Beyond Science Scores: What PISA 2025 Tells Us About STEAM Education in Estonia
From scientific inquiry to computational problem solving — why Estonia’s results point towards a deeper understanding of STEAM
By Mehmet ALTAYFounder & International Projects DirectorTeachers Education Academy
When education systems talk about STEAM, the conversation often begins with activities.
Robotics. Coding. 3D printing. Experiments. Makerspaces. Engineering challenges. Digital tools.
All of these can contribute to excellent learning.
But PISA 2025 invites us to ask a more fundamental question:
What should students actually become capable of doing through STEAM education?
Estonia provides an especially interesting case.
In PISA 2025, Estonia achieved 527 points in science, placing it among the world's highest-performing education systems. Within the European Union, Estonia ranks first in science. It also achieved 541 points in computational problem solving, again placing it first among EU Member States and among the strongest OECD education systems internationally.
But the most interesting story is not the ranking.
It is what Estonian students appear particularly capable of doing.
And that has important implications for how we understand STEAM education.
First, an important distinction: PISA does not measure “STEAM performance”
We should be careful with terminology.
PISA 2025 assesses science, mathematics, reading and computational problem solving. It does not provide a national STEAM score, nor does it evaluate whether a particular country's STEAM strategy has “succeeded”.
Therefore, Estonia's PISA results should not be presented as proof that STEAM education caused its performance.
That would go beyond the evidence.
What PISA does provide, however, is something highly relevant.
The competencies it assesses—particularly scientific inquiry, evidence evaluation, problem solving, experimentation, data analysis, computational thinking and self-regulated learning—overlap strongly with the capabilities that high-quality STEAM education seeks to develop. PISA's new Learning in the Digital World assessment explicitly combines computational and scientific inquiry practices with students' capacity to regulate their own learning.
This makes Estonia's results particularly interesting from a STEAM perspective.
Estonia's science strength is not simply knowing more science
Estonia scored 527 in science, compared with an OECD average of 482.
But looking inside the science score reveals something more interesting.
Estonia's results across the three science competency areas were:
Science competency | Estonia |
Explaining phenomena scientifically | 521 |
Evaluating designs for scientific enquiry | 530 |
Evaluating scientific information for decision making | 528 |
Estonia's strongest relative area is therefore scientific enquiry. The OECD specifically identifies evaluating designs for scientific enquiry as a relative strength for Estonia.
That matters enormously for STEAM.
Because there is a fundamental difference between teaching students to know science and teaching students to think scientifically.
A student can memorise a scientific principle.
A different level of learning occurs when the student can ask a question, formulate an investigation, determine what evidence is needed, conduct or evaluate an experiment, analyse results, recognise weaknesses and defend a conclusion.
That second form of learning is much closer to what meaningful STEAM education should be trying to achieve.
Estonia's PISA result therefore suggests an important principle:
The goal should not simply be more science knowledge. It should be greater capacity for scientific inquiry.
From “What do you know?” to “What can you investigate?”
This distinction reflects a broader change within PISA itself.
The PISA 2025 science framework places considerable emphasis on students' ability to engage with scientific information and evidence rather than simply reproduce established knowledge.
Students are expected to evaluate scientific investigations, interpret evidence critically and research and use scientific information when making decisions.
This is remarkably relevant to contemporary STEAM pedagogy.
Consider two science activities.
In the first, a teacher demonstrates an experiment. Students follow instructions, record the expected result and complete a worksheet.
In the second, students encounter a problem.
They must determine:
What do we need to know?
What variables matter?
How could we test our idea?
What data should we collect?
Does our evidence support our hypothesis?
What went wrong?
What should we change?
Both lessons may involve experiments.
But only one truly develops inquiry.
This is why the number of experiments, robots, tablets or STEAM projects in a school tells us very little about the quality of STEAM learning.
The deeper question is:
Who is doing the thinking?
Computational problem solving strengthens the Estonia story
Estonia's second remarkable result comes from PISA 2025's new computational problem-solving assessment.
Estonian students achieved a mean score of:
541
The OECD average is 500.
Estonia is among the strongest OECD education systems internationally in this new domain.
This assessment is especially relevant to STEAM because it is not simply a test of computer skills.
PISA describes computational problem solving as students' capacity to engage in an iterative and self-regulated process of knowledge building and problem solving using computational tools and practices.
