Discussing artificial intelligence and education is beginning to seem, simultaneously, inevitable and exhausting. Inevitable because AI is already part (implicitly and explicitly) of our daily practices of communication, information search, creation, production, evaluation, work, learning, and leisure. Exhausting because, in a very short time, reports, conferences, guides, decalogues, headlines, promises, and fears have multiplied. It might seem that everything has already been said, but perhaps the most important thing is still pending, i.e., to conceive AI from a truly educational perspective.
For this, it is advisable to start with an idea that is often lost in the noise: we are not dealing with a technology that has arrived from outside to be installed in an intact educational world. This is what Carlos Magro stresses repeatedly: we cannot take the integration of AI for granted; it must be questioned. But it is true that we already live in a post-digital condition. Digital is not a separate layer of life, nor is it a dimension that is activated when we turn on a device and deactivated when we turn it off. Technology has become hybridized with our ways of living. It has left a mark on the way we attend, read, write, create, remember, work, relate, inform ourselves, and learn. Therefore, the question is not simply how to introduce AI into education, but how to understand education in a world already deeply penetrated by digital technologies and AI.
This post-digital view helps to escape from a very common simplification: “AI is just a tool; everything depends on how we use it.” The sentence is comfortable, reassuring, and partly true. People decide, interpret, give it meaning, and assume responsibility. But it is also an insufficient statement. AI is not a hammer, a whiteboard, or a more sophisticated word processor. It is not a neutral tool suspended in the cloud. It is a socio-material phenomenon comprised of infrastructures, data centers, energy, minerals, supply chains, visible and invisible human labor, political decisions, business models, geopolitical interests, languages, biases, norms, imaginaries, and concrete ways of reorganizing life and social structures.
Saying that “it depends on how we use it” can remind us that there is no technological determinism, but it can also be dangerous if it prevents us from seeing that uses never occur in a vacuum. We use technologies designed by someone, under specific conditions and perspectives, with particular purposes, and within specific economic and cultural ecosystems. AI has real effects in the physical world: it consumes resources, requires infrastructure, concentrates power, reorganizes jobs, conditions decisions, produces dependencies, and affects fundamental rights. It also has technical, social, ethical, environmental, epistemological, ideological, political, mercantile, and, of course, educational implications. It is not enough, therefore, to teach how to use it well. We need to ask ourselves what world it sustains, what world it produces, and what world we want to build with it or in opposition to it.
From this complexity, the debate on AI and education becomes more interesting and… uncomfortable because it forces us to move beyond enthusiasm and rejection. Above all, it forces us to review an important part of what is being written and said on the subject. There is a lot of talk about ethics, responsibility, transparency, privacy, and bias. All of this is necessary. However, I have the feeling that we continue to consider AI from an excessively solutionist, mercantilist, and hyper-productivist perspective, often disguised as ethical concern. How many times have you seen infographics made with oversaturated, hyper-stimulating AI about the ethical implications of this technology? And how many times have we discussed the tools available to improve our productivity?
Solutionism occurs when AI is presented as an answer before the questions have been properly formulated. Personalization of learning (often confused with individualization), automation of teaching tasks, generation of materials, immediate correction (a high-risk practice that should not be done), permanent tutoring, predictive analysis of performance, and process optimization are promised. Some of these possibilities may be useful in certain contexts; I would be naïve if I denied it. But education is not a collection of technical problems waiting for a more powerful tool. Educating implies relationship, context, conflict, friction, care, time, interpretation, professional judgment, and shared construction of meaning. None of that is solved by simply adding an algorithmic layer.
I believe that there are two logics behind solutionism. On the one hand, market logic appears when educational questions are subordinated to market questions: which platform do we lease, which license do we buy, which tool saves the most time, which provider offers the best functionalities, which solution promises not to leave us behind…? Educational AI (if one can say that) is not developing in a neutral space, but in a technopolitical, capitalist, and geopolitical race to dominate infrastructures, data, standards, narratives, and markets. Therefore, when a university, an administration, or an educational center decides to integrate AI, it is not just making a technical decision. It also involves decisions on digital sovereignty, technological dependence, data protection, institutional autonomy, and an innovation orientation.
On the other hand, hyperproductivist logic appears when AI is justified, above all, by its ability to accelerate: generating more materials in less time, correcting faster (I insist, it should not be done), summarizing more documents, producing more reports, generating more activities, answering more emails… automating more and more processes. The problem is not wanting to alleviate real workloads, which exist and are many. The problem is that we end up confusing educational improvement with increased productivity. We have learned to wrap acceleration in ethical vocabulary: responsible use, sustainable efficiency, inclusive innovation… but if we do not discuss the ends, ethics risks becoming a superficial layer that allows us to continue doing the same thing, faster and with greater technological dependence.
