In the results released on 8 September 2026, PISA 2025 reported a new assessment of computational problem solving. The aim is not to check whether students have memorised a programming language. The focus is on whether they can learn and solve problems in a digital environment using computational tools, test hypotheses, interpret data, build executable representations and correct solutions when they do not work.
According to the OECD's PISA 2025 Results Volume I, the scale measures students' capacity to make progress towards a goal and build functional solutions with tools such as modelling and programming. The PISA 2025 assessment framework organises the competency into three official practices: conducting experiments, analysing data and building and debugging computational artefacts. Self-regulation appears as a supporting component: maintaining engagement, monitoring progress, adapting strategies and evaluating one's own performance.
What “computational problem solving” means in PISA
In PISA 2025, solving a problem computationally means using a digital environment to represent a system, act on it and learn from what happens. A student may manipulate a simulation, assemble a model, organise a flow or combine programming blocks. The key is to turn a question into actions that can be executed, observed and revised.
This brings the assessment closer to situations in which a solution is not captured by a single multiple-choice response. Students need to make progress through a sequence: understand the goal, experiment, observe results, adjust a representation and decide whether the solution is adequate. That is why the OECD describes the process as iterative and self-regulated.
The three practices the test seeks to measure
| Official practice | What the student does | What evidence a task can produce |
|---|---|---|
| Conduct experiments | Uses models, simulations or other tools to generate data; formulates hypotheses and tests variables in a controlled way. | Which parameters were changed, in what order, how many attempts were made and whether the test isolated the relevant variable. |
| Analyse data | Interprets data to infer how parts of a system are related and to predict changes. | Reading graphs and tables, comparing results, identifying patterns and choosing a conclusion supported by the data. |
| Build and debug computational artefacts | Decomposes the problem, creates an executable representation and adjusts algorithms or models. | Blocks, connections and parameters added; use of loops or conditions; corrections after failures; quality of the final model or program. |
1. Experiment and analyse: learn from how the system behaves
Author-created example: imagine a simulation of water use in a school. The student can separately change the number of students, tap flow rate and time of use. A strong strategy is not to change everything at once, but to form a hypothesis — for example, “will reducing usage time have a larger effect than reducing flow rate?” — test comparable scenarios and then read the resulting data.
Two connected practices appear in this kind of task. Conducting experiments involves deciding what to vary and what to control; analysing data involves using the simulation output to support a conclusion or prediction. The assessment can record not only the final answer but also actions during the process, such as experiments run, changes to the artefact and requests for feedback.
2. Model: turn relationships into an executable representation
Author-created example: suppose students receive information about a small transport network: the number of buses affects waiting time, but more buses also increase cost. They need to assemble a model that represents those relationships, run scenarios and check whether the output makes sense given the supplied data.
“Modelling” is central to the experience, but there is an important nuance: modelling is not listed as a separate fourth official practice. In the OECD framework, models and modelling tools are means through which students conduct experiments, analyse data and build or debug an artefact. Tools envisaged for the assessment include block-based programming, executable concept maps, flowcharts and simulations.
3. Build and debug: make the solution work and fix it
Author-created example: in a block-based programming environment, a student needs to create a rule for a robot to collect objects on a grid. The first version works in some cases but fails when it encounters an obstacle. The student must identify the cause, change a condition, test again and remove redundant steps.
The OECD links this practice to decomposition and abstraction, identifying patterns, using control structures such as loops and conditionals, creating representations that a computer can execute, and adapting or debugging algorithms and models. Getting it right on the first attempt is therefore only one possibility; detecting and correcting an error is also evidence of the skill.
What task format appears in the assessment
Each unit in the digital domain is structured as a small learning module. The final framework describes 30-minute units organised in a common progression:
- Intro + Show: presentation of the context and goal, followed by four or five short items on relevant prior knowledge;
- Learn: a guided tutorial followed by three or four interactive tasks, with worked examples and feedback;
- Reflect: brief self-report questions about confidence, difficulty, interest and self-evaluation;
- Apply: one or two more open and complex tasks in which students apply the practices they learned;
- Final Reflect: another reflection on performance, effort and affective state.
PISA's rotated design means that units can use different tools. The framework cites, for example, a simulation-based environment in one unit and a block-based programming environment in another. The scale therefore does not depend on one piece of software or a single way of programming.
An official example shows how the logic works
In material released by the OECD, the “Conservation” unit uses an ecosystem as its context. In simpler stages, students complete a model, connect variables and compare model output with observed data. In an application stage, they need to build a model of another ecosystem, run it to generate data and check whether those data adequately reproduce real behaviour.
In another subtask, students add a variable related to the fishing rate, test different values and identify the highest value compatible with coral survival. The same activity brings together all three practices: building/debugging the model, experimenting with different rates and analysing the resulting data. The OECD's released-item material illustrates how science, data and digital tools can be integrated in one task.
What role does self-regulation play?
The assessment also observes how students manage their own process. The competency model includes three self-regulation dimensions: maintaining motivation and task engagement, monitoring progress and adapting strategies, and evaluating performance. The task provides resources such as tutorials, examples, feedback and annotated solutions; using them appropriately can reveal how students respond to difficulty and adjust their approach.
It is important not to mistake this for “extra points for behaviour”. In the assessment framework, specific self-regulation indicators are used to contextualise performance on the computational problem-solving scale and do not directly contribute to task scores. The OECD combines self-reports with response-process data to study these patterns.
Does the test measure programming?
It measures programming as one possible tool, but it is not limited to programming. Students may work with blocks, models, flowcharts or simulations. What matters is whether they can build an executable solution, produce evidence through experimentation, interpret results and revise their strategy.
This also explains why the domain is not simply an extension of mathematics or science. PISA 2025 results show strong correlations with those areas, but the OECD says the assessment was designed to capture a related yet non-reducible set of competencies. A distinctive feature is that students can learn from resources and feedback during the assessment itself.
How to recognise a typical computational problem-solving task
A task in this domain usually asks for more than selecting a ready-made response. There is a system to explore, a representation to build or modify and a goal that requires cycles of testing and revision. To identify the skill being assessed, ask:
- does the student need to generate data by changing variables or running a simulation?
- do they need to interpret those data to reach a conclusion or make a prediction?
- do they need to build, complete or correct a model, algorithm or flow?
- is there an opportunity to use feedback, revise the strategy and try again?
That combination captures the value of the new domain: PISA 2025 does not only look at whether students know digital concepts, but whether they can use computational tools to learn, test ideas and produce a functional solution. For methodology, the main reference is the PISA 2025 Assessment and Analytical Framework; for the results released on 8 September 2026, see PISA 2025 Results Volume I and the OECD's PISA frequently asked questions.