Students may need to use modelling or programming tools, conduct experiments, analyse data and build or debug computational artefacts. Their success also depends on whether they can monitor progress, adapt strategies, maintain motivation and continue working when a task becomes difficult.
That definition deserves attention.
Because it moves digital education well beyond:
Can the student use technology?
towards:
Can the student use technology to investigate and solve something they do not already know how to solve?
That is a much higher educational ambition.
Science and computational thinking should not live in separate worlds
Perhaps the strongest STEAM message emerging from PISA 2025 is the relationship between scientific inquiry and computational problem solving.
The framework behind Learning in the Digital World explicitly recognises synergies between inquiry-based learning and computational thinking. Its computational practices include experimentation, data generation and analysis, modelling, problem decomposition and building or debugging computational artefacts.
This is exactly where STEAM can become more than an acronym.
Imagine students investigating energy consumption in their school.
Science helps them understand energy.
Mathematics helps them analyse consumption data.
Technology allows them to collect and visualise information.
Computational thinking helps them model patterns.
Engineering allows them to design possible solutions.
Design and creativity help them communicate or prototype those solutions.
Now the disciplines have a reason to interact.
They are not simply five subjects placed beside each other.
They become different ways of understanding and solving the same problem.
And this leads to one of the most important lessons we can draw from Estonia's PISA performance:
Strong STEAM education is not primarily about combining subjects. It is about combining ways of thinking.
Less technology activity. More problem solving with technology.
This distinction becomes particularly important as schools invest increasingly in educational technology.
A classroom containing robots is not automatically a STEAM classroom.
Neither is a classroom using tablets.
Neither is a coding lesson.
Neither is a 3D printer.
Technology can certainly create extraordinary learning opportunities. OECD's Learning in the Digital World framework specifically recognises the potential of technology to allow students to explore complex phenomena and create digital representations of their ideas.
But technology is a means.
The learning objective remains the development of human capability.
A robotics activity in which every student follows identical instructions to construct an identical robot may develop certain technical skills.
A different challenge might ask students to design a robot capable of solving an unfamiliar problem under defined constraints.
Now students must:
analyse → design → test → fail → debug → redesign → explain.
The technology may be identical.
The learning is not.
So perhaps one of the clearest messages for STEAM education after PISA 2025 is:
Less “technology activity”. More “problem-solving activity using technology”.
Failure should be part of STEAM learning
There is another important dimension hidden inside computational problem solving.
PISA's assessment does not assume students immediately know how to solve every task.
They encounter increasingly complex problems, receive resources and feedback, monitor their progress and adapt their strategies. Self-regulated learning is therefore part of successful performance.
This has an important classroom implication.
In weak STEAM activities, success sometimes means reproducing the teacher's intended product.
Everyone's bridge looks similar.
Everyone's robot follows the same route.
Everyone's experiment reaches the expected conclusion.
But authentic problem solving contains uncertainty.
A bridge collapses.
A model produces unexpected results.
A program does not work.
An experiment contradicts the hypothesis.
A design solves one problem but creates another.
These are not necessarily failures of learning.
They can be the moments when the deepest learning begins.
High-quality STEAM should therefore create safe opportunities for students to:
test → fail → analyse → modify → test again.
The final product matters.
But the reasoning that produced it matters more.
Estonia also shows that strong science should not be reserved for an elite
Another important dimension of Estonia's PISA performance is how broadly science competence is distributed.
Estonia is not simply producing a small group of exceptional science students while leaving large numbers behind.
PISA places Estonia among the small group of education systems internationally with very low shares of students below baseline science proficiency, while the country simultaneously maintains high overall performance.
This is highly relevant to STEAM policy.
STEAM should not become an enrichment programme only for students already interested in science, robotics or technology.
Nor should it become something reserved for gifted students or specialised schools.
The larger educational ambition should be:
How can scientific thinking and problem solving become accessible to every learner?
That changes the design of STEAM education.
It moves us away from identifying a small number of future engineers and scientists and towards developing a scientifically and technologically capable generation.
Not every child will become an engineer.
But every child will live in a world increasingly shaped by science, technology, data and artificial intelligence.
Scientific reasoning is therefore becoming part of citizenship.
What about the “A” in STEAM?
This requires particular caution.
PISA 2025 does not directly assess arts integration.
We therefore cannot use Estonia's PISA results to claim that adding Arts to STEM caused stronger academic performance.
But the “A” in STEAM can still play an important pedagogical role—provided we understand it properly.
