How we think.
AI does not teach. Judgment does.
Education is an act of dialogue — not transmission.
From Socrates to Comenius, from Comenius to Freire, the deepest education has never been depository. It is a conversation between someone who asks and someone who dares to think aloud. The arrival of artificial intelligence does not exempt us from that demand; it radicalizes it.
AI is not a neutral tool.
It is a new interlocutor — opaque, prone to error, trained on biases it did not choose. To pretend it is just "an assistant" hides its dialogical nature. When a teacher asks AI for help, they are not consulting an encyclopedia: they are conversing with an agent that has its own ways of answering.
The quality of learning depends on the quality of the dialogue.
We have shown this empirically in our first paper: when the teacher poses instrumental questions, they receive instrumental answers. When they co-design, the AI co-designs with them. The mirror always returns what is asked of it — the question is what we ask.
The teacher will not be replaced. They will be augmented — or diminished.
The false debate is whether AI will replace teachers. The real debate is whether we will use it to free pedagogical time — or to automate teaching into content management. The decision lies in every interaction, in every prompt.
We publish in the open, in Spanish, from Mexico.
Most research on AI and education is produced in English and designed for Anglophone contexts. But the Mexican classroom, with the New Mexican School and its multicultural reality, poses its own questions. Our contribution is from here, for here, with the perspective of elsewhere.
Empirical, not prophetic.
The field of AI + education is full of manifestos without data and predictions without method. We analyze real conversations, validate frameworks with expert panels, and publish the code. Where others speculate, we count.