The Same AI Does Not Mean the Same Opportunity to Learn

It is tempting to imagine that giving everyone access to the same AI will give everyone a similar opportunity to learn. The tool can answer patiently, adapt its language, and remain available outside classroom hours. Those possibilities matter greatly, especially where other support has been difficult to obtain.

Still, access is a beginning rather than a complete educational arrangement. What a person can do with an answer depends partly on the conditions around them. If we overlook those conditions, we may celebrate a lower barrier while leaving another barrier untouched.

The same learner, under different conditions

In the previous part, I considered how we notice the boundary of our understanding. Now imagine the same learner in two different circumstances. In one, they have time to study, a basic introduction to the subject, and someone who can discuss an uncertain answer. In the other, they are studying after a long working day, in a less familiar language, with nobody available to check their interpretation.

The AI may be identical in both situations. The opportunity to learn is not. One situation allows the learner to turn an answer into a question, test that question, and receive feedback. The other may encourage them to accept the first usable response simply because their time and energy are limited.

This is a hypothetical comparison, not a judgement about particular groups. Its purpose is to shift attention away from supposedly capable and incapable people. The same person can engage very differently when the available support changes.

Educational inequality is often discussed as a difference in possession: who has a device, a connection, or a subscription. We should also ask who has the conditions that make these resources useful. A screen can provide access to explanations without providing a secure place to ask what those explanations mean.

Foundational knowledge still matters

Prior knowledge helps a learner place new information. It can make an unusual claim easier to notice, suggest a worthwhile follow-up question, or reveal that a familiar word is being used in an unfamiliar way. AI can help develop this foundation, but it does not make the foundation irrelevant.

Imagine reading an AI-generated explanation of a scientific result. Someone familiar with the difference between correlation and causation may notice an unjustified conclusion. Someone who has not yet been taught that distinction may find the same passage entirely persuasive. The second person needs an opportunity to learn the distinction, not criticism for lacking it.

Language can create another layer. A translation may make a document accessible, yet leave the reader uncertain about an institutional term or a qualification hidden in the original phrasing. What appears to be a language problem may also be a problem of context.

These are reasons to continue investing in education, rather than reasons to restrict AI to people who already know enough. Foundational teaching and AI assistance can support one another. A patient introduction can make the tool more useful; the tool can then provide additional ways to practise and explore.

An inequality that could grow, not a future already decided

I am concerned that AI could widen some educational differences. People with stronger foundations, more time, and better guidance may be able to use it to extend their thinking, while others receive mainly convenient answers. If that pattern becomes established, apparent equality of access could coexist with unequal growth in capability.

This is a risk to examine, not a claim that every educational gap has already increased because of AI. Different tools, teaching arrangements, and institutional choices can produce different outcomes. It would be as careless to declare inevitable harm as to promise automatic equality.

UNESCO's guidance on generative AI in education and research offers a relevant policy perspective. It emphasises human-centred use, human capacity, and ethical, safe, equitable and meaningful adoption. That guidance is not experimental evidence that a particular inequality has worsened. It identifies responsibilities that should shape how the technology is introduced.

The practical implication is that distributing access should not be the only measure of success. We also need to ask whether learners can make sense of what they receive, participate safely, and develop the capacity to question and use it.

Support has to reach the moment of difficulty

A course may introduce an AI tool and demonstrate several impressive uses. The learner's real difficulty may emerge later, when an answer contradicts another source or when a practice problem cannot be solved with the example they remember. Support needs to reach that moment, not only the introductory demonstration.

Sometimes the missing element is feedback. Sometimes it is a teacher who recognises that a learner needs a more basic explanation. Sometimes it is a peer who makes asking less intimidating. These forms of support are different, and no single tool should be expected to reproduce all of them.

For a learner balancing work, family, and study, flexibility can be especially important. Yet flexibility should not mean being left alone with unlimited material. A manageable sequence, an achievable next step, and a clear route to help may be more valuable than another large collection of resources.

I think good educational design makes these routes visible. It tells a person where to begin, what they are working towards, and what to do when they cannot proceed. That structure can protect the usefulness of AI instead of competing with it.

A kinder way to discuss capability

When a learner accepts a weak answer, it is easy to say that they should have asked a better question. Sometimes that is true, but it does not explain how they were supposed to know what a better question would be. The ability to ask develops through experience, examples, feedback, and encouragement.

We should therefore be careful about making confident AI use a new test of personal worth. Someone who needs help identifying a source or breaking down a problem is not less deserving of the technology's benefits. They may be precisely the person for whom thoughtful educational support makes the greatest difference.

That support need not remove difficulty. Learning often requires sustained effort. It should make the effort meaningful and possible, while helping the learner distinguish a productive challenge from a situation in which they simply lack the necessary guidance.

Equal access needs a deeper companion

I would like the promise of accessible AI to be accompanied by a promise of accessible understanding. We cannot guarantee identical outcomes, but we can take responsibility for more than placing a tool in someone's hands. We can build environments in which questions are welcomed and judgement can develop.

One part of that responsibility is learning how information is produced, presented, checked, and shared. If we teach only how to obtain a response, we leave learners without enough support for deciding what that response deserves. The next part considers this wider form of literacy.

Next: “What We Need to Learn Beyond Using AI,” available on 22 October 2026 at 9:00 AM KST. What should education teach when information can be generated more quickly than we can evaluate it?

Source note: UNESCO, “Guidance for generative AI in education and research” (2023). This is policy guidance for human-centred and equitable adoption, not evidence that AI inevitably increases educational inequality. https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research