for producing answers, it can become a support for critical learning. Analysing generated responses, comparing sources, and detecting biases and errors are all exercises that strengthen analytical skills and intellectual rigour.

However, these opportunities come with significant risks. The first is the deterioration of cognitive abilities through passive or systematic use. The automation of essential intellectual tasks (writing, structuring thought, memorisation) can lead to a reduction in mental effort and a weakening of fundamental skills. Recent work by the MIT Media Lab on the use of ChatGPT in writing tasks suggests precisely that too early or too heavy reliance on a conversational agent can reduce cognitive engagement, weaken memorisation and diminish the sense of ownership of the work produced.

This risk is all the more worrying because it occurs in a context already weakened in the Wallonia-Brussels Federation. The 2022 PISA results show a decline in mathematics, science and reading, with a drop of 21 points in mathematics since 2018 and 23 points in reading over ten years, while inequalities linked to social background remain among the most pronounced. If used without oversight, AI could amplify these trends by encouraging the search for immediate solutions at the expense of effort, concentration and the appropriation of knowledge.

Another danger lies in the transformation of our relationship with knowledge. When answers are instantly available, the risk is that speed is favoured over understanding, recall over appropriation, and the assembly of content over the personal construction of reasoning. The pupil or student may then become a mere orchestrator of generated content, without genuine integration of knowledge, without lasting consolidation of what has been learnt, and without sufficient development of intellectual autonomy.

Finally, AI raises major ethical, cultural and democratic issues. Algorithmic biases, the standardisation of content, dependence on external models and the dominance of predominantly English-language corpora can influence learning and weigh on neutrality, pluralism and educational sovereignty. Without a clear framework, without validation of the tools, and without vigilance over content, these technologies risk conflicting with the fundamental principles of education in the Wallonia-Brussels Federation.

Recommendations

• Learning to become irreplaceable

The first mission of education in the age of artificial intelligence is to train individuals capable of remaining irreplaceable. Learning can no longer be conceived as a phase limited to the beginning of life, but must be understood as a continuous process, shaping the entire personal, academic and professional journey. In a rapidly changing technological environment, the ability to learn, unlearn and relearn becomes a central skill, perhaps the most decisive of all.

In this context, ceasing to learn directly exposes one to the risk of being left behind. Rapid technological developments are transforming professions, rendering certain skills obsolete and creating new ones at a steady pace. Education must therefore prepare pupils and students for this instability, not by giving them the illusion of definitive knowledge, but by developing their autonomy, intellectual curiosity, appetite for effort and ability to adapt to changing realities.

This implies a transformation of teaching methods and assessment models. Diplomas remain important reference points in social, academic and professional organisation, but skills (their updating, transferability and capacity to be applied in new situations) are becoming increasingly decisive. The aim must be to train individuals capable of structuring original thought, solving complex problems, formulating good questions and collaborating intelligently with AI tools without ever dissolving into them.

• Empowering and supporting teachers

Faced with the lightning-fast emergence of AI, empowering teachers is an immediate priority. It is essential to encourage a culture of self-training, drawing on existing tools: Pix courses dedicated to AI, guides, tutorials, resources for education about and through digital technology, and online micro-learning modules. This skills development cannot wait for the full completion of traditional training schemes, as these tools are already being used daily by pupils and students.

This approach fits within a broader culture of lifelong learning. Teachers must embody the model they pass on: that of a professional capable of evolving, questioning their practices, updating their knowledge and adapting to a rapidly changing environment. This exemplary role is essential. It determines the credibility of the school institution at the very moment when it must demonstrate that it remains capable of guiding young people through a world marked by profound technological disruption.

At the same time, a comprehensive training plan must be implemented from the start of the 2026-2027 school year. This must cover the educational uses of AI, its administrative applications, critical analysis of its biases and limitations, as well as its ethical and legal implications, particularly with regard to the GDPR and the EU AI Act. The aim is to ensure that every teacher masters these tools while understanding their risks, so as to integrate them judiciously into their professional practice and thereby safeguard the quality of the public education service.

• Deploying AI in a controlled way

The integration of AI into education must be part of a clear and structured strategy. It cannot consist of a disorderly accumulation of local initiatives, scattered private solutions or improvised practices. An ethical and secure framework is essential to guarantee compliance with European standards, ensure data protection, clarify responsibilities and preserve the trust of the education community in the tools used.

