What We Need to Learn Beyond Using AI

Learning to use AI can begin with a practical skill: describing what we need clearly enough to receive a useful response. That skill deserves attention. But it cannot carry the whole responsibility of education in an environment where information can be generated, rewritten, and distributed almost instantly.

We also need to understand what we are receiving, what it leaves out, and what happens when we pass it to someone else. This is where Media and Information Literacy, or MIL, becomes important. It gives us a wider frame than learning to operate a particular tool.

Literacy includes the life of information

UNESCO connects MIL with critical thinking, ethical awareness, and responsible participation in media and information environments. Its discussion of digital competencies places AI literacy within this wider context. The concern extends beyond distinguishing a true statement from a false one.

Information has a source, a purpose, an audience, and a route through which it reaches us. A message may be factually accurate while omitting a relevant limitation. An image may be authentic while appearing in the wrong context. A summary may faithfully reproduce one perspective while leaving another invisible.

An AI answer introduces additional questions. Which claims can we trace to reliable evidence? Has the response silently interpreted an ambiguous request? What assumptions connect the facts to the conclusion? These questions matter even when the language is polite and the formatting is excellent.

MIL does not require treating everything as suspicious. It helps us decide what attention a piece of information needs, and why. The aim is more informed participation, not a permanent inability to trust anything.

A practical way to restore context

Mike Caulfield's SIFT approach offers four useful moves: Stop, Investigate the source, Find better coverage, and Trace claims, quotations, or media to their original context. The approach encourages us to step outside an isolated item instead of evaluating it only by how convincing it looks.

Stopping can mean pausing before sharing something that immediately confirms what we believe. Investigating the source means considering who is speaking and what relevant expertise or interests they have. Finding better coverage means looking for dependable work on the underlying claim, rather than collecting more repetitions of the same item.

Tracing to the original context is particularly useful with AI-generated summaries. Suppose an answer says that a study proves a teaching method works. Opening the original study may reveal that it examined one subject, one age group, and a particular comparison. The finding may remain valuable, but its meaning becomes more precise.

This example is hypothetical, yet it also describes the discipline I have tried to apply in this series. Research can help us think without being made to promise more than it actually examined. Respecting a study includes respecting its limits.

Verification should match the consequence

It would be exhausting to investigate every casual suggestion with the same intensity. A proposed title for a notebook and a claim affecting a student's qualification do not carry the same consequences. Good judgement includes deciding where careful checking is necessary.

We can begin by asking who might be affected if the information is wrong. Does it influence health, money, rights, educational access, or another person's reputation? If so, we should seek appropriate authoritative sources and qualified guidance, rather than rely on the confidence of a generated answer.

For less consequential tasks, a lighter check may be sufficient. We might compare examples, inspect a citation, or ask which assumption changes the conclusion. The standard should be proportionate, but proportionate does not mean absent.

Source quality also depends on the question. An official policy document can establish what an institution states; it cannot by itself establish that the policy achieved its intended effect. A personal account can reveal experience; it cannot alone show how common that experience is. Learning to make these distinctions is part of learning to use evidence.

We become producers as well as readers

AI makes it easier to create a polished message before we have fully considered its audience. A learner can produce a persuasive essay, an organisation can generate a public explanation, and a colleague can circulate a summary. At that point, literacy includes responsibility for what we send.

If I use a generated paragraph in my own communication, the other person encounters it as my message. They may reasonably expect that I have checked its meaning and can explain its relevance. The tool's involvement does not remove my responsibility for the decision to share it.

There are also questions of privacy and fairness. Does the material contain another person's confidential information? Does the explanation represent a group through an oversimplified stereotype? Have I made an uncertain claim sound settled? These concerns cannot be resolved simply by improving the prompt.

Responsible creation need not become a declaration that AI is unwelcome. It means considering disclosure where required or relevant, following the rules of the learning or working environment, and being honest about what we have actually examined. The recipient deserves more than text that happens to look complete.

Education has to make judgement practicable

A lesson about MIL should do more than tell learners to think critically. That instruction can be difficult to act on without examples. Learners need opportunities to compare a summary with its source, identify an omitted condition, and discuss why two credible accounts emphasise different aspects of the same event.

These activities also need foundations. Someone cannot evaluate an unfamiliar technical claim solely through a general checklist. Subject knowledge, language support, and access to appropriate expertise remain important companions to information literacy.

A teacher's role can be especially valuable in making the process visible. They can show not only which answer they accept, but which uncertainty they noticed and how they decided what to check. That demonstration teaches a way of working, rather than merely delivering another conclusion.

I would like educational environments to make room for revising an answer after checking it. If correction is treated only as failure, learners may defend a weak claim instead of improving it. Responsible judgement includes being able to say that a source changed what we thought.

From operating a tool to participating responsibly

Knowing how to obtain an answer is useful. Knowing how to interpret, question, apply, and share it gives that skill a more dependable foundation. MIL brings these activities into the same conversation, reminding us that information is part of a relationship with other people.

The remaining question is how to make this relationship livable. We cannot check everything endlessly, and we should not have to refuse assistance to prove that we are learning. The final part will consider a practical balance: how AI can extend our thinking while we continue to develop the capacities that make its help meaningful.

Next: “How Shall We Learn and Live with AI?” available on 24 October 2026 at 9:00 AM KST.

Sources: UNESCO, “Media and Information Literacy and Digital Competencies” (2025). https://www.unesco.org/en/articles/media-and-information-literacy-and-digital-competencies

Mike Caulfield, “SIFT (The Four Moves)” (2019). This is a practical information-evaluation approach, not a guarantee that every claim can be settled through a checklist. https://hapgood.us/2019/06/19/sift-the-four-moves/