AI in Education: The Real Risk Is Not Cheating — It Is Outsourcing Thinking
- Krzysztof Kosman
- 11 hours ago
- 8 min read
Artificial intelligence has already entered the classroom.
Students use it to write essays, solve problems and prepare presentations. Teachers use it to create lesson plans, generate exercises and reduce administrative work. Parents are trying to understand whether AI is a useful learning tool or simply another shortcut.
But the most important question is not whether someone used AI.
The real question is: what happened to their thinking in the process?
This was the central theme of my conversation with Dorota Janczak, a digital education expert who has spent more than 20 years helping teachers introduce technology into their work.
Instead of treating AI as either a miracle or a threat, Dorota encourages us to look at something much more difficult: whether the technology supports human development or quietly replaces the mental effort that learning requires.
The problem is not using AI in Education. The problem is using it without thinking
Our conversation began with a public debate in Poland sometimes referred to as “Tokarczukgate”. The discussion started after Nobel Prize-winning author Olga Tokarczuk reportedly mentioned using artificial intelligence in her work. The reaction was immediate. Some people treated the use of AI as a natural part of modern creative work. Others saw it as a betrayal of human creativity.
Dorota’s response was simple:
“I am not ashamed that I use AI. I would be ashamed to use it without thinking.”
This distinction matters far beyond literature.
The same question is already appearing in schools. Is a student cheating whenever they use ChatGPT? Is a teacher less competent because they use AI to prepare materials? Should people disclose when artificial intelligence has supported their work?
A blanket ban does not solve the problem. Neither does unlimited access.
What matters is how the tool was used, why it was used and whether the person remained intellectually responsible for the result.
AI can help someone compare ideas, identify weaknesses, generate counterarguments or improve the presentation of an existing concept.
It can also produce the entire answer while the user contributes almost nothing.
From the outside, both results may look equally polished. From an educational perspective, however, they are completely different.
Cognitive debt: when convenience weakens the ability to think
One of the strongest ideas in the conversation was the concept of cognitive debt.
The analogy is straightforward: an unused muscle becomes weaker. The same may happen with mental abilities that we constantly outsource to technology.
For an experienced adult, this can mean losing some sharpness in skills that have already been developed.
For a child, the risk may be greater. Their “mental muscles” are still being built.
Students need to practise reading, analysing, comparing, questioning, remembering, explaining and creating. These activities require effort. That effort is not an unfortunate side effect of learning. It is the mechanism through which learning happens.
AI tools, on the other hand, are often designed to reduce effort.
They summarise texts, prepare presentations, generate mind maps, suggest answers and create complete written assignments. This makes them useful productivity tools, but it also creates a fundamental conflict.
Technology is designed to save cognitive energy. Education often needs students to spend it.
This does not mean that AI has no place in learning. It means that its role must be designed much more carefully than many schools and technology companies currently assume.
A generated presentation is not evidence of learning
Many generative AI tools appear highly educational at first glance.
Upload a document and the system can create:
a summary,
a quiz,
a presentation,
a podcast,
a mind map,
study notes,
or a complete visual explanation.
The output may be impressive. But where exactly did the learning happen?
A student does not necessarily learn by looking at a presentation generated from a source document. They learn by selecting information, deciding what matters, identifying connections, organising ideas and translating knowledge into their own structure.
When AI performs all these steps, the student receives the result without practising the process.
This creates an illusion of competence.
We read a clear summary and feel that we understand the topic. We look at a well-designed presentation and feel prepared. But recognising information is not the same as being able to recall it, explain it or use it independently.
This problem existed before generative AI. Students have always confused reading notes with learning.
AI makes that illusion much easier to produce and much harder to notice.
We need to assess the process, not only the final answer
Traditional education often evaluates the final result.
A teacher receives an essay, a presentation, a calculation or a completed worksheet. The student is graded on the quality of the finished product.
Generative AI exposes the weakness of this approach.
If the final answer is all that matters, it may be impossible to know whether the student understood the topic, copied from a classmate or generated the work with an AI tool.
The solution is not to create “AI-proof” assignments. Almost any task can eventually be completed with technology.
A better approach is to make the student’s reasoning visible.
Teachers can ask:
How did you reach this conclusion?
Which sources did you use?
Why do you consider this information reliable?
What did you change after receiving feedback?
Which part was difficult?
What surprised you?
Where did you use AI?
Which decisions remained yours?
These questions shift the focus from producing an answer to understanding how the answer was created.
They also prepare students for the real world.
In professional life, people will use AI. The valuable skill will not be pretending that the tool does not exist. It will be knowing how to use it critically, verify its output and remain responsible for the final decision.
Start on paper, then bring in AI
One of Dorota’s most practical recommendations was also one of the simplest:
Start without AI.
Before asking a model to generate ideas, students should first try to produce their own.
Before requesting a summary, they should read and identify the main points themselves.
Before generating a visual, they should decide what they want to communicate.
This sequence matters.
Imagine asking children to brainstorm with an AI system from the first minute. The model will often produce more ideas, more quickly and in more polished language than the children can.
The natural reaction may be:
“I will never come up with something this good, so why should I even try?”
Used too early, AI can weaken confidence and discourage creativity.
Used later, it can support an idea that already belongs to the learner.
A student might first design a poster on paper, select the key message and decide what information should appear. AI can then help improve the visual design.
In this case, the tool is not replacing the student’s thinking. It is helping communicate it.
That is a much healthier division of labour.
Teachers do not need to know every tool
There is a widespread fear that students will always understand new technology better than their teachers.
In some cases, they may know more applications, shortcuts and features.
But that does not make the teacher irrelevant.
A teacher’s value is not based on knowing every new tool. It comes from understanding learning, development, responsibility, ethics and the needs of the student.
Teachers can help students ask better questions:
Is this answer logical?
What could be missing?
What assumptions is the model making?
Is the information accurate?
Who is responsible for the final result?
What personal data should never be entered?
Does this use of AI support learning or avoid it?
These are not purely technical questions.
They are educational and human questions. They are exactly where teachers should have the strongest role.
However, teachers also need support. They are already overloaded, and the expectation that every teacher should independently become an AI expert is unrealistic.
Schools need practical training, shared standards and time for experimentation.
Technology without a change in teaching methods will not improve education. New devices and software cannot compensate for old tasks, old assessment models and a lack of understanding about how students learn.
Parents need to understand AI too
Schools cannot handle this transition alone.
Children may encounter generative AI at home, through friends or on their own devices long before it is introduced during a lesson.
Parents therefore need to understand more than how to activate parental controls.
They need to explain that AI does not “know” things in the same way a person does. It can generate confident, convincing and completely incorrect answers.
It can also feel human.
A chatbot may appear patient, friendly and supportive. It may always be available and rarely challenge the user in the uncomfortable way another person might.
That creates a different type of risk.
Children also need to learn how to deal with disagreement, criticism and frustration. Real relationships involve other people with their own needs, moods and boundaries.
An AI system optimised to keep the user engaged is not a replacement for those experiences.
Parents can reduce the risk by keeping technology use visible and social. Devices can be used in shared spaces. Families can discuss what the child is doing, what they are watching and how an AI system responded.
The point is not constant surveillance.
The point is to make sure that technology does not remove the human conversation around it.
The language we use about AI matters
We often say that AI “thinks”, “lies”, “understands”, “creates” or “wants” something.
These words make the technology easier to discuss, but they also make it appear more human than it is.
Lying requires intention. A language model does not intentionally deceive in the human sense. It generates an answer that may be false.
Creating, understanding and imagining also carry human meanings that do not perfectly describe what these systems do.
This may sound like a small linguistic issue, but it shapes how children and adults relate to the technology.
When we speak about AI as if it were a person, we make it easier to trust, admire or fear it in ways that may not be justified.
Teachers and parents should therefore help children understand what is actually happening behind the interface.
The system may sound intelligent. That does not mean there is a conscious mind on the other side of the conversation.
What should schools do now?
Schools do not need a perfect long-term AI strategy before taking the first step.
But ignoring the topic is no longer a responsible option.
Dorota suggested beginning with a small internal group of interested teachers, supported by school leadership. This group can explore tools, discuss risks and begin drafting basic rules.
Even a temporary rule is better than silence.
A school might initially decide not to use generative AI with students while teachers receive training and experiment privately. That is still a conscious position.
The next steps should include:
Creating clear rules
Students, teachers and parents should understand when AI is allowed, when it is not and when its use must be disclosed.
Training teachers
Training should not focus only on prompts and features. It should include learning science, assessment, privacy, copyright, ethics and cognitive risks.
Testing tools before classroom use
Teachers should first try to learn with the tool themselves. This helps reveal where it supports reflection and where it encourages intellectual shortcuts.
Redesigning assignments
Tasks should make thinking visible through stages, reflection, discussion, drafts and explanations.
Sharing experience
Schools need spaces where teachers can discuss what worked, what failed and what they learned.
Moving slowly, but deliberately
Education is often criticised for changing too slowly. In this case, caution has value. Poorly introduced AI may do more harm than a delayed implementation.
Moving slowly does not mean doing nothing.
It means experimenting carefully, observing outcomes and improving the approach step by step.
AI should strengthen learning, not replace it
The future of AI in education will not be decided by whether schools ban or adopt ChatGPT.
It will be decided by thousands of smaller choices.
Do we ask the student for an answer, or for the reasoning behind it?
Do we use AI before forming an idea, or after?
Do we reward polished output, or genuine progress?
Do teachers hide their use of technology, or model transparent and responsible behaviour?
Do parents leave children alone with persuasive systems, or remain part of the conversation?
Artificial intelligence can support education. It can remove repetitive work, increase access to information, help personalise practice and allow teachers to create new learning experiences.
But it can also help students avoid the exact effort that education is supposed to develop.
The goal is not to protect the old school from new technology.
The goal is to make sure that, in a world full of generated answers, students still learn how to think.
This article is based on a conversation between Dorota Janczak and Krzysztof Kosman for the EdTech Dots podcast.
Watch more conversations about AI, technology and the future of education at:
