Knowledge · Ira Berg, 12 November 2026
How teachers can make their students' share of thinking visible
In short: A finished text reveals little about how much of the student's own thinking went into it. Oral follow up questions, interim stages and tools that give students feedback themselves show this more reliably than any text analysis afterwards.
Why the finished result reveals little
Two students can hand in the same text: one copied it from an AI answer, the other worked it out with AI support and really understood it. From the outside both texts often look equally good. If you only assess the finished result, you can hardly see this difference, and this is exactly where many teachers get frustrated.
Three ways to make the share of thinking visible
1. Process instead of product
If you ask for interim stages, drafts or a short record of the steps, you see more than in the finished text. A simple question like “What did you try yourself before you asked the AI?” often gets honest answers.
2. Oral follow up questions
A short, unannounced question about their own text quickly shows whether the content was understood. It does not need to feel like an exam. A short chat at the desk is often enough.
3. Self assessment by the students
Tools like Monkey Score give students feedback right while they work with AI on how much they read, questioned and added themselves. This is not control by the teacher but a mirror for the students, and it often leads them to ask better questions on their own. Teachers do not see individual results, only the effect in class discussion.
What the research says
A study with almost 1,000 students in maths classes (Bastani and others, PNAS 2025) shows that using AI while practising can make performance in the following exam without AI worse, unless the AI only gives hints instead of finished solutions. So the decisive difference was not whether AI was used, but how. More on the research
Clear rules as a basis
Visibility works best when everyone knows what is expected. The template AI rules for the classroom uses a simple traffic light model for this.
FAQ
What exactly does share of thinking mean?
The part of a piece of work that is based on the student's own reading, questioning and adding rather than on simply taking over an AI answer. It cannot be measured exactly, but it can be estimated from behaviour such as reading time and follow up questions.
Do you need technical tools for this?
Not necessarily. Oral questions and interim stages also work without software. Tools like Monkey Score can help in addition, because they give students feedback that is otherwise missing in everyday school life.
Prompt to try
Copy this straight into ChatGPT, Gemini or Claude:
“Write a short reflection question that students answer in writing after an AI supported task, so they can honestly assess their own share of thinking.”