How Problem Posing Works | Learn by Building the Question Before Solving It

eduKateSG Learning Node Series · 0050

Solving a question asks whether you can navigate a problem someone else designed. Posing a question asks whether you understand the machinery well enough to design the problem yourself.

Problem posing is the deliberate construction, reformulation or extension of a problem. The learner must decide what information matters, which relationships are possible, what can be asked, what constraints make the question meaningful and whether the resulting problem is actually solvable.

That makes problem posing more than a creativity exercise. Done carefully, it becomes a diagnostic window into conceptual structure.

Quick Read: When the Learner Becomes the Problem Designer

Problem solving begins with a problem and searches for a solution.
Problem posing constructs or transforms the problem itself.
A learner can pose by creating a new problem, changing a condition, reversing a known question, completing an incomplete situation or asking a new question from the same data.
Mathematics education has the strongest research tradition here. Recent meta-analyses report positive average effects of problem-posing interventions on mathematical cognitive outcomes and problem-posing competence.
The method is not automatically productive. Learners can generate trivial, impossible, ambiguous or mathematically empty questions.
Good problem posing therefore needs constraints, examples, feedback and quality criteria.
The goal is not to replace solving. Posing and solving should form a loop: construct → test → solve → inspect → revise.

If solving reveals whether you can use a structure, posing can reveal whether you understand what makes that structure possible.

The Ready-Made Question Problem

School usually presents questions as finished objects.

The numbers have already been chosen. The diagram has already been drawn. The unnecessary information has already been removed. The target is already stated. The problem setter has quietly performed a large amount of intellectual work before the student sees the page.

This can hide something important: a well-formed question is itself a model of the domain.

To create a good ratio problem, you must know which quantities can legitimately be compared. To create a science investigation question, you must know which variable can change and what outcome can be measured. To create a comprehension question, you must know what the text actually supports.

Problem posing returns some of that design work to the learner.

What Counts as Problem Posing?

Problem posing can happen before, during or after solving. It includes several families of task.

Free posing: create a problem from a broad topic or situation.
Semi-structured posing: create a problem from a diagram, table, story, equation or set of objects.
Structured posing: modify a given problem under explicit constraints.
Reformulation: restate an existing problem in a different but equivalent form.
Extension: change one condition and ask what new problem appears.
Reversal: give what was previously unknown and make something else the target.

A Simple Example: The Question Behind 24 ÷ 6

A learner can calculate 24 ÷ 6 = 4 without understanding all the situations division can represent.

Ask the learner to create three different word problems whose calculation is 24 ÷ 6.

Twenty-four objects shared equally among six people.
Twenty-four metres divided into lengths of six metres.
A quantity 24 that is six times another quantity.

Now the arithmetic stays fixed while the semantic structure changes. The posing task exposes whether the learner owns division as a relationship rather than only as a button sequence.

Why Posing Can Deepen the Model

To pose a valid problem, the learner must coordinate several things that ordinary solving can sometimes leave implicit:

which entities exist;
which quantities or properties matter;
which relationships connect them;
which information is sufficient;
which information is redundant;
which unknown can be determined;
which assumptions are legitimate;
what answer form would count as complete.

This is one reason problem posing can function as a knowledge audit. Weak conceptual links become visible during construction.

What the Recent Meta-Analyses Say

A 2025 meta-analysis in the Journal for Research in Mathematics Education synthesised 26 quantitative studies published from 2002 to 2024. It reported a positive, statistically significant average effect of problem-posing interventions on cognitive mathematics outcomes, with Hedges’ g around 0.53. Effects were larger in the analysed studies when technology-supported environments were used and when learners posed problems with the support of problem examples.

A separate meta-analysis focused on problem-posing competence also reviewed 26 intervention studies and found a medium positive overall effect, while showing meaningful variation across intervention features.

Earlier meta-analytic work likewise reported positive average effects on mathematics achievement and related outcomes.

The useful conclusion is not “problem posing always works.” It is that there is now a substantial research base showing that well-designed posing interventions can support mathematics learning, while design features and context matter.

Singapore: Problem Posing Is Not Foreign to the Curriculum Conversation

Recent Singapore-focused scholarship has examined how problem posing sits inside the mathematics curriculum and classroom. That matters because posing should not be treated as an exotic enrichment task disconnected from problem solving. It can be used to deepen the same mathematical relationships learners are already expected to understand.

The practical question is where posing adds diagnostic or generative value without crowding out the foundational instruction and practice students still need.

Problem Posing Versus Problem Solving

Problem Solving owns the process of moving from uncertainty to a valid route and solution.

Problem posing owns the design or transformation of the question space itself.

The strongest learning loop often combines them:

pose → solve → test the result → inspect the construction → revise the posed problem → solve again.

Problem Posing Versus the Generation Effect

The Generation Effect owns the memory benefit that can arise when learners generate target information rather than merely read it.

Problem posing is not simply generating an answer. It asks the learner to generate the conditions, relationships and target that make a problem worth solving.

Problem Posing Versus Generative Learning

Generative Learning is the broader family of activities in which learners actively produce meaning, such as explaining, mapping, drawing or teaching.

