Duties in the Age of AI

The assignment arrived flawless. Clear structure, smooth transitions, correct bibliography. The history teacher read it twice and found nothing to correct, which was precisely the problem: that tenth-grade student had never written like that before. She could have failed him without proof, she could have passed him knowing the truth, or she could do what she did: call him to her desk and ask him to explain, in his own words and without a paper in front of him, what he had learned. The conversation lasted four minutes and was more revealing than the entire assignment. Without knowing it, that teacher had resolved with a single gesture what half the educational system is still debating in the abstract.

The data indicates that this scene is no longer anecdotal. According to the Pew Research Center, the percentage of American teenagers aged 13 to 17 who admit to using ChatGPT for schoolwork doubled in just one year: from 13% in 2023 to 26% in 2024. In Spain, there is no equivalent series yet, but the climate is measured in other surveys: Ipsos' Education Monitor 2024 found that 63% of Spaniards believe ChatGPT should be banned in schools, one of the highest percentages among the countries studied. Between the growing use by students and the majority desire for a ban lies a chasm, and school faculties are currently living in that chasm.

The Misformulated Question

To ban or to permit is a comfortable dilemma because it seems to require only a circular. It is also a sterile dilemma. A total ban drives usage underground and relinquishes the opportunity to educate about it, in addition to running into a stubborn fact: the tool will be present in the adult lives of these students, and a school that ignores it entirely delivers them to it without critical criteria. Total enthusiasm makes the symmetrical mistake, and in worse company: as the UNESCO 2023 Global Education Monitoring Report warns, much of the advice educational systems receive on technology comes from those who sell it, and special caution is warranted with sellers of euphoria. Neither technopanic nor techno-evangelism: the fertile question is not whether AI enters the classroom, but what we ask of the student once it has entered.

Because the traditional homework assignment—summarize this topic, write this commentary, answer these questions—was designed for a world where producing text required having thought. That equivalence has been broken. A language model produces text without anyone having thought, and punishing the student for discovering this is to mistake the messenger for the problem. What has become obsolete is not the effort: it is a specific format of accrediting it.

Redesigning the Task, Not Chasing the Student

Students themselves, when asked seriously, point the way. A study published in 2024 in the journal Tecnología, Ciencia y Educación with Spanish university students found that 92.5% consider the tool useful, while at the same time lucidly pointing out its Achilles' heel: the difficulty of verifying the accuracy of what it returns. Therein lies a magnificent new task: giving a student an AI-generated response and asking them to audit it, find the error, and cross-reference the sources. This is a task that AI cannot do for them, because the task itself consists of supervising it.

The general logic is this: shifting the weight of assessment toward what only happens in front of a teacher. The conversation where what is written must be defended. Writing in the classroom, by hand or without a safety net, not as a punishment but as a gymnasium. Work that requires personal experience—interview your grandfather, measure the courtyard, observe your street—and that no model can invent without it being obvious. Oral presentations where questions arrive unannounced. None of this is new; it is as old as Socrates. AI has not invented a method: it has restored value to the timeless ones.

Administrations, meanwhile, are attempting a balancing act. Autonomous communities have begun to publish guides on the use of artificial intelligence in schools, with different emphases but a core agreement: neither an open bar nor a veto, but supervised use and pedagogical criteria. The fine regulation, the one that decides what is done on Tuesday in Spanish class, will not be written by any official bulletin. It will be written by each school faculty.

Criteria Cannot Be Downloaded

And here the issue shifts to another level, because redesigning homework requires something prior: teachers with their own criteria about what is worth learning and how it is demonstrated. An educator who only knows the inherited format is defenseless against its obsolescence; one who knows why they ask for what they ask for can change the format without losing the substance. The difference between the two is not provided by any tool tutorial, which expires with every update. It is provided by deeper training: knowing what it means to comprehend, what it means to think for oneself, and what distinguishes information from knowledge. Questions from the philosophy of education that have suddenly become departmental urgencies.

At Nomba Education we formulate it this way: we do not train executors, that is what AI is for; we train creators. This applies to students and, first, to their teachers. A school faculty that has thought deeply about what is irreplaceable in their profession does not fear the tool: they put it in its place, which is beneath them. Homework in the age of AI will be corrected, as always, by a human being who knows what they are looking for. The training of that human being is the true pending task.

Cover image generated with artificial intelligence.