Deployment must be gradual, based on trials and a controlled scaling-up. This requires making practical toolkits available to teachers, school leaders and education stakeholders, the considered introduction of generative AI tools to support administrative and pedagogical tasks, and a gradual familiarisation of pupils with digital technology and AI. This scaling-up must be aligned with planned curriculum developments, notably the digital annex of the FMTT-N framework and future frameworks based on DigComp.

Scientific evaluation of these initiatives is essential. Pilot experiments, conducted in representative schools, would make it possible to observe the real effects of these tools on learning, school organisation and teachers' working conditions. This evaluation is essential to adjust public policy, identify the most relevant practices, avoid fads and ensure the long-term effectiveness of investments made.

• Integrating AI literacy from primary school onwards

Learning about digital technology and AI must begin at a very young age. It is essential to train citizens capable of understanding the technologies that shape their daily environment rather than passively enduring them. In this respect, digital literacy can no longer be considered a peripheral skill; it becomes a fundamental dimension of general education, on a par with mastery of language, mathematics or science.

At primary level, this involves an introduction to algorithmic logic, visual programming and the development of critical thinking. Pupils must learn to verify information, distinguish facts from opinions, and understand that a digital tool is neither neutral nor infallible. This early familiarisation helps build a clear-eyed relationship with technology and lays the foundations for a thoughtful use of AI.

At secondary level, this learning must be deepened through the acquisition of more explicit technical skills, such as learning simple programming languages, but also through the study of the ethical, democratic and social challenges of AI. This requires increased time devoted to digital education in school timetables, as well as sufficient investment in equipment, content and support. Without this structure, the aim of making pupils autonomous in the digital world will remain unfulfilled.

• Guaranteeing fundamental learning

The integration of AI must not come at the expense of basic skills. Mastery of reading, writing, arithmetic, memory, attention and reasoning remains the indispensable foundation on which any relevant use of technology rests. Without these fundamentals, AI risks becoming an impoverishing substitute rather than a tool for empowerment.

AI-free learning spaces must be maintained to guarantee pupils' cognitive autonomy and allow for an authentic assessment of their achievements. Basic skills must be assessable without recourse to AI, in a sufficiently protected setting to gauge how far a pupil can genuinely read, understand, write, calculate or reason independently. This requirement is all the more important in a context where AI-assisted cheating is becoming technically easy.

Clear rules must therefore govern permitted and prohibited uses, to prevent cheating and safeguard academic integrity. Outside explicitly regulated pedagogical uses, mobile phones and screens must be strictly regulated in educational institutions. More generally, any certifying assessment must uphold the principles of authenticity, responsibility and transparency, without which the act of assessment loses its credibility.

• Developing critical use of AI

It is essential to teach pupils and students to use AI critically and independently. These tools are already used de facto, sometimes intensively, outside any school framework. The challenge is therefore not to ignore their existence or believe in a purely formal ban, but to build the knowledge and reflexes that enable controlled, thoughtful and intellectually demanding use.

This means understanding the existing tools, their specific features, how they work, their limitations and their areas of uncertainty. Formulating relevant queries becomes a core skill, drawing on language, logic, the ability to frame problems and conceptual clarity. Learning to question an AI is not simply learning a new technique: it is also learning to think more precisely about what one is looking for, what one expects and what one can legitimately question.

Pupils must also learn to verify answers, detect biases, errors, hallucinations and excessive simplifications, and to cross-check sources. The aim is to develop a genuine culture of critical thinking in a world saturated with automatically generated content. This capacity for verification and detachment will become all the more important as AI systems become more convincing in form while remaining fallible in substance.

• Creating a public educational AI

The development of a free educational AI is a major lever for ensuring equal opportunities. In a landscape where private solutions are multiplying, often with varying levels of quality and powerful commercial logics, providing a public tool based on the frameworks and validated resources of the Wallonia-Brussels Federation would offer a reliable, common and accessible foundation for all.

This tool could act as an educational tutor, re-explaining content in a personalised way, offering tailored exercises, correcting certain work, questioning pupils and supporting their motivation. Such an AI could also serve as a safety net in exceptional situations: prolonged absences, teacher shortages, strike days, health crises or other disruptions. It would thus help strengthen educational continuity, without claiming to replace school.

It must, however, be clearly established that this public AI should be designed as a complement to the teacher, not a substitute. The human remains at the heart of the educational process. The technological tool can amplify possibilities for support, repetition and personalisation, but it replaces neither the educational relationship, nor pedagogical authority, nor the embodied example that the teacher represents in the intellectual and civic education of young people.

