How Problem Solving Works | From Uncertainty to a Usable Route

Problem solving is the process of reducing the gap between a current state and a desired state when the route is not immediately obvious.

In one line: problem solving works when we represent the problem accurately, define the goal and constraints, select a plausible route, monitor what happens, and change the route when evidence shows it is not working.

Routine exercises are useful, but not every exercise is a genuine problem. If the method is already known and the learner only has to execute it, the main challenge may be accuracy or fluency. A problem becomes more demanding when the learner has to decide what the situation means, which knowledge applies, what information matters, and what to try first.

What Is a Problem?

A problem contains three broad elements:

a current state;
a desired state or goal; and
a route that is incomplete, uncertain or constrained.

The problem-solving process therefore asks:

Where am I? Where do I need to get to? What prevents a direct move? What route is worth trying next?

1. Represent the Problem Before Solving It

Many failures happen before the first calculation or action. The learner misreads the situation.

Problem representation identifies the objects, relationships, constraints and target. It may take the form of a diagram, equation, timeline, table, list of knowns and unknowns, causal map or verbal restatement.

A good representation removes irrelevant detail and makes important structure visible.

2. Clarify the Goal

Vague goals produce vague searches. “Fix my mathematics” is too large. “Work out why I choose the wrong method in ratio word problems” creates a more useful problem.

A clear goal also allows the solver to recognise completion. Without a success condition, activity can continue without progress.

3. Identify Constraints and Resources

Problems live inside constraints: time, rules, available information, tools, money, memory, safety, prior knowledge or social requirements.

Constraints are not merely obstacles. They shape the solution space. A mathematics proof has different admissible moves from an engineering design problem. A school decision differs from a laboratory experiment because the consequences, evidence and reversibility are different.

4. Retrieve Candidate Strategies

Solvers draw on prior knowledge: formulas, examples, heuristics, analogies, procedures and memories of similar situations.

Novices often struggle because they have fewer strategies to retrieve or because they recognise problems by surface features. Experts are more likely to notice deeper structure and retrieve a suitable family of approaches quickly.

5. Strategy Selection Is a Judgement

Knowing several methods is not enough. The solver must choose among them.

Selection may depend on efficiency, reliability, available data, reversibility, cost or the need for explanation. Sometimes the best first move is not a complete solution but an information-gathering action that reduces uncertainty.

This is why strong problem solving often looks like a sequence of smaller decisions rather than one flash of insight.

6. Execute While Monitoring the Route

Once a strategy begins, the solver should not switch off judgement. Ask whether the intermediate results make sense, whether the route is moving toward the goal, and whether a constraint has been violated.

Monitoring can catch a poor route early. This matters because persistence is not the same as repeating a failing strategy for longer.

7. Feedback Decides Whether to Continue, Repair or Reroute

Problem solving is iterative. Results from one move become information for the next.

Try → observe → compare with goal → continue, repair or reroute.

A failed attempt is not automatically wasted if it removes a possibility, reveals a hidden constraint or exposes a misunderstanding.

8. Decomposition Makes Large Problems Tractable

Large problems can exceed working memory because too many relationships must be managed at once. Decomposition breaks the problem into smaller subproblems whose outputs can later be recombined.

In writing: clarify the claim, gather evidence, organise paragraphs, then refine language. In mathematics: identify known relationships, solve an intermediate quantity, then use it to reach the final target. In life decisions: separate what is known, what is uncertain, what is reversible and what must be decided now.

9. External Representations Carry Cognitive Load

Writing, drawing and modelling allow parts of the problem to exist outside the head. This reduces the amount that must remain actively held and makes relationships easier to inspect.

The US Institute of Education Sciences practice guide on mathematical problem solving