Problem posing is one specific generative act: constructing a legitimate problem from knowledge of the domain.

The Quality Test: Is It Actually a Problem?

A learner-generated question should pass several tests.

Validity: are the relationships internally correct?
Sufficiency: is enough information supplied?
Solvability: can the requested target actually be determined?
Non-triviality: does the problem require the intended idea rather than collapse immediately?
Clarity: is the wording interpretable without hidden assumptions?
Alignment: does the problem genuinely exercise the concept being learned?
Answerability: can the learner verify a solution independently?

The Constraint Ladder

Free problem posing can be cognitively expensive. Novices often benefit from constraints before freedom.

Level 1 — Change one number. Keep the structure fixed and preserve solvability.
Level 2 — Change the unknown. Reverse what is given and what must be found.
Level 3 — Change one condition. Ask what new relationship appears.
Level 4 — Use the same representation to ask a different question.
Level 5 — Change representation. Turn a graph into a story, an equation into a diagram or a table into a prediction.
Level 6 — Create a near-miss. Pose a problem that looks similar but requires a different method.
Level 7 — Free construction. Build a problem from the concept with minimal scaffolding.

Mathematics: Change the Unknown

Suppose a standard question gives the radius of a circle and asks for its area.

A posing sequence can ask students to redesign it:

give the area and ask for the radius;
give the circumference and ask for the area;
compare two circles under a fixed ratio of radii;
create a question where rounding changes the final interpretation;
create a realistic context in which the area matters but circumference does not.

Each new problem tests whether the learner understands the relationships among quantities rather than one familiar input-output route.

Additional Mathematics: Pose the Condition That Forces the Method

Advanced mathematics becomes especially revealing when the learner must create a problem that requires a method.

For differentiation, ask: “Create a function and a condition such that finding a stationary point is necessary to answer the question.”

Now the student must understand not only how to differentiate, but why the derivative is relevant to the target state.

Science: Pose a Question the Evidence Can Answer

Science problem posing requires a discipline that casual curiosity does not: the question must match the evidence that could reasonably be collected.

Given a plant-growth dataset, learners can pose:

a descriptive question about a pattern;
a comparison between two conditions;
a prediction for an unmeasured value;
a question the current data cannot answer;
a follow-up investigation that would answer it.

The distinction between answerable and unanswerable questions is part of scientific literacy.

English: Build the Question From the Text

Give a learner a paragraph and ask for three comprehension questions at different levels.

one whose answer is directly stated;
one requiring inference from two clues;
one that looks plausible but cannot be answered from the passage.

To construct the questions, the learner must distinguish text evidence from assumption. The third task is especially useful because it requires understanding the boundary of the evidence.

History and Humanities: Change the Lens, Not the Facts

Given the same source set, learners can pose different legitimate questions:

What does this source suggest about public opinion?
What evidence would be needed to test the author’s claim?
How might an economic historian frame the same event differently?
Which question cannot be answered reliably from these sources alone?

Problem posing becomes a lesson in the relationship between question, evidence and disciplinary lens.

Vocabulary: Pose a Context That Forces the Word

Instead of asking for a sentence using reluctant, ask the learner to design a situation where reluctant is the most precise word and a near synonym such as unwilling would lose something important.

The learner is effectively posing a lexical discrimination problem.

Training and Professional Learning: Build the Failure Scenario

Problem posing is useful beyond school. In professional training, ask learners to create a scenario that would expose a known failure mode.

A cybersecurity trainee can design a phishing case that defeats superficial checking. A nurse can construct a handover scenario where one missing detail changes the decision. An engineer can pose a boundary condition under which a normally safe assumption fails.

Building the scenario requires understanding how the system can break.

Cross-Domain Lens: Test Engineers Do This for a Living

Software testing is partly the art of posing questions to a system.

What happens at the boundary? What if the input is empty? What if two events occur in the wrong order? What if the normal assumption is violated?

A good test case is a posed problem designed to reveal whether the system’s model is robust. Learner-generated problems can play a similar role: they test the learner’s model by forcing it to produce cases, not merely consume them.

Cross-Domain Lens: Science Advances by Better Questions

A scientific field does not progress only by solving a fixed list of questions. Progress also depends on reframing the problem, identifying a measurable unknown and asking a question that separates competing explanations.

School problem posing is much smaller in scale, but the intellectual move is related: knowledge is used to define what is worth asking next.

Failure Mode 1: Trivial Mutation

A learner changes 5 to 6 and calls it a new problem.

This can be useful at the first rung of the ladder, but it does not yet demonstrate deep structural understanding.

Repair: constrain the next pose to change a relationship, unknown or representation rather than only a number.

Failure Mode 2: Impossible Problem

The learner asks for an answer that cannot be determined from the information supplied.

Do not treat this only as failure. It can be diagnostic gold. Ask which missing piece would make the problem solvable.

Failure Mode 3: Ambiguous Language

The mathematics may be sound while the wording permits multiple interpretations.

Repair: make learners exchange posed problems and solve each other’s versions. Ambiguity becomes visible when another mind encounters the question.