• Creating a high-performing guidance AI

AI can significantly improve guidance systems by providing reliable, up-to-date and personalised information on study pathways, career prospects, employability rates and salary expectations. In a system where many pupils and students struggle to navigate the diversity of courses, training programmes and professions, such a tool can help make information more objective and facilitate better-informed choices.

A high-quality guidance AI could combine educational, economic and sectoral data to provide finer, better-documented support tailored to individual profiles. It would help not only to better understand the range of training available, but also to grasp the concrete consequences of choices made, linking personal aspirations to labour-market realities and future economic needs.

Its integration must, however, be coherent with existing arrangements, whether guidance services, career hubs ("cités des métiers"), Walloon initiatives or French-speaking tools under development. The aim should not be to add another tool on top of others, but to strengthen the coherence of the guidance ecosystem, offering simpler, richer and more continuous access to quality information.

• Enshrining the right to lifelong learning

AI offers the opportunity to rethink education as a permanent right. Learning must no longer be seen as a specific period confined to the early years, but as a possibility open throughout life, compatible with discontinuous careers, career changes, family constraints or regular upskilling needs. The State must progressively guarantee everyone access to training, including through educational AI tools.

This requires guaranteeing a minimum baseline of digital and information literacy for the entire population. Without this baseline, the promise of lifelong learning risks mainly benefiting those who already have the cultural, technical and social resources to seize it. The development of educational AI must therefore be accompanied by an explicit effort on digital inclusion, so that everyone can concretely exercise their right to learn.

Educational platforms must therefore be structured to offer coherent, modular and personalised pathways, building on the digital ecosystem already deployed by the Wallonia-Brussels Federation and the Regions. Such an architecture would allow resources to be pooled, content shared, teachers supported, educational continuity facilitated, and a modern conception of the right to education strengthened, based on personalisation, flexibility and continuity.

• Enshrining cognitive portability

Cognitive portability represents a major shift in how learning is envisaged in the age of digital assistants. It refers to an individual's ability to transfer, from one AI system to another, all their interactions, preferences, habits, knowledge and cognitive history built up over time. In the longer term, this continuity could become an essential element of educational personalisation.

This "digital self" represents a valuable resource for continuity in learning. It gathers together the traces of a journey, the difficulties encountered, progress made, methodological preferences and the most effective forms of support. From an educational perspective, it forms an anchoring point for increasingly individualised learning, provided this personalisation remains in the service of empowerment rather than increased dependence on a single operator.

Its implementation must be accompanied by strict guarantees on data protection, confidentiality, sovereignty and individual control. A first step would be to enshrine, in the architecture of the Wallonia-Brussels Federation's platform, the principle of a single account per pupil throughout their schooling. In the longer term, this principle could also be linked to other public tools for tracking certifications and pathways, to better connect initial training, continuing education and job placement.

Conclusion

The advent of AI in education constitutes a systemic upheaval. In a very short time, it has transformed the conditions of knowledge production, challenged traditional methods of assessment, and shifted the boundaries between what pupils produce themselves and what they can now delegate to a machine. The abruptness of this change explains both the fascination it generates and the concerns it raises. Above all, it requires moving beyond a wait-and-see or defensive stance towards clarity, structure and decisiveness.

Generative AI clearly holds considerable promise. It opens up the possibility of hyper-personalised education, available at all times, capable of adapting explanations, exercises, pace and learning formats to each pupil or student. For the first time, individualised tutoring, once reserved for a socially privileged minority, can be envisaged on a large scale. In a system marked by uneven levels, teacher shortages and unequal support, this prospect cannot be ignored.

But this promise has its downside, and that downside is all the more worrying because it strikes at the very heart of the educational mission. Reading a long text, sustaining attention, memorising, structuring thought, writing, arguing, prioritising information, exercising judgement: all these skills risk being short-circuited if AI becomes a systematic substitute for intellectual effort. The problem is not merely that the tool helps; the problem arises when it excuses one from learning. At that point, technology no longer strengthens autonomy: it erodes it.

The danger is compounded by the fact that AI is not arriving in a neutral or robust educational landscape, but in an already fragile environment. The 2022 PISA results show a decline in pupils' levels in mathematics, science and reading in the Wallonia-Brussels Federation, along with one of the highest levels of social inequality among comparable systems. The school climate is also deteriorating, with more noise, disruption and significant staffing difficulties reported by a large proportion of school leaders. In such a context, AI could become either an accelerator of decline or an instrument of recovery, depending on how it is used.

