Does AI Really Help Students Learn? A Major Experiment in Polish Schools
A teacher launches a new tool. An interactive simulation appears on the screen, students look up from their desks, and the room comes alive. The lesson suddenly has energy. From the outside, it looks like a success.
But what will the students remember a week later?
That question is much harder than asking whether an app is attractive, how many times students clicked, or whether they say they enjoyed using it. In education, the wow effect comes quickly. Evidence that technology genuinely improves learning requires patience, a sound method, and a comparison with what would have happened without it.
This is the test being prepared by the Educational Research Institute in Poland. The Formula for STEM competition will select eight digital tools for a large-scale study in Polish schools. In an EdTech Dots conversation, Michał Klepka from the Institute describes a plan involving 720 classes, eight experiments, and one fundamental criterion: measurable growth in knowledge.
Eight tools and 720 classrooms
Formula for STEM is not a contest for the most futuristic presentation. The Institute is looking for solutions that already exist, work in practice, and can enter a real school, complete with its timetable, available devices, internet connection, and teachers who do not have several spare hours for every new piece of technology.
The selection committee will choose eight digital tools. According to the plan outlined in the interview, each will become the subject of a separate experiment, with the full study covering 720 classes across Poland.
That scale matters. A tool that succeeds in a small pilot run by its creators or a group of enthusiasts may not perform equally well in daily practice. Polish schools also differ from schools in the United Kingdom, the Netherlands, Singapore, or the Czech Republic. Their organization, infrastructure, curriculum, and working culture are different. If the findings are to support serious decisions, the solution has to be tested here.
For an EdTech company, this is a rare opportunity. Instead of another case study built around a handful of opinions, the company may gain evidence showing whether its product genuinely supports learning.
The tool, the teaching method, and an ordinary lesson
The most interesting part of the project begins when the technology is separated from the way the lesson is taught.
Imagine a series of biology lessons on the circulatory system. In one class, the teacher uses project-based learning together with a digital simulation of blood flow. In another, the teacher follows the same teaching method but works with textbooks, illustrations, and other non-digital materials. In a third class, the topic is taught in the usual way, without the tested method and without the tool.
This setup makes it possible to answer three different questions:
Did students gain more from the combination of the method and the technology?
Did the teaching method alone produce a similar result?
Was either result different from that of an ordinary lesson?
Without this comparison, it is easy to credit an app for an effect that actually came from a well-designed lesson plan. The opposite mistake is also possible: a good tool can be rejected because it was paired with weak teaching practice.
That is why the competition evaluates more than the product. Applicants must also prepare three consecutive lesson plans aligned with the curriculum. The technology has to become part of a specific learning process, not hover above it as an impressive add-on.
The real test starts after the lesson
The first measurement will establish what students know before the intervention. The next will show what changed after the lessons. A later measurement is meant to answer an even more important question: how much of that knowledge remained over time?
Retention is what separates momentary engagement from learning.
An app may make students more willing to participate. It may also help them complete a task faster. Both outcomes can be valuable, but neither proves that students understand the topic better or can return to that knowledge later.
Michał Klepka expresses the success criterion plainly: the researchers need to see growth in students' knowledge. The presence of technology is not the result. The result is a change that can be measured credibly.
Teachers' time is also at stake
When a company introduces a new tool into a school, it asks much more of a teacher than simply logging into another application.
The teacher has to learn the interface, adapt materials, revise lesson plans, anticipate technical problems, and understand how the product behaves in a classroom. It may take several attempts before the tool becomes a natural part of the work. If it ultimately proves ineffective, the cost is not limited to the licence fee. It also includes the teacher's time and energy.
This is one of the strongest arguments for rigorous EdTech evaluation. Schools cannot embrace every novelty simply because it looks modern. At the same time, they should not close themselves off from change. They need reliable evidence about which solutions deserve the effort.
The study is not designed to grade teachers. It is designed to measure changes in students' knowledge. Teachers participating in the experiment are expected to receive preparation both for using the tool and for teaching the relevant part of the curriculum. This matters because even an excellent product cannot show its value in a study if the person using it has not been shown how to make it part of a lesson.
What kind of tool has a chance?
The Institute is not looking for an idea captured on a slide or a prototype whose development will only begin after the competition. It is seeking mature solutions that can operate in primary or secondary schools and are available in Polish.
The eligible subjects include:
mathematics,
physics,
chemistry,
biology,
geography,
computer science,
general science.
A strong candidate should have a clear educational purpose, a credible team, the rights required to provide the solution, and the ability to support the research. The application asks about previous work with schools, existing evidence of effectiveness, consultation with teachers, digital accessibility, licensing, support for larger numbers of users, and the ability to provide structured data about student activity.
This is not a competition for every product. But if your tool is already in use, addresses a specific challenge in STEM education, and your team wants an honest test of its effectiveness, it is difficult to imagine a more interesting opportunity.
What about AI?
Artificial intelligence now appears in almost every conversation about the future of education. Yet the label alone still tells us nothing about whether students learn more effectively.
A tool can generate exercises, suggest the next step, personalize content, or support a teacher. Each of these functions may be valuable. Each may also make a task faster without deepening understanding.
AI-based solutions therefore receive no special exemption. Applicants must explain how they ensure safety, transparency, human oversight, and data protection. Then comes the decisive test: when paired with a specific teaching method, does the tool improve learning outcomes?
The point is not to prove that AI is good or bad. The point is to identify the conditions in which it becomes useful.
What does an EdTech company gain?
The greatest value is not the title of competition winner. It is the opportunity to join a large-scale study conducted in real schools and planned for the 2027/2028 school year.
For product teams, that creates an opportunity to:
see how the tool performs outside a controlled pilot,
obtain data about learning outcomes,
separate the effect of the technology from the effect of the teaching method,
discover constraints that remain invisible in a product demonstration,
build stronger foundations for conversations with schools and public institutions.
An honest result may confirm the product team's assumptions. It may also show that the solution needs to change. Either way, the company receives something more useful than praise: information on which it can base its next decision.
How to apply to Formula for STEM
Applications are open until 21 September 2026 at 11:59 p.m. Participation is free and voluntary.
A complete submission includes:
the online application form,
a video showing how the digital tool works,
three consecutive lesson plans prepared according to the organizer's guidelines.
The documentation is detailed because selected solutions are intended for serious research, not a showcase pilot. Teams should begin by reviewing the technical, methodological, and formal requirements before preparing their materials.
More information is available on the competition page and the project page.
Evidence first, excitement second
Polish schools do not need another application that looks impressive on a conference screen but falls apart when it meets an ordinary lesson. They need solutions that can make it through a normal school day and still deliver value to students.
Formula for STEM asks EdTech teams a demanding question: are you ready not only to demonstrate your product, but also to let it be tested rigorously?
If the answer is yes, applications remain open until 21 September. Your tool may become part of a substantial experiment into the role of educational technology in Polish schools.
The conversation with Michał Klepka is hosted by Krzysztof Kosman.
This episode is sponsored by 1000ideas. We help companies turn ideas into working digital products, including applications, educational platforms, web and mobile systems, AI tools, and business automation. Learn more about 1000ideas.
