When AI Writes the Work, Who Does the Thinking? The Enterprise Stakes of Cognitive Offloading

When fiction professor Micah Nathan began reading the first student submissions of the semester at Massachusetts Institute of Technology, he sensed something was wrong almost immediately. The stories were polished, grammatically impeccable, and structurally sound. Characters developed on cue, metaphors appeared exactly where they should, and every sentence seemed to glide effortlessly into the next. On paper, they looked like competent short stories. Yet they lacked the one quality that matters most in any piece of fiction: the unmistakable presence of a human mind struggling to say something real. Nathan recognised what many experienced readers now instinctively detect. The stories had been generated by artificial intelligence.

What happened next did not become a disciplinary case or a debate over academic misconduct. Instead, it evolved into one of the most revealing classroom conversations of Nathan's eight-year teaching career. In confronting two students who admitted they had used AI to write their workshop submissions, Nathan uncovered a truth that extends far beyond the walls of MIT. The issue was not simply whether students should be allowed to use AI. It was whether they still understood what writing is for.

For many students, especially those immersed in science, technology, engineering, and mathematics, writing can feel intimidating because it lacks the certainty of equations and formulas. In technical disciplines, there is often a correct answer and a clear method for arriving at it. Fiction offers no such guarantees. A story either moves the reader or it does not. A sentence either feels true or it falls flat. The feedback can be deeply personal because writing is not just a technical exercise; it is a visible expression of thought, emotion, and perception. To submit a story to workshop is to expose one's inner world to public scrutiny.

That vulnerability, Nathan discovered, lies at the heart of why students turn to AI.

One student, fighting back tears, confessed that she used AI because she was afraid of looking stupid. She loved writing, she said, but could not bear the possibility that her classmates might see her work as immature or flawed. What began as a simple grammar check became a series of increasingly invasive suggestions until the AI offered to rewrite the entire story. She accepted. Another student admitted he had a compelling idea but no idea how to begin. Rather than risk stumbling through the opening pages, he outsourced the task to a machine.

Their confessions exposed a deeply human motivation behind AI use: fear.

Fear of imperfection. Fear of criticism. Fear of confronting the gap between imagination and execution.

ChatGPT promises relief from all three. With a single prompt, it transforms uncertainty into polished prose, replacing the anxiety of creation with the illusion of competence. For students accustomed to high achievement, that promise can be irresistible.

Nathan does not deny that AI-generated writing can appear impressive at first glance. In fact, he describes it as "perfectly mediocre." It is competent, coherent, and stylistically plausible. But it is also emotionally hollow. The prose imitates thought without actually thinking. It mimics feeling without having felt anything. Nathan borrows Alfred, Lord Tennyson's famous phrase from Maud to capture this quality: "Faultily faultless, icily regular, splendidly null."

Those four words may be among the most precise descriptions of AI writing ever written.

The sentences are technically correct. The structure is orderly. The language is smooth. And yet, somehow, nothing is there.

Experienced readers sense this absence even when they cannot fully articulate it. The story moves, but no consciousness seems to inhabit it. The text performs the gestures of literature without revealing the lived experience that gives literature meaning.

By contrast, student writing is often gloriously imperfect.

Characters behave inconsistently. Dialogue sounds unnatural. Plots lose direction. Sentences wobble under the weight of what the writer is trying to say.

Yet these flaws are not signs of failure. They are signs of effort. They show a mind grappling with ideas that are not yet fully formed.

Nathan compares novice writers to foals learning to walk. Their legs tremble. Their movements are awkward. But within those stumbles lies the promise of future grace. If a foal emerged from the womb already running flawlessly, we would know something essential had been bypassed. The same is true of writing. The awkwardness is evidence that learning is taking place.

This insight touches on a central truth about education: the value of writing lies not primarily in the finished product, but in the transformation that occurs during the process.

Writing is a form of thinking.

Students begin with fragments of emotion, intuition, and half-formed ideas. Through drafting and revision, they discover what they actually believe. The struggle to choose the right word, reorder a sentence, or clarify an image forces them to refine their understanding. When AI supplies the polished draft, the visible product remains, but the intellectual journey disappears.

The student receives a story. But not the growth.

Nathan's concerns align with a growing body of research into what psychologists call cognitive offloading,the practice of outsourcing mental effort to external tools. A 2025 study from the MIT Media Lab reported that participants who used ChatGPT to write essays exhibited lower neural connectivity than those who wrote independently. Other emerging studies suggest that excessive reliance on AI may reduce persistence, problem-solving endurance, and executive functioning.

Although the research remains early, the underlying concern is intuitive. Skills strengthen through use and weaken through neglect. If students consistently outsource difficult thinking, they may gradually lose the mental stamina required for independent analysis and creative work.

The debate in Nathan's classroom quickly expanded beyond fiction. Students asked thoughtful and entirely reasonable questions. If AI helps them articulate their ideas, why is that wrong? How is using AI different from working with a human editor? At an institution renowned for pioneering artificial intelligence research, should they not be encouraged to use the most advanced tools available?

Nathan's answer was nuanced. He does not reject AI as inherently unethical or dangerous. AI can be tremendously useful for brainstorming, editing, and handling routine tasks. But fiction writing serves a different purpose. It is not about efficiency. It is about self-discovery.

To support this argument, Nathan turned to George Orwell and his 1946 essay Confessions of a Book Reviewer. Orwell describes the numbing effect of producing endless responses to books that no longer evoke genuine reactions. Over time, the process erodes judgement and reduces criticism to performance. Nathan sees a modern parallel in AI-generated writing. Students submit work that looks like thought, but the underlying intellectual engagement has not occurred.

The danger, then, is not simply bad writing. It is simulated thinking.

This distinction matters far beyond the classroom. In an economy increasingly shaped by AI, the temptation to bypass cognitive friction is becoming universal. Professionals can now generate emails, presentations, reports, marketing copy, and software prototypes in minutes. The efficiency gains are undeniable. Yet many of the activities most essential to human development derive their value precisely from the effort they require.

Writing. Design. Problem-solving. Decision-making.

The friction is not a flaw in these processes. It is the mechanism through which understanding is built.

Nathan has since revised his syllabus to state clearly that he does not want students using AI to write their stories. This is not a moral judgement against technology. It is a pedagogical decision rooted in the purpose of workshop. Peer review only works when there is a real author behind the words — someone whose thinking is visible on the page and who can respond meaningfully to criticism.

In this sense, the classroom becomes more than a place to teach craft. It becomes a sanctuary for authorship, a space where students are allowed to struggle, fail, and gradually discover their own voices.

Since that pivotal workshop, Nathan says the atmosphere in his classes has changed. Students speak more openly about frustration and self-doubt. They have begun to understand that difficulty is not evidence of inadequacy, but a sign that important work is taking place. The moments when words fail are no longer treated as proof of weakness. They are recognised as essential steps in the creative process.

That may be the most enduring lesson AI can offer educators.

Machines can generate text with astonishing fluency. They can imitate voice, structure, and style. But they cannot replicate the inner transformation that occurs when a person wrestles with ideas and forces them into language.

And in education, that transformation is the point.

In the end, Micah Nathan is not asking his students for perfection. He is asking for something far more valuable: their thoughts, their voices, their uncertainties, their unfinished sentences, and the imperfect but unmistakably human struggle that turns language into meaning.