Another paradox must also be taken seriously: the time and efficiency gains enabled by AI can be cancelled out by the attentional environment in which this technology operates. The very digital ecosystems that offer powerful learning tools are also those that fragment concentration, reinforce screen addiction and accustom young people to instant gratification. The risk is thus to create pupils and students capable of quickly obtaining a formally correct result, but increasingly unwilling to sustain the inner effort needed for understanding, ownership and independent thought.

In this context, two symmetrical mistakes must be avoided. The first would be to demonise AI, to claim to permanently exclude it from educational practice, or to view it merely as an external threat. Such a stance would condemn education to rapidly falling behind the social, cultural and professional realities young people will have to face. The second mistake would be the opposite: believing that AI alone constitutes the solution to the system's structural difficulties, and gradually delegating to it functions that in reality belong to a human, political and institutional educational project.

The only reasonable path is one of mastery. This is not about adding yet another technology to an already saturated system, but about evolving practices in a balanced, gradual and responsible way. This evolution must combine teacher training, support for institutions, rigorous experimentation, scientific evaluation and robust ethical safeguards. Without this, AI will impose itself de facto, but without coherence, without control, and mainly to the benefit of the best-equipped, best-trained or most well-resourced players.

It is with this in mind that the complementary role of three major learning instruments must be reaffirmed. The first is reading. It remains the sanctuary of inwardness, concentration, fruitful slowness and encounter with the singularity of great works and great ideas. In a world dominated by instantaneity, preserving time for deep, screen-free reading becomes a major educational imperative, because that is where nuance, depth and true independence of mind are formed.

The second instrument is digital tutoring. AI can become a pedagogical aide of remarkable power: it explains, repeats, questions, corrects, rephrases, motivates and adapts. It can provide each pupil with a form of individualised assistance, available 24 hours a day, seven days a week, without fatigue or impatience. But this strength must be precisely framed: AI must remain an intellectual servant, not a master. It must support the pupil's progress, not spare them the effort that alone allows skills to be consolidated.

The third mechanism is the school itself, understood as an embodied, collective, political and democratic space. School is not just a place for transmitting content; it is a place of presence, encouragement, authority, confrontation with otherness and learning shared rules. There, one learns to listen, to speak, to meet a requirement, to measure oneself against peers, to admire a teacher, to receive criticism and to take part in a learning community. None of this can be replaced by a chatbot, however powerful it may be.

The challenge for contemporary education is therefore to skilfully orchestrate these three paths (reading, AI, school) without allowing any one of them to overwhelm the other two. If reading alone reigns, we risk neglecting the decisive contributions of personalisation and interactivity. If AI alone dominates, we encourage cognitive dependence and the erosion of inner effort. If the mass school in its most traditional form alone persists, we risk failing to respond to the growing diversity of needs, or to make use of tools that could strengthen support. The balance is demanding, but it is the condition for a fruitful use of AI in education.

In this respect, the development of a well-designed AI mentoring system probably represents one of the most promising avenues. In higher education as in compulsory schooling, such a mechanism could partly offset certain budgetary and organisational constraints, provide individualised explanations, generate exercises, support drilling, assist with revision and strengthen supervision outside the classroom. In a context of teacher shortages, it can help free up human time for what is most irreplaceable: presence, guidance, judgement and the pedagogical relationship.

Yet these systems must be expressly designed to develop the intellectual potential of pupils and students, not to increase their dependence or passivity. A well-conceived educational AI must encourage checking, reformulating, arguing, comparing and deepening understanding. It should not merely provide answers, but encourage more demanding intellectual work. Only under this condition can it strengthen freedom rather than diminish it.

Fundamentally, the question posed by artificial intelligence in education is not primarily technological; it is political, cultural and anthropological. It concerns a certain idea of the human being, of learning, of authority, of democracy and of emancipation. Do we want to train individuals capable of intelligently using powerful tools, or individuals who depend on these tools to think, write and decide in their place? The answer to this question will determine the very nature of the school of coming decades.

This is why the urgency is real. It is not a matter of giving in to panic, nor of surrendering to technophile enthusiasm, but of acting with method, rigour and determination. AI is already here. It is already transforming behaviours, expectations and relationships with knowledge. The real political choice is therefore no longer whether it will enter education, but under what conditions, in service of what aims, with what guarantees and for whose benefit.

If this framework rises to the challenge, artificial intelligence can become a powerful lever for emancipation, equal opportunity, support for teachers and a revival of educational ambition. If, on the contrary, it is left to improvisation, to the market alone, or to fascination with immediate performance, it risks worsening existing vulnerabilities. The full responsibility of public authorities now lies in ensuring that this revolution serves human intelligence, personal freedom and the quality of education, from school to university.