Educational technology has for years been presented as a driver of equity, based on a rationale that is to dispute: when there is a shortage of teachers, infrastructure or resources, digital platforms make it possible to reach further and reach more students.
Artificial intelligence takes that capacity to a whole new level. Systems capable of adapting content, providing immediate feedback or supporting learning processes without constant teacher intervention open up possibilities that, just a decade ago, were unthinkable. Recent reports from the Inter-American Development Bank and other multilateral organisations highlight precisely this potential: expanding coverage, personalising learning pathways and sustaining learning in contexts where the education system struggles to reach students.
However, this very logic contains within it a much less obvious risk to which we must pay close attention. For when a technological solution becomes the most viable option in certain contexts, it may cease to be a complement and begin, in practice, to shape the predominant educational model.
That being the case, we must ask ourselves: could technology be defining the kind of education received by those who have the least? Is technology reshaping the definition of educational inequality? These are the questions we will reflect on in this article.
Greater access, but not the same education
Traditional equity metrics remain necessary, but they are increasingly less meaningful on their own. Knowing how many pupils are enrolled in school, how many have access to devices, or how many hours they spend online allows us to gauge the scale of the system, but it does not accurately describe the kind of learning they are experiencing.
International assessments such as PISA, led by the OECD, have for years shown that students with similar levels of access and school enrolment can exhibit significant gaps in reading comprehension, mathematical reasoning or complex problem-solving. These differences cannot be explained solely by the resources available, but rather by the nature of the learning experiences to which they are exposed.
The incorporation of artificial intelligence introduces a new variable into this equation. On the one hand, it enables a level of practice and feedback that is difficult to achieve in traditional models. In contexts with overcrowded classrooms or limited teaching time, this can translate into concrete improvements in certain areas, particularly procedural skills.
At the same time, however, it tends to favour forms of learning that can be organised in a clear, sequential and verifiable manner. This is a direct consequence of how it works: it manages best those elements that can be broken down, assessed and optimised.
This does not mean that other forms of learning disappear. But they may lose their central role if they are not deliberately safeguarded in the pedagogical design. Access to technology, in itself, says little about the quality of learning. Two students may use similar tools and develop very different skills. The difference lies not in access, but in how that technology organises the student’s time, attention and tasks – in what takes up the bulk of their learning experience.
This is where the analysis shifts from being quantitative to qualitative. It is not enough to measure how many have access. What should matter to us is what kind of cognitive processes they are being systematically exposed to.
What is included… and what is left out
Artificial intelligence-based systems tend to organise learning around structured tasks: exercises, sequences, verifiable answers, optimised pathways. This helps to improve certain aspects of learning. But it pushes others into the background – aspects that are more difficult to systematise and which are fundamental to an education that must prepare students for uncertain contexts: thinking with others, arguing a case, exploring without an immediate answer, making mistakes and reformulating or constructing complex ideas.
Here it is worth clarifying what ‘pushing into the background’ means. It is not that these skills disappear, but rather that they require conditions which are not always present in highly technology-mediated environments: unfragmented time, sustained interaction, ambiguity, and the possibility of straying from the planned path.
Educational research has long highlighted this tension. Studies compiled by the OECD on high-quality teaching or global competences emphasise that deep learning does not occur solely through repeated practice or immediate feedback, but rather through more open-ended processes, where the student must interpret, make connections and construct meaning.
Even in the workplace, the evidence points in the same direction. Internal studies by companies such as Google have shown that the skills that distinguish high-performing employees are, for the most part, not technical, but rather linked to communication, collaboration or the ability to integrate complex information.
The problem is not that artificial intelligence cannot contribute to these processes. It can, and increasingly so. The problem is that, when it forms the core of the learning experience, it tends to favour what can be managed most clearly: well-defined tasks, explicit objectives and assessable outcomes. And that introduces dangerous imbalances.
If a significant proportion of learning time is devoted to solving structured tasks, whilst other dimensions are relegated to residual or unplanned moments, the balance of the educational process shifts. And with it, the type of skills that are developed over the long term, resulting in learning experiences that may differ depending on the context.
When we talk about equity, we must ask ourselves whether all students, regardless of their background, have access not only to content, but also to learning processes that involve interpretation, dialogue, creativity, reasoning and deep understanding.
The risk: normalising different experiences
This is where the analysis becomes more complex. Not because there is an explicit decision to offer a different education to one group compared to another, but because the solutions adopted in contexts of scarcity may tend to rely more heavily on technology-mediated models. It is a pragmatic response to very specific constraints: teaching time, high pupil-teacher ratios, and a lack of specialisation.
And over time, this pattern may cement a very significant difference:
- For some students, learning is built through constant interaction with teachers and fellow pupils, with space for conversation, reflection and exploration.
- For others, learning is organised primarily through interfaces, exercises and automated systems.
This is not a formal division. It is a difference in the day-to-day experience of learning. And that difference matters. It matters because opportunities to develop certain skills are not distributed in the abstract, but through repeated practice. What a student devotes time, attention and effort to on a sustained basis ultimately defines the kind of thinking they develop.
If most of their experience is geared towards solving closed-ended tasks, they will have more opportunities to hone accuracy, speed or procedural mastery. If, on the other hand, their learning environment systematically incorporates discussion, interpretation and collaborative construction, they will develop other skills that require a different kind of exposure.
It is not a question of prioritising one over the other, but of recognising that they are not equivalent. Research into educational inequality has long highlighted that the most persistent differences do not always lie in access to content, but in the type of cognitive demands faced by pupils. Studies compiled by the OECD show that pupils from more advantaged backgrounds tend to be more exposed to open-ended tasks, which require them to interpret, justify or connect ideas, whilst other contexts prioritise more structured exercises.
The incorporation of artificial intelligence may amplify this trend if it is not designed with pedagogical intent. And this is where the most difficult-to-detect risk lies, because rather than an overt gap, it is a progressive divergence in the quality and nature of learning experiences.
Equity as a shared experience
This shift compels us to re-examine the very concept of educational equity. For decades, the aim has been to expand access: to ensure that all pupils are in school, have access to materials, and can participate in the system. That remains a necessary condition. But it is beginning to be insufficient. Because equity is not limited to access; rather, it is determined by the experience. By what actually happens when a pupil learns: what kind of tasks they undertake, what kind of interaction they engage in, and what kind of thinking is required of them.
Talking about equity therefore involves asking whether these experiences are comparable. Whether all students, regardless of their background, have access not only to content, but also to learning processes that include interpretation, dialogue, creation, argumentation and deep understanding.
When this is no longer the case, inequality does not disappear. It transforms. It shifts from access to the quality and nature of learning. And it becomes harder to detect, because it is no longer expressed in visible indicators, but in cumulative differences in opportunities to develop certain skills.
In this scenario, technology can work both ways. It can help to democratise rich learning experiences if designed to reinforce them. Or it can entrench more limited models if used as the primary solution where the system struggles most to sustain other forms of teaching.
Artificial intelligence opens up a real opportunity to expand the possibilities of education. It allows us to go further, to better sustain learning and to offer support that was previously not viable.
But it also forces us to make more precise decisions. Not about whether or not to use it, but about what must remain common to any educational experience. What learning outcomes, what practices and what forms of interaction must be guaranteed for all students. If that definition is not made explicit, the system will tend to resolve its tensions in an unequal manner. And then technology will not act as a leveller, but as a silent organiser of differences.
Not because it generates inequality in itself, but because it can contribute to establishing, without declaring or intending to do so, different ways of learning depending on the context. That is where educational equity is redefined, in practice.
