The same prompt, two classrooms. In one, a private university in Bogotá with an institutional license to a paid model, stable connectivity, and students who arrive with the assistant already open in the browser. In the other, a provincial public university where the bandwidth is not enough to load the free interface and where three students share one phone. “Integrating artificial intelligence into education” names, in each case, two worlds that do not touch.
The reference frameworks we inherited do not see that second world. UNESCO proposes a global framework that presupposes generalized infrastructure, operative regulation, and continuous teacher training — conditions that, across much of Latin America, are not met. The Spanish framework is designed for the Spanish system, with its own rules. Neither is wrong for its own context. The problem appears when they are brought to the South as a universal template.
Here I need a distinction Catherine Walsh sharpened, and it changes the whole conversation. It is one thing to treat context as an adjustment variable: we have a universal model and we adapt it to local conditions. It is another to treat it as a constitutive condition: local conditions determine what kind of model is pertinent. In the first case context modifies implementation. In the second it modifies theory. Walsh called the first functional interculturality — adding diversity to a system that stays intact — and the second critical interculturality — questioning the system that produces the need to “include” the other.
A framework that adapts the European model to Bogotá is functional interculturality. A framework that starts from Bogotá’s conditions to ask what competences a teacher needs facing AI is critical interculturality. The difference is not a nuance. It is a direction.
Aníbal Quijano named the mechanism with a precision that still stings: the coloniality of knowledge. The presumption that valid knowledge is produced in the North and applied in the South. Competence frameworks reproduce that logic without meaning to: they are written in Brussels or Madrid and “incorporated” in Bogotá or Mexico City. The very word we use — “implement” — betrays the position. One implements what another thought.
Feenberg helps show why this matters technically and not only politically. Every technology passes through two moments: separating it from its origin to make it a means — he calls this primary instrumentalization — and reinserting it into a social context loaded with values, meanings, and power relations — secondary instrumentalization. Existing frameworks deal almost only with the first: they teach how to use the tool, use it well, use it responsibly. The second — what it means to reinsert it into a Bogotá classroom with forty-five students, intermittent connectivity, and no national regulation — is left out. And that is not an implementation detail. It is a theoretical omission.
The model I designed reverses the direction. It does not adapt an imported theory to a local reality. It thinks from Latin American experience and offers itself as a contribution to the global debate. The digital divide stops being an external precondition someone will close with investment, and becomes a constitutive condition of the pedagogical problem itself.
When my university’s vice-rector for research — she from engineering, I from linguistics — and I talk about this, we agree on something that, said aloud, sounds obvious and, in practice, is ignored all the time: to think from Latin America is not to apply someone else’s theory to one’s own reality. It is to produce theory from an experience that does not fit the borrowed molds.
The same prompt, two classrooms. Whoever does not see the difference is teaching in a country that does not exist.
