The Homework Machine: AI, Learning and the Skipped Struggle
Written by The Pilgrim ·
What do we actually lose when a student skips the struggle and accepts a polished answer from a machine? Is the finished answer the point of education, or is it merely its least important product? And if difficulty is where learning lives, what becomes of the learner who has been quietly relieved of all difficulty?
These are not questions about technology in any narrow sense. They are questions about what education is for, and they have been worth asking long before artificial intelligence entered the classroom. But the arrival of capable language models, tools that can produce coherent essays, solve mathematical problems and summarise complex arguments in seconds, has made the questions urgent in a way that can no longer be deferred. Students across every level of education now have access to a homework machine, and many of them are using it. The machine is impressive. It is also, in a particular and important way, dangerous.
The danger is not that the answers are wrong, though they often are. The danger is that they are finished. When a student sits with a difficult problem, something happens in that sitting. There is confusion, and then a partial clearing of confusion, and then fresh confusion again. There is the moment of reaching for a word that will not come, and the slower moment of finding something close to it. There is the frustration of a paragraph that refuses to cohere, followed by the quiet satisfaction of the sentence that finally does. None of this is comfortable. All of it is necessary. The cognitive scientists call it desirable difficulty, the idea that obstacles placed in the path of learning, when they are of the right kind, actually strengthen what is retained and understood. The struggle is not the unfortunate price of learning. It is the mechanism by which learning occurs.
When a student delegates that struggle to a language model, what arrives in return is the surface of thought without the substance of it. The essay exists, but the thinking that would have produced it has not taken place. The answer is present, but the capacity to reach it independently has not grown. This is the central deception of the homework machine: it appears to help the student while actually depriving the student of the only thing that education can genuinely offer, which is the transformation of the person doing the work.
One might ask whether this is truly new. Students have always found ways to avoid hard thinking. They have copied from one another, bought essays, skimmed when they should have read. This is all true, and the temptation to take the easier path is not an invention of the digital age. But there is a difference of scale and of friction. Previously, outsourcing your essay required effort, social coordination and a degree of moral self-consciousness. Now it requires typing a sentence into a box and waiting several seconds. The barrier has been reduced so dramatically that avoidance has become the path of least resistance rather than the path of deliberate evasion. That shift in friction is a shift in kind, not merely in degree.
There is also something worth examining in the way students understand what they are doing. Many of them, one suspects, do not experience using an AI tool as cheating. They experience it as getting help, as using a resource, as being efficient. The language of productivity has colonised the language of learning, and within that vocabulary, spending two hours on a difficult problem looks wasteful while generating an instant answer looks smart. But education was never meant to be efficient in that sense. It was meant to be formative, which is a word we would do well to reclaim. The point is not to produce outputs. The point is to produce a certain kind of person, one capable of sustained attention, of tolerating uncertainty, of revising their own thinking in the face of evidence.
What is quietly lost when the effort is outsourced is precisely the formation of that person. A student who has never wrestled with a difficult argument has not developed the intellectual muscle needed to evaluate one. A student who has never fought to express a complex idea in their own words has not built the relationship with language that makes genuine communication possible. These capacities do not emerge from reading finished texts, however excellent. They emerge from the painful, private, productive work of trying and failing and trying again. Remove that work, and you remove the conditions under which a thinking person is made.
None of this is an argument for treating artificial intelligence as an enemy. It is a tool, and like all tools, its value depends entirely on how it is used. A student who uses a language model to interrogate their own argument, to seek a counterpoint they had not considered, or to check whether their reasoning holds under pressure, is using it in service of their own thinking. A student who uses it to replace their own thinking altogether is doing something categorically different. The difference matters enormously, and it is the kind of difference that critical thinking exists to illuminate.
Teachers and institutions bear some responsibility here, not in policing every submission, but in making the purpose of the struggle visible and legible to students who may never have had it explained to them. If we do not tell students why the difficulty is the point, we should not be entirely surprised when they route around it. The case for productive struggle needs to be made, and made well, and made repeatedly.
But the deeper responsibility belongs to the students themselves, and perhaps to each of us who has ever reached for the easier answer when the harder one was available. What kind of mind do we want to inhabit by the time we are finished learning? What are we willing to earn, and what are we willing only to borrow? The homework machine will always be there, patient and generous and entirely indifferent to our growth. The question is whether we will choose, in spite of everything, to do the work ourselves.