The Ghost in the Machine

BACK-TO-SCHOOL ESSAY: How AI is killing take-home assignments and forcing education to rely on labs and tests

As a student, I found a sort of comfort in the quiet, deliberate work of an assignment. There was something reassuring about taking time to refine thoughts and handing in a thing that was my own. Tests were another matter — annoying and stressful. Eventually, and this may surprise some readers, I earned a master’s degree in computer science, with a specialization in artificial intelligence. Back then the challenge was distinguishing a circle from a square.

And here we are. Generative AI today can write a plan for taking over the world, for making a better lasagna, and for anything in between. It is forcing a fundamental shift — in the labor market obviously, but also in education.

Among other disruptions, the traditional model of homework, essays, and take-home assignments is becoming obsolete. With AI tools capable of producing well-structured essays, solving complex mathematical equations, and even mimicking personal writing styles, educators face a wrenching dilemma: how can academic integrity be maintained when students can generate perfect assignments with the click of a button?

The conclusion, however inevitable, seems unfortunate: education is drifting toward in-person assessments, timed tests, and hands-on demonstrations carried out under watchful eyes, with devices firmly out of reach.

The very power that makes generative AI astonishing — its ability to absorb oceans of data and return fluent, human-like responses in seconds — is what undermines traditional assignments. A student can now submit an essay indistinguishable from original thought, even tailored to sound like their own voice. Plagiarism software is helpless, since the words are not copied but freshly minted. And so, the take-home essay, once a cornerstone of learning, begins to lose its credibility.

This loss is not trivial. Assignments and exams measure different things, cultivate different skills, and demand different kinds of effort. In what might be called the “Before Times,” assignments encouraged depth: long hours of research, careful synthesis of ideas, the crafting of a persuasive argument. They nurtured creativity, allowing students to explore tangents, experiment with structure, and weave external sources into their own voice. Perhaps most importantly, they taught patience and time management: the discipline of planning, revising, and refining over days or weeks.

Exams, by contrast, prize immediacy. They test recall under pressure, procedural fluency, and the ability to perform on command. Conducted in controlled environments, they can measure knowledge cleanly. Yet they also strip away the space for reflection and creativity, rewarding standardization and often penalizing the student whose strength lies in slow, deliberate thought.

What can be done?

Some see this is just an extension of the old cheating paradigms—basically, that parents used to do homework as well. It was always cheating. Of course, what this misses is that AI is incomparably better than the average parent. What AI has done, argues engineering professor Josh Brake, is not to kill homework but to expose what has always been true: assignments only matter if students choose to engage with them. Workarounds—whether from parents, older students, the internet, or now AI—have always existed. The “homework apocalypse” is thus perhaps less about destruction than revelation: the illusion that teachers can enforce learning through control is gone. What remains is the real task—building trust and helping students see the value in doing the work for their own growth.

Of course, few are the parents or older students who could compete with AI for most purposes.

So the direction of travel is clear. US national data shows (Or is it “show”? I must ask AI) a meaningful uptick in the number of students reporting no assigned homework on a typical day: 37% of 13-year-olds in 2023, compared to only 29% in 2020 and 21% in 2012. In Australia, schools like St Benedict’s in Townsville have ended mandatory homework — except for daily reading — citing research that finds minimal benefit for young students. Similar trends are reported all over the world — the question is what can come instead or what adjustments can be made.

California is weighing a “Healthy Homework Act” to curb excessive assignments, while China’s Double Reduction Policy already caps daily homework by grade level and curtails off-campus tutoring. Poland, too, has moved to sharply restrict primary school homework in line with political promises. Even at the university level, instructors are adapting to student disengagement by loosening deadlines and reducing workload expectations. Together, these changes point to a broad recalibration of homework across education systems.

Some educators argue that assignments can still be rescued. Experiments abound: monitored in-class writing sessions, hybrid models that require research at home but completion on site, portfolios assembled over time that make outsourcing harder, even assignments where AI is not banned but interrogated — students asked to critique, improve, or contextualize what the machine produces. These ideas are clever, but none erase the central problem: as long as AI can do the work, doubt clings to the authenticity of the product.

And so the tide pushes education toward tests, orals, and laboratories. Universities will tighten their rules: more handwritten exams, more closed-book conditions, more oral defenses of written work. Surveillance, biometric authentication, even AI-powered proctoring may become commonplace.

Professors know this. I hear it when I speak with them: some ignore, some adapt, many despair, and at least one I know has walked away. The consensus, whether spoken aloud or not, is that a profound issue is upon us. “We’re totally fucked,” one American professor told me, preferring not to be quoted by name because he would be fired for stating the obvious. “We’re raising a generation that will be unlike any other in post-industrial society. It’s an honor system, and we’ll see where that goes.”

Many of those who remain more optimistic argue that schools at least should be an AI-free zone by banning phones. The case for doing so is strong: smartphones are unrivaled distraction machines, fragmenting attention and pulling students away from the sustained focus that learning demands. Teachers report that social interactions during breaks have collapsed into silent scrolling, while incidents of cyberbullying and surreptitious filming have multiplied. Evidence from countries like France, which banned phones in schools in 2018, suggests such measures can improve concentration and even narrow achievement gaps by helping disadvantaged students most. In this view, creating a phone-free environment is less about technophobia than about preserving the rare spaces where children can concentrate, connect, and learn without the constant tug of the digital world.

Still, outright bans raise real complications. Parents often want their children reachable in emergencies, and phones can be potent educational tools when used well, whether for quick research, language practice, or coding apps. Critics of bans argue that shielding students from devices only postpones the need to teach responsible use in a world where smartphones will be unavoidable. A growing number of schools are therefore experimenting with middle-ground solutions: locking devices away during the day but allowing them before and after school, or restricting use to supervised educational purposes. Whether phones are banned or tightly regulated, the challenge remains the same — how to give students both the focus they need to learn and the skills they’ll need to navigate a digital society.

Whichever way this goes, the transformation carries a heavy cost. An education built mainly on timed tests will favor a particular kind of student — the quick thinker with nerves of steel — while penalizing the slow and meticulous. It risks elevating memorization at the expense of deep understanding. It discourages the long-form reflection and independent wrestling with ideas that assignments once required.

This is one of progress’s cruel ironies: in solving problems, it erases practices that cultivated resilience, creativity, and independent judgment. The take-home essay was never just a task; it was a rehearsal for adult life, a space where one learned to wrestle with complexity in solitude. Its disappearance signals not only a pedagogical shift but a cultural one.

There’s no way around it: Education must adapt or be steamrolled. Yet we should not pretend that nothing is lost. In trading essays for labs and exams, we risk producing graduates more efficient but perhaps less thoughtful — trained to pass tests and perform tasks, but less practiced in the thinking power that made homo sapiens something special.

It touches something deeper about how knowledge is formed and how character is built. Writing an essay was never only about the grade; it was about wrestling with ambiguity, shaping arguments, and learning to sit with a question until clarity emerged. It was about building stamina for the slow work of thought — a discipline that will be harder to teach in a world where machines do the wrestling for us. It prepared us for gainful employment and productive labor — which, of course, are the other areas AI threatens. That is no coincidence.

My assessment at this juncture is that we need to consider massive, brutal and problematic regulation of AI — something akin to the smartphone bans in schools. Progress always carried a cost, as the Luddites feared, but the net result tended to be positive. For most of history, going with efficiency and economic growth was wise, and paid handsome dividends in longer life spans and better quality of those lives. I am not so sure that this equation will hold for long.

So while we still call this season “back-to-school,” in reality we are actually entering a completely new frontier, with children as the vanguard. En garde!