Failure Mode 4: The Problem Does Not Need the Intended Idea

A learner is asked to create a percentage problem but designs one that can be answered by simple subtraction without using percentage reasoning.

Repair: add an alignment test—“What concept must a solver use, and why can the problem not be solved correctly without it?”

Failure Mode 5: Free Posing Before the Knowledge Exists

Novices can become lost when told simply to “create a hard question.” They may spend cognitive effort inventing a story rather than reasoning about the target concept.

Repair: provide a source problem, example, representation or constraint. Recent meta-analytic work suggests that problem examples can strengthen the effectiveness of problem-posing interventions.

The Pose–Solve–Audit Loop

Pose: construct the problem.
Predict: state what knowledge the solver should need.
Solve: solve your own problem.
Audit: check validity, sufficiency and clarity.
Exchange: let another learner solve it without explanation.
Diagnose: compare their interpretation with your intended structure.
Revise: improve the problem.
Generalise: state what you learned about the underlying concept from building it.

A 30-Minute Home-Study Routine

Minutes 0–5: solve one representative problem and identify its givens, relationships and target.
Minutes 5–10: pose a near variant by changing the unknown.
Minutes 10–15: pose a structurally different problem using the same concept.
Minutes 15–20: solve both and repair any invalid or missing information.
Minutes 20–25: pose one problem that looks similar but should use a different method.
Minutes 25–30: write the rule that separates the two problem families.

A Teacher Protocol

Choose a concept where relationships can be varied meaningfully.
Start with structured or semi-structured posing.
Show examples of strong and weak posed problems.
Define quality criteria before asking for originality.
Require learners to solve their own questions.
Use peer solving to expose ambiguity.
Ask learners to explain which conceptual feature their problem was designed to test.
Move toward freer posing only after quality stabilises.

A Student Protocol

Begin with a problem you understand.
Identify what is given, what is related and what is asked.
Change one structural element.
Write the new question precisely.
Predict the method before solving.
Solve it completely.
Check whether all information was necessary and sufficient.
Ask someone else to solve it.
Revise from the mismatch between what you intended and what they understood.

Using AI as a Problem-Posing Partner

AI can generate unlimited questions. That is exactly why the learner should not surrender the posing work entirely.

A more productive pattern is to ask the learner to pose first, then use AI as a critic:

Is the problem solvable?
Is any information missing?
Can it be solved without the intended concept?
Is there another legitimate interpretation?
What one modification would make it more diagnostic?

Preliminary research on AI-supported problem posing suggests that different forms of direct and indirect AI scaffolding may shape reflection differently. The evidence is still young. The robust design principle is older: assistance should improve the learner’s construction process, not replace it.

What Problem Posing Diagnoses

Concept boundaries: does the learner know what counts?
Dependency knowledge: which facts must be known first?
Constraint knowledge: what conditions make the relationship valid?
Representation: can the learner move between story, symbol, table, graph and diagram?
Evidence discipline: can the question be answered from the available information?
Method selection: can the learner design a question that genuinely demands a target method?

Canonical Owner Boundaries

This page owns problem posing: constructing, reformulating or extending problems as a learning activity.

Problem Solving owns finding routes through already-defined problems.
The Generation Effect owns memory effects from producing target information rather than merely reading it.
Generative Learning owns the wider family of meaning-producing learning activities.
Questions by Difficulty owns the dimensions that make questions harder or easier.
This page owns the learner’s act of designing or transforming the problem itself.

Evidence and Limits

The strongest quantitative evidence base for problem posing is in mathematics education. Recent meta-analyses support meaningful positive average effects, but the studies vary in age group, duration, technology use, scaffolding and outcome measures.

That evidence should not be casually stretched into a claim that learner-generated questions improve every subject equally. In science, language and professional training, the mechanism is plausible and the activity can be educationally valuable, but local assessment is still needed.

Problem posing also consumes time. For foundational fluency, direct explanation and ordinary practice may sometimes be the better use of a lesson. The method earns its place when building the question reveals structural understanding that solving alone would leave hidden.

The Return Path

Return to a familiar worksheet problem.

Solve it once.

Then stop being the candidate.

Become the examiner, engineer, scientist or author who must decide what the problem should be.

Change the unknown. Remove one fact. Add a constraint. Reverse the relationship. Ask what becomes impossible. Make somebody else solve it.

You may discover that constructing a legitimate question is harder than answering one—and that the difficulty exposes exactly what you know.

Problem posing works when building the question forces the learner to reveal the structure that ordinary answering can keep hidden.

Use This Tomorrow

Take one question you can solve. Create three versions: one with a different unknown, one with one condition changed, and one that looks similar but requires a different method. Solve all three. For each, write one sentence explaining why the supplied information is sufficient and why the intended method is necessary. If you cannot do that, the posing task has found the next thing to learn.

Research and Further Reading

eduKateSG Learning Node Series · 0050 of the continuing series. Previous: 0049 — How Subgoal Labeling Works. Continue through the Study & Learning Methods Hub and the wider eduKateSG Learning Hubs.