Arts should not simply mean decorating the final product.
Its deeper contribution can involve:
creativity, design, communication, visualisation, perspective, imagination and multiple possible solutions.
An engineering problem rarely has only one acceptable design.
A scientific discovery still needs to be communicated.
Complex data often need visual representation.
An innovative product must respond to human needs.
Students need imagination to move from understanding what exists to proposing what could exist.
Those capabilities are compatible with the open-ended, iterative problem solving that PISA increasingly values.
But this connection should be understood as a pedagogical interpretation, not as a direct PISA finding.
That distinction matters.
STEAM in the age of AI
There is another reason these findings matter now.
Artificial intelligence can increasingly provide students with answers.
It can write code.
It can generate explanations.
It can analyse data.
It can suggest experimental designs.
It can even propose engineering solutions.
If STEAM education is primarily about producing an answer or final product, AI will increasingly be able to perform much of that work.
This makes process more important than ever.
The educational value moves towards:
asking the right question;
designing the investigation;
deciding which evidence matters;
testing assumptions;
evaluating generated information;
identifying errors;
explaining decisions;
and
defending a solution.
This is consistent with the broader message emerging from PISA 2025: technology should deepen learning rather than provide shortcuts around the cognitive effort required to learn.
The AI age therefore does not make STEAM less important.
It makes deep STEAM more important.
From STEAM activities to STEAM pedagogy
Perhaps the biggest mistake schools can make is to define STEAM through equipment.
A school buys robotics kits.
It establishes a makerspace.
It purchases 3D printers.
It introduces coding.
It organises a STEAM Week.
All of these may be valuable.
But none automatically creates STEAM pedagogy.
A deeper model would begin with learning behaviours.
Can students:
ask investigable questions?
develop hypotheses?
design tests?
analyse evidence?
use mathematics to interpret results?
use digital tools to model a problem?
design and test solutions?
recognise failure and modify their approach?
communicate their reasoning?
work independently when the solution is not obvious?
Those questions describe student capability.
And student capability—not equipment—is ultimately what matters.
What can we learn from Estonia?
The lesson should not be:
“Copy Estonia's STEAM model.”
PISA does not provide evidence for such a conclusion, and education systems cannot simply be transplanted from one national context to another.
A more useful interpretation is this:
Estonia demonstrates what is possible when students achieve high levels of scientific reasoning and computational problem solving at the same time.
That combination should interest every school trying to develop meaningful STEAM education.
The direction is clear.
Move from:
following instructions → designing investigations
memorising content → applying knowledge
using devices → solving problems with technology
producing products → explaining decisions
avoiding failure → learning through iteration
separate subjects → connected ways of thinking
STEAM activities → STEAM pedagogy
That is a much deeper transformation than simply introducing more technology into classrooms.
Beyond the acronym
STEAM has become one of the most widely used terms in contemporary education.
But there is a danger whenever an educational idea becomes popular.
The label can become more visible than the learning.
PISA 2025 offers a useful reminder of what should remain underneath that label.
Students need strong foundational knowledge.
But they also need to know what to do with that knowledge.
They need to investigate.
Experiment.
Model.
Analyse.
Evaluate.
Design.
Adapt.
Create.
And solve problems whose answers are not immediately obvious.
Estonia's results do not prove that STEAM produced its educational success.
They tell us something more useful.
They show that one of Europe's strongest education systems is exceptionally successful in developing precisely the kinds of scientific inquiry and computational problem-solving capabilities that meaningful STEAM education should seek to cultivate.
That should shift the STEAM conversation.
Away from:
How many STEAM activities do we provide?
Towards:
What kind of thinkers are those activities helping our students become?
Because the real promise of STEAM is not that every student learns to build a robot.
It is that students learn to approach an unfamiliar problem and say:
“I don't know the answer yet — but I know how to investigate it.”
That may be one of the most important capabilities education can develop for the future.
Mehmet ALTAYFounder & International Projects DirectorTeachers Education Academy
This independent analysis was prepared by Teachers Education Academy using OECD PISA 2025 findings and the PISA 2025 assessment frameworks. PISA does not directly assess STEAM education or establish a causal relationship between STEAM provision and Estonia's performance. References to STEAM in this article are pedagogical interpretations of the scientific inquiry and computational problem-solving competencies assessed by PISA. Interpretations and conclusions are those of Teachers Education Academy and should not be understood as official OECD or Estonian government positions.
Comments