The real danger is not that students will use artificial intelligence, but that they will stop practising the mental labour through which judgement, memory and creativity are formed. When answers arrive too quickly, learners may bypass uncertainty, effort, revision and reflection, gradually losing the habits that make independent thought, original expression and responsible decision-making possible.
A child asks an artificial intelligence tool to explain a poem. Within seconds, the machine produces a fluent summary, identifies the imagery, offers a theme, and even suggests a paragraph for an assignment. The answer looks useful, perhaps even impressive. But a quieter question remains: what did the child not have to do?
That question should now sit at the centre of education policy. Artificial intelligence is entering classrooms, homework, tutoring, assessment and communication faster than most schools and universities can govern it. The public debate has largely focused on plagiarism, misinformation, privacy and bias. These are serious concerns. Yet they do not reach the bigger educational risk. The greater danger is that students may become visibly productive while becoming less practised in the mental labour through which learning actually happens.
This is the difference between cognitive offloading and cognitive substitution. Human beings have always used tools to extend the mind. A notebook helps memory. A calculator helps with computation. A dictionary helps expand vocabulary. Such tools are not the enemy of learning. But generative AI is different in scale and character. It does not just store information or speed up a calculation. It can summarise, draft, translate, solve, compare, evaluate and advise. It can perform many of the visible signs of understanding before the learner has had the chance to struggle towards it.
Education is not valuable simply because it produces correct answers. It is valuable because it forms persons capable of remembering, interpreting, questioning, writing, judging and creating. These capacities do not appear fully formed. They are built through repetition, difficulty, feedback, error and revision. A student who never retrieves an idea from memory will not strengthen their memory. A student who never writes a first draft will not develop compositional stamina. A student who never compares weak arguments will not build judgement. A student who is spared every difficulty may also be deprived of the very conditions that make intellectual independence possible.
The everyday examples are already familiar. A student asks AI for the “main points” before reading the chapter. Another asks it to “make this sound better” before discovering her own voice. A third requests a solution to a mathematical problem before attempting to understand the problem’s structure. Each act appears small, even harmless. But repeated over months and years, these small bypasses may change the learner’s relationship with effort. The risk is not a single dramatic act of replacement. It is the slow normalisation of not trying first.
The Seduction of Speed
AI has entered education with promises of speed, accuracy, personalisation, and optimisation, addressing real pressures. Teachers are overworked, students overloaded, and institutions are expected to do more with less. Tools that reduce administrative tasks, generate feedback, support multilingual learners, or enhance accessibility should be seriously considered.
AI support differs from replacement; assistance is not replacement, hints are not answers, and feedback differs from authorship. A scaffold must help learners stand on their own; if it never disappears, it creates dependence.
This matters most for children and young learners, whose executive function, attention, and sense of ownership are still developing. Introducing unregulated reliance on AI can hinder their cognitive growth.
Some educational obstacles are wasteful and should be removed, but not all effort is bad. Productive struggle, retrieval, repeated writing, and slow reading build durable knowledge and judgement. A system too frictionless may impair learning, as pressure fosters competence.
The language of convenience must be used carefully: while it can aid struggling students and support revision or structure, it can also prevent persistence, replace reading, or rob discovery. Tools can support or hinder learning depending on how they are used.
From Classroom Dependency to Social Vulnerability
The problem does not end at the classroom door. A learner who grows accustomed to machine-generated answers may become an adult less prepared to resist machine-shaped choices. Digital platforms already organise attention, rank options, personalise advertisements, predict preferences and steer decisions. Shopping suggestions, travel destinations, entertainment feeds and political messages increasingly arrive via systems designed to capture attention and influence behaviour.
This is why cognitive substitution in education may translate into decision vulnerability in society. A student who has not practised doubt is more easily reassured by authoritative-looking output. A student who has not practised comparison is more easily steered by rankings and recommendations. A student who has not practised writing is more dependent on borrowed expression. A student who has not practised judgement is more exposed to decision architectures designed by platforms, advertisers and political actors.
Brian Solis once described “digital Darwinism” as the condition in which technology and society evolve faster than institutions can adapt. Education now faces precisely that mismatch. AI tools are moving faster than curriculum policy, assessment design, teacher training and child-protection frameworks. In that vacuum, vendors and platforms can quietly become pedagogical authorities. If an AI tool decides what a student reads, how feedback is framed, which errors matter and which learning path appears optimal, then a portion of educational judgement shifts away from teachers and public institutions.
Policy Must Arrive Before Dependency
International policy advocacy is beginning to address AI in education. UNESCO calls for guidance on generative AI, emphasising human agency, teacher roles, inclusion, and age appropriateness. The EU’s AI Act labels certain educational AI as high-risk, while the U.S. Department of Education promotes human involvement, transparency, and trust. The UK notes that evidence on pupil AI use is still emerging and warns schools to exercise caution.