Here’s an Assignment We Can’t Avoid Any Longer: Reinventing Homework for an AI World

Today, any student can photograph a class assignment, upload it to a chatbot and receive a complete solution in seconds. This hasn’t created a homework problem; it has exposed one. For as long as anybody can remember, homework has been an educational constant. An almost automatic extension of the school day, legitimized by tradition, social pressure and a seemingly reasonable intuition: practice makes perfect.

However, that logic, which was never entirely clear, begins to crack with unusual speed. And not only for pedagogical reasons, but also because of a technological transformation that has profoundly altered the very meaning of “doing homework”.

The most visible trigger is generative AI. The fact that any student can get a chatbot to provide an essay that is indistinguishable from one written by a human means that the cost of appearing to be learning has fallen practically to zero. And that, in educational terms, is devastating.

But it would be a mistake to think that AI has created the problem. Because if a task can be outsourced without the student having to think, it is likely that the task was already poor from the beginning. AI has not rendered homework useless: it has exposed the fragility of its design.

The data accompanies that feeling. A majority of teenagers already use chatbots, with more than half that number using them for schoolwork, while a significant proportion admit to using them intensively. At this point, continuing to defend the traditional model of homework ignores the learning environment in which students operate today.

At the same time, something is also changing from within the system. Some school districts in the United States are beginning to abandon mandatory homework, arguing that they don’t significantly improve academic performance and can increase family stress. It’s not an ideological revolution, but a pragmatic adaptation: when a tool stops working, it’s time to rethink it.

For years, research on homework has been contradictory: some studies find modest benefits, others virtually none, and many stress that variables such as motivation, self-regulation, or context outweigh time spent. But there is a fact that changes the discussion: an article with precise measurements of study time based on real data, not on questionnaires, demonstrates with more than 6,000 students from three countries solving more than 1.1 million exercises that there is a clear and quite intuitive pattern: dedicating more time to study improves performance in mathematics, especially in students with a lower starting point, and these effects can be maintained over time when learning is real.

But here’s the key: it’s not simply about “doing more,” it’s about how it’s done. The study itself insists that the quality of the study, that is, improving while practicing, is decisive. And that completely changes the discussion. Because what this work analyzes is structured practice in a digital environment with immediate feedback, adaptation and follow-up. That is, the exact opposite of the repetitive worksheet that a student can photograph and solve with a chatbot without needing to think at all.

The evidence on overload does not help to defend the classical model either. We have known for some time that beyond a certain threshold, around two hours a day, the benefits stagnate and negative effects such as stress and exhaustion appear.

The academic results do not invite optimism either. Recent data points to a worrying deterioration, especially among the worst-performing students, which shows that even the traditional homework model is not fulfilling its reinforcing function. In response, teachers are already reacting. A significant proportion claim to have reduced the amount of home study in recent years, with others simply stopping assignments.

The conclusion is beginning to be clear: homework, as we have understood it for decades, is obsolete. Not because all work outside the classroom is useless, but because its dominant format (repetitive, mechanical, easily delegable) has ceased to make sense in a world with AI.

But abolishing homework without further ado is not the solution either. Because, as the evidence shows, learning takes time. Without time investment, there is simply no skill acquisition. The question, therefore, is not whether homework should disappear, but whether we are capable of reinventing it.

“Good homework” in the age of AI cannot consist of solving routine exercises or writing generic texts. It has to demand something that requires significant student intervention: comparison, judgment, contextualization, explanation, error detection. It has to leave a cognitive trace. That implies, for example, asking students to use AI explicitly and critically: to document how they have used it, to contrast answers, to identify limitations, to explain their decisions. It is not a question of banning the tool, but of making it part of the learning process, because they will live in a world in which AI will play an important role.

It also involves recovering forms of assessment that are more difficult to automate: presentations, oratory and applying knowledge in real contexts. In a world where text can be generated automatically, comprehension has to be demonstrated in another way. And, above all, it implies accepting an uncomfortable idea: much of the homework assigned over the years was ineffective, and is now irrelevant.

In Louisiana, a school district has decided to eliminate mandatory homework, the Facebook post announcing the move garnering the most Likes in its history. Freeing up time for rest, reading, or creativity is not lowering the bar: it may, in fact, be the only way to raise it again.

Homework has finally been laid to rest. The question now is how we use AI to benefit a greater number of students, particularly those who face challenges at home and school that impair their performance or who require a different approach to learning.

—

This post was previously published on Enrique Dans’ blog.

***