In the spring of 2024, when GoSchool was still just an experiment, we organized a workshop for university educators. Among other things, we asked them to consider: which educational practices might be strengthened by artificial intelligence, and which might lose their meaning. Most red sticky notes had the label “multiple-choice questions “”, closely followed by “essay writing”.”
The case reminded me of a field experiment by a Boston research group, which they published in December 2025. The research sought to answer how the use of AI affects essay writing – the cognitive processes and brain activity of the participants, as well as the quality of the texts produced.
The research showed that students using AI exerted less cognitive effort, showed reduced brain activity in areas responsible for deep processing, and the essays they wrote proved to be less substantial, personal, or memorable.
Interestingly, while teachers judged these essays to be weaker, an automatic text evaluation AI rated them as better – purely based on formal characteristics.
What Follows from this? Essay writing will become unsuitable for measuring school performance in the near future. Students will complete essays without studying, and teachers will evaluate them with machines. Everyone benefits, and the system appears to work. The measured results don’t even worsen, in fact – with one small exception: we are measuring the performance of language models, not the students’.
This state can only be maintained as long as the student remains within the school walls. As soon as they enter the job market – and truly need to apply their knowledge – the lack of invested effort quickly becomes apparent.
Essay writing – beyond some specialized training – was never a goal, but rather a tool: a substitute method to assess what a student is capable of with the knowledge they have acquired. However, since language models are capable of “acquiring” and articulating knowledge, this function is lost, and a fundamental tool for assessment slips from our hands.
In parallel with all this, AI makes every participant more efficient. Both students and teachers outsource demanding cognitive work – just as a lever reduces muscle work. This is not laziness, but human nature: where possible, we save energy.
The Key is how AI is Applied
One of the most important lessons from the Boston research was the result of the “brain-to-LLM “” group. In this experimental condition, students first formulated their thoughts independently, and only then involved the language model in essay writing. The result: higher cognitive activity, more substantial and personal texts – not only compared to the “LLM-to-brain”” group, but also compared to students who wrote entirely without AI.
For us, this is not only a methodological but also a pedagogical guideline: the problem is not AI, but the timing of outsourcing thinking. One of GoSchool’s core principles is that the student should not start from the machine, but from their own question, hypothesis, or viewpoint – and receive feedback on this – in real time, without further burdening the teacher.
When founding GoSchool, we started from the premise of accepting the rise of AI. We are not naive techno-optimists, but we personally experience how generative AI scales our information consumption capabilities and makes us more efficient in our work.
That’s why we decided to experiment with bringing AI into schools – not as a solution, but as a tool – and observe, gather feedback, and adapt. Always keeping in mind: the goal is not the technology itself, but to strengthen education. If AI boosts performance in the economy, it must also do so in schools – but to achieve this, we must also raise the bar.
If the student is more efficient, then we must measure their knowledge differently. We need tasks that, even with the use of AI, compel the same brain activity that was previously required for essay writing. If an athlete uses a lever, they must be made to lift heavier weights with it.
GoSchool was created because we believe that education cannot develop without embracing artificial intelligence. Teachers and students together, because, in our opinion, education is their collaboration.
Not long after the aforementioned workshop, I ran into my high school homeroom teacher at an alumni meeting. Piarist monks dedicate their lives to learning, education, and upbringing. My homeroom teacher at the time spoke seven (or nine?) languages, taught English, biology, chemistry, and could step in for math at any time. In addition, he took us sailing, skiing, hiking, rowing, and cycling.
When I told him about the project, I remarked: I don’t dare tell the teachers that students are using the system instead of reading the book – and they say they understand the material faster this way. He laughed: “Today’s teachers were not socialized in scholasticism. Back then, discussing knowledge was the value; the book was just a tool.”
Discussion can still be a good tool for active learning today. If a student has to defend what they have learned, AI can help them delve deeper and more comprehensively into a topic than if they were to skim through only a fraction of the available literature. They can ask questions, seek alternative viewpoints, clarify concepts, and explore related topics – and those who are better at this will perform better in discussions. And others will learn from them too.
In the world of artificial intelligence, we must rethink how we assess students’ acquired knowledge. Which past methods must we abandon – and what new forms are emerging in their place?
It is likely that not only measurement, but the entire methodology of education will transform. There will be tools that lose their meaning, and there will be those that gain new meaning. Discussion, more complex tasks, or even consciously integrated LLMs (as the “brain-to-LLM” model shows) are all promising directions.
This is a rapidly changing, still evolving field – and many questions await answers. At GoSchool, we are working to find answers to these questions together.
If you are interested in collaborative thinking, join us.