Vineeth K Kolarath
On 25-10-2026, I spoke to a group of selected students in Kannur at the district-level Deshabhimani Aksharamuttam Quiz. My topic was simple to say, but harder to practise: use AI as a mentor, not as an answer machine.
This essay is a longer version of that idea.
The answer machine habit
Think about a situation many students will recognise.
You are working on a homework question. It looks difficult, and you are not sure where to begin. So you type it into an AI assistant.
A neat answer appears in seconds. You copy it, perhaps change a few words, and submit it.
The homework is done.
But have you learned how to solve the problem?
That is the question worth asking.
A teacher does not set a mathematics problem because the world needs one more solved equation. The problem is there to help you practise. You are expected to think, try a method, make mistakes and work towards understanding.
The final answer is only one part of that process.
When an AI assistant gives you the complete solution before you have made an effort, it can take away the very practice the question was meant to provide.
The assignment may be finished. The learning may not have happened.
That is what I mean by using AI as an answer machine. It feels efficient because the work gets done quickly. But getting the work done and learning how to do it are not always the same thing.
What a good mentor does
Think about a teacher or mentor who has genuinely helped you learn something.
Did they simply give you the answer every time you got stuck?
Usually, a good mentor does something different.
They ask what you have already tried. They give you a hint when you need one. They let you work through a mistake and help you understand why it happened. They explain the idea behind a method, not just the steps. Sometimes they ask you to explain the idea back to them.
They do not rush to finish the work for you.
Their aim is to help you become capable of doing it yourself.
That is the distinction I want students to remember.
An answer machine gives you something to use. A good mentor helps you develop the ability to produce it yourself.
The same AI tool can play either role
Here is the interesting part. An AI assistant can help in both ways.
It can give you a finished answer. But it can also ask questions, offer hints one at a time, examine your reasoning, point out where you went wrong and explain an idea in different ways.
The tool may be the same. The way you use it changes the experience.
Consider these two requests:
Request one:
"Solve this equation."
Request two:
"I am trying to solve this equation. Do not give me the answer. Ask me one question at a time to help me find it myself."
The first asks the machine to do the work. The second asks it to help you learn how to do the work.
Or consider writing an essay.
Request one:
"Write an essay on the importance of forests."
Request two:
"Here is my essay on forests. Do not rewrite it. Tell me which paragraph is weakest and ask me questions that would help me improve it."
The first can replace your writing effort. The second gives you a chance to improve your own work.
You do not need to know complicated prompting techniques to make this change. You only need to be clear about what you want.
Are you trying to finish a task, or are you trying to get better at doing it?
Both are legitimate goals. But they are not the same goal.
Why the struggle matters
It is tempting to think that struggling with a problem is a waste of time. If a machine can remove that struggle, why not let it?
Because some kinds of effort are part of learning.
Researchers in cognitive psychology use the term desirable difficulties for learning conditions that may feel harder at first but can improve later retention and the ability to use what has been learned in other situations.
Trying to remember something before looking it up is one example. Solving a problem before seeing its solution is another. Explaining an idea in your own words is another.
These activities require you to work with what you have learned rather than simply look at the finished answer.
But there is an important qualification. Not every difficulty is useful. A problem that is far beyond your current ability can leave you confused rather than help you learn. The challenge needs to be appropriate, and you need enough guidance to make progress.
The aim is not to make learning unnecessarily difficult. It is to avoid removing the thinking that helps learning happen.
There is also some direct research on what happens when students use AI this way.
In a 2025 study published in the Proceedings of the National Academy of Sciences (PNAS), researchers led by Hamsa Bastani studied nearly a thousand high-school mathematics students in Turkey.
During practice sessions, some students used a GPT-4-based assistant with a standard chat interface. Another group used a specially designed version called GPT Tutor. A third group did not receive AI assistance.
The standard assistant helped students perform better during practice. But when the students later had to complete an exam without AI, those who had used the standard assistant scored about 17% lower than the control group.
The researchers also tested GPT Tutor. It was designed to guide students with hints rather than simply give away answers. It also included teacher-provided solutions, common student mistakes and guidance for responding to them.
Students using this version performed much better during practice. On the unassisted exam, however, their results were statistically indistinguishable from those of students who had not used AI.
In other words, the specially designed tutor avoided the decline seen with the standard assistant. It did not demonstrate that AI tutoring made students better at the subject than learning without AI.
The study took place in one high school and focused on mathematics. It does not prove that every student, subject or AI tool will produce the same results. Nor does it settle the question of long-term learning.
But it gives us an important lesson: doing better while AI is helping you does not necessarily mean you have learned enough to do the work on your own.
The way the tool helps matters.
What AI cannot do as a mentor
I believe an AI assistant can be useful when you are learning. But we should be honest about its limitations.
It can be confidently wrong.
As I explained in What Is a Thinking Machine? and How a Machine Learns from Examples, AI systems can produce fluent answers that are incorrect.
A neatly explained solution is not necessarily a correct solution. Check important answers against your textbook, your teacher, reliable sources and your own reasoning.
It does not know your whole situation.
A good teacher may notice that you are struggling, remember what you found difficult last month and understand how you respond to different explanations.
An AI assistant only has the context and information available to it. It may not know what you have already learned, where your confusion began or whether you are ready for the next step.
You can help by explaining your level, showing your attempt and telling it what you do not understand. But its picture of your learning is still incomplete.
It cannot take responsibility for your learning.
An AI assistant can offer explanations and questions. It cannot make you practise, decide whether you have understood something or take responsibility for your progress.
If you ask it to do all the work, it may do exactly that. The decision to keep thinking remains yours.
It needs to be used with care.
After my talk in Kannur, one of the questions students asked me was about privacy. It was an important question.
Do not casually share passwords, personal identifying details, private documents or information about other people. Be especially careful with school records, photographs and anything confidential. Understand the privacy settings and data practices of the service you use, and follow your school's guidance.
You can discuss an idea or ask for help with a problem without handing over details of your private life.
So treat AI as one learning aid among several. Your teachers, parents, friends, books and your own effort still matter.
A few habits worth building
You do not need a complicated set of rules. A few simple habits can change the way you use AI.
Try before you ask.
Spend a little time on the problem yourself. Write down what you understand and where you are stuck, even if your attempt is wrong.
Then show the attempt to AI and ask it to help you improve it.
Ask for hints, not complete answers.
Try saying:
"Give me one hint at a time. Do not reveal the complete solution unless I ask."
Work with the hint before asking for another.
Ask it to question you.
Try:
"Ask me questions to check whether I really understand this topic. Ask one question at a time."
Do not just read the questions. Try to answer them without looking things up.
Ask why, not only what.
Instead of asking, "What is the formula?", ask, "Why does this formula work, and when should I use it?"
Understanding the reason behind a method makes it easier to recognise when that method is useful.
Explain it back.
Once you think you understand an idea, explain it in your own words. Ask the AI to identify what you have misunderstood or left out.
If you cannot explain an idea simply, you may need to spend more time with it.
Check what it tells you.
Compare important answers with your textbook, notes or teacher. If they disagree, do not simply choose the answer that sounds more convincing. Investigate the difference.
That investigation is part of learning.
Be honest about what you are trying to do.
Sometimes you genuinely need a quick answer. That is fine. AI can be useful for getting work done.
But recognise the difference between using an answer and learning how to produce one.
Do not mistake the first for the second.
A word to teachers and parents
Many teachers and parents worry that AI will make students stop thinking.
It is a reasonable concern. The research discussed above shows that the way students use AI can affect what they learn.
But simply telling students never to use these tools may not prepare them for a world in which AI assistants are increasingly available.
We need to teach them how to use these tools thoughtfully.
A teacher might ask students to submit their first attempt along with the help they received from AI. A mathematics assignment might require students to explain why a method works, not just provide the final answer. A writing exercise might ask students to show how they revised their own draft after receiving feedback.
These approaches make the student's thinking visible.
Teachers can also demonstrate the process themselves. Ask AI a question. Examine its answer. Check it against another source. Point out where it is wrong or incomplete.
Students learn something valuable when they see an adult question a machine rather than accept everything it says.
The aim is not to make AI the teacher. It is to help students use it without giving up their own responsibility to learn.
Try this
The next time you are about to ask AI for an answer, pause for a moment.
Make your own attempt first. It does not have to be correct. It only needs to show what you have understood so far.
Then try this:
"Here is my attempt. Do not give me the answer. Tell me whether I am on the right track, point out any mistake and give me one hint to help me continue."
Work with that hint. If you are still stuck, ask another question.
When you reach the answer, close the conversation and try solving a similar problem on your own.
That last step matters. It helps you find out whether you can use what you have learned without the machine beside you.
In Bring a Rough Idea, I asked you to bring a rough idea. This time, bring your own rough attempt.
Let the machine help you improve it, not replace it.
That is how an answer machine can become a learning companion: one question at a time.