Perhaps that is why education views AI with a mix of interest, anxiety, suspicion, and even fear. This is understandable, especially in higher education studies. Education faculties have a great responsibility: we train future teachers and, therefore, we participate indirectly in the training of all the professional generations to come. We are not just another actor within the system. We are a strategic node and must think about how we learn, how we teach, how we evaluate, and how we build critical citizenship in a society permeated by AI. However, integrating AI into initial and continuing teacher training is not easy. It changes the rules of the game because it forces us to leave our comfort zones, even when we say we innovate. It forces us to review practices that seemed stable: what it means to write, solve a task, evaluate learning, accompany a process, author a work; to consider what critical thinking means and what place the teaching criterion occupies when automated systems can mediate part of the textual, visual, or analytical production.
For years, we have talked about educational innovation, digital competence, active methodologies, continuous formative and summative assessment, and educational transformation. However, AI raises a more radical question: Were we really transforming education, or simply modernizing some of its forms? Because if a student can generate an essay in seconds, I’m sure the problem isn’t just how to detect if she’s done so. Perhaps the questions concern what kind of essay we should assign, what process we want to provoke, what evidence of learning we consider valuable, and how we can design educational experiences where thinking matters more than delivering.
The same happens with university teaching. If AI can generate a rubric, an activity, an explanation, a presentation, or an evaluation proposal, the value of the teaching staff does not disappear; it is simply displaced and transformed. We have been saying for years that it is not enough to produce materials or disseminate content: the AI phenomenon and its integration into education have brought out our colors and challenge us to make the teaching task more complex. That complexity entails none other than designing contexts, asking good questions, accompanying processes, sustaining conversations, deciding on criteria, reading a situation correctly, deciding when a technology adds value, and when it impoverishes the experience. AI does not eliminate pedagogy. On the contrary, when it is integrated without pedagogy, its absence becomes clearer.
Meanwhile, we can observe that events occur at three levels. At the macro level, international institutions are starting to mobilize. PISA will integrate media and artificial intelligence literacy by 2029. The OECD and the European Commission have promoted the AILit framework on AI literacy aimed at primary and secondary education. European regulation, for its part, already requires AI literacy as a relevant obligation for those who provide and deploy AI systems. These actions indicate that AI literacy is no longer a specialized interest and is beginning to be recognized as a vital dimension for critical citizenship.
But just because something fits into an international framework does not mean that it automatically fits well in the classroom. Policies can guide, create a common language, and legitimize priorities, but they do not replace situated pedagogical work. A competency framework alone does not transform teacher training (we have already experienced this with the DigComp framework for digital competency for years); I am afraid that an international test will not solve the meaning of the curriculum either. For its part, regulation does not guarantee a critical culture. The macro level is necessary, but it needs to engage with what happens in institutions, schools, classrooms, and in the daily decisions of teachers and students.
At the meso level, many universities, networks, and educational centers are beginning to consider AI governance strategies. This word, “governance,” is key because the integration of AI cannot depend solely on each teacher’s curiosity, prudence, enthusiasm, or fear. It needs institutional conditions, shared criteria, infrastructures, training, spaces for reflection and deliberation, and technical, legal, ethical, and pedagogical support. It also needs research and transfer, as well as clear policies on data, privacy, evaluation, authorship, and acceptable uses. That doesn’t just mean writing a user guide or deciding which tools are allowed. It means asking ourselves what educational-pedagogical, technological, and ethical ecosystem we want to build. It also involves discussing whether we are committed to our own infrastructure or commercial platforms, which dependencies we accept, which data we expose, which internal capabilities we develop, and how we prevent digital transformation from being reduced to buying licenses. A university does not become more innovative by leasing more tools. They become more responsible when they can align their technological decisions with their educational and social mission.
The meso level must also address the redesign of teaching and learning processes. It is not enough to add a statement on AI in the teaching guides. Tasks, methodologies, evaluation criteria, and forms of accompaniment must be reviewed. It is also necessary to decide when it makes sense to use AI as an object of study, given its didactic relevance, and when to use it as a tool. A distinction must be made between learning about AI, learning with AI, and learning in an AI world. Teachers must be trained not only to use applications but also to make informed, critical, and contextualized pedagogical decisions.
Finally, there is the micro level: the personal, the professional, the day-to-day. There, as in so many other social debates, we tend to polarize: either we embrace AI as an inevitable promise, or we detest it as an absolute threat. At one extreme, uncritical enthusiasm sees AI as an opportunity to innovate, save time, customize, produce, and avoid being left behind; at the other, absolute rejection interprets it as a threat to authorship, thought, evaluation, privacy, teaching work, or social justice. Both extremes have partially valid reasons, but neither position is sufficient. We must escape that looping polarization. Neither fascination nor rejection: we need critical judgment.
And critical judgment is cultivated. It does not appear spontaneously just by having access to a tool or prohibiting it. It requires knowledge, skills, and attitudes. Knowledge to understand, at least basically, what AI is, how these systems work, what they do and what they don’t do, where their results come from and what limits they have (AI literacy). It requires skills to use them when they add value, to assess AI responses, protect data, contrast information, ask good questions, and design meaningful learning processes (digital competence). Attitudes to act with prudence, curiosity, responsibility, justice, critical awareness, and sensitivity to social and environmental impacts (critical and ethical perspectives). The key competence will not mean using AI all the time. It will be knowing when not to use it. And, when it is used, know why, for what, under what conditions, and with what consequences. This idea is tremendously important in education because the goal should not be to train obedient users of intelligent systems, but rather to develop people capable of making decisions about those systems.
That’s why all three levels matter: the macro, the meso, and the micro. They must be kept in alignment. We need to build coherence between regulation, institution, teaching practice, and educational sense, with leadership. This implies remembering that the purpose is not to adapt education to AI, but to educate so that AI does not develop on the margins of democracy, social justice, and ordinary life. The underlying question is not what AI can do for education, but what education should do in the face of a society that AI is already reorganizing.
From the universities, this question challenges us fully. It is up to us to respond to our role as catalysts for social change. This implies research, yes, but also transferring knowledge, accompanying institutions, engaging in dialogue with educational centers, training professionals, reviewing our own teaching, and disseminating clearly. The university cannot limit itself to observing the phenomenon from the outside, nor to reproducing the story that AI is about to revolutionize everything magically. We have a responsibility to make public debate more complex, dismantle simplifications, and provide frameworks for understanding and moving forward. We also have the responsibility of training generations that will work surrounded by AI. Some will exercise professions that we can hardly imagine today. Others will perform well-known but profoundly transformed jobs. In any case, they will need skills to make decisions in contexts mediated by algorithmic systems. Not only technical skills, but also digital, critical, ethical, communicative, political, and pedagogical skills. AI will not only affect those who program AI: it will affect teachers, journalists, health workers, lawyers, engineers, social workers, artists, public administrations, companies, families, and citizens. This is where education has an irreplaceable task: to prevent society from believing the story of automatic revolution. AI does not come to save education, nor to inevitably destroy it. It is a complex, post-digital, and socio-material phenomenon that we must understand, challenge, and guide to build.
If something is transformed, it will be people who transform it. Perhaps that is why the great challenge is not to integrate artificial intelligence into education. The challenge is to build an education capable of training people who not only live in a society with AI but also participate in the construction of a society – and an AI – that is more just, equitable, sustainable, pedagogical, and humane. It is clear to me: AI will not magically revolutionize education. That promise belongs more to technological marketing than to pedagogical thinking. The revolution, if it deserves the name, will depend on our collective decisions: what we teach, what we evaluate, what we research, what we regulate, what we buy, what we reject, what we imagine, and what we take care of. In short, it will depend on whether we can prioritize human, social, and pedagogical intelligence over artificial intelligence.
Acerca de la autora
Amaia Arroyo-Sagasta is a professor and researcher at the Faculty of Humanities and Education Sciences (HUHEZI) at Mondragon Unibertsitatea and coordinator of the Digital Innovation specialization within the Primary Education degree program. She also holds a PhD in Communication and Education from the National University of Distance Education (UNED), where she received the Extraordinary Doctorate Award. She holds a master’s degree in Communication and Education in the Digital Network from UNED and a master’s degree in ICT in Education from Mondragon Unibertsitatea. For nearly 20 years, she has trained teachers in the pedagogical use of digital technology.
References
Arroyo-Sagasta, A. & Magro, C. (2026). When evaluating is learning. Authentic assessment for student-centred education. In J.M. Muñoz, N. Lorenzo, X. Suñé, M.A. Prats & C. López (coords.) ODITE 2026: Keys to a new education in the age of AI. pp. 42-29. https://ciberespiral.org/es/download/23047/?tmstv=1781710592
Bilbao-Eraña, A. & Arroyo-Sagasta, A. (2025). Fostering AI literacy in pre-service teachers: impact of a training intervention on awareness, attitude, and trust in AI. Frontiers in Education, 10, 1668078. https://doi.org/10.3389/feduc.2025.1668078
Castañeda, L., Postigo-Fuentes, AY., & Arroyo-Sagasta, A. (2025). Beyond Tools, Toward Power Structures: A Critical Review of AI in Primary Education. Spanish Journal of Comparative Education, 48, 73-95. https://doi.org/10.5944/reec.48.2025.45126
European Commission. (2025, November 19). AI Literacy: Questions and Answers. Shaping Europe’s digital future. https://digital-strategy.ec.europa.eu/es/faqs/ai-literacy-questions-answers
Cosgrove, J., & Cachia, R. (2025). DigComp 3.0: European Digital Competence Framework (5th ed.). Publications Office of the European Union. https://data.europa.eu/doi/10.2760/0001149
Organization for Economic Co-operation and Development. (n.d.). PISA 2029 media and artificial intelligence literacy. OECD. OECD. https://www.oecd.org/en/about/projects/pisa-2029-media-and-artificial-intelligence-literacy.html
Organization for Economic Co-operation and Development, & European Commission. (2026). Empowering learners for the age of AI: An AI literacy framework for primary and secondary education. AILit Framework. https://ailiteracyframework.org/
Translation by Daniel Wetta
