The question that actually decides it
Most advice about AI and study starts with rules. Allowed, not allowed, disclose it, don't. That is a real question, and we cover it later. But it is the wrong *first* question, because the answer tells you nothing about what happened to you.
Here is a better one.
After the session, what changed — the document, or you?
Both kinds of help produce a finished assignment. Only one produces a person who could do it again.
The same tool, two directions
Same subject. Same chatbot. Same ten minutes.
| Avoiding the work | Doing the work | |---|---| | "Solve question 4." | "I got x = 3, the answer key says x = −1. Here is my working. Which line is wrong? Don't tell me why yet." | | "Write an introduction on the causes of the Nigerian Civil War." | "Here is my introduction. Argue against my thesis as harshly as an examiner would." | | "Explain recursion." | "Explain recursion, then give me three problems and withhold the solutions." | | "Summarise chapter 6." | "I've read chapter 6. Ask me six questions on it, one at a time, and don't accept a vague answer." |
The right column is not slower because it is more virtuous. It is slower because you are the one doing the expensive part, and the expensive part is the point.
The load-bearing part
Every assignment has one thing it exists to build. Find it, and you know what not to hand over.
- In a problem set, the load-bearing part is the solving. Not the arithmetic, not the neat presentation.
- In an essay, it is the argument — which claims you make, in what order, and what you leave out.
- In a lab report, it is the interpretation. Not the method section you have written eleven times.
- In a translation exercise, it is choosing between two defensible renderings.
Everything else is scaffolding, and scaffolding is fair game. Reformatting a bibliography into APA is not a skill anybody is examining. Turning your bullet points into a table is not either. If you spend the saved hour on the load-bearing part, the tool has done its job.
The transfer test
Here is a check you can run today, and it costs ten minutes.
Finish a piece of work with AI help. Close the laptop. Wait ten minutes — get water, walk somewhere. Then take blank paper and redo the central thing: solve the problem again, or write your thesis and its three supports from memory.
Whatever comes back is yours. Whatever does not, you watched somebody else do.
This is uncomfortable the first time, because the gap is usually much bigger than it felt. Reading a clear explanation produces a strong, immediate feeling of understanding. That feeling is real, and it is not knowledge. Cognitive psychologists call it the fluency illusion, and AI is very good at generating it, because AI is very good at being clear.
Neither extreme survives contact with reality
Refusing to touch these tools is not integrity, it is just a tax you pay in hours. A student in Lagos or Lima or Lucknow with no tutor, no office hours, and a class of 300 has something now that did not exist before: a patient thing that will explain the same idea six different ways at two in the morning. Telling that student to switch it off is not advice, it is a lecture.
And going the other way — running every task through the model and shipping the output — reliably produces someone who cannot tell when the output is wrong. That is not a moral failure. It is a mechanical one, and we will look at the evidence in the next lesson.
The useful position is in the middle and it is specific: delegate the scaffolding, keep the load-bearing part, and test yourself on paper afterwards.
One habit to start with
Before you type anything into a chat box, finish this sentence out loud:
"I am asking for this because I want to be able to _____ afterwards."
If the honest ending is "submit it," you already know what kind of help this is. That is not always the wrong answer — some weeks are like that. But say it plainly to yourself, because the cost is real and it arrives later, in a room where the tool is not allowed.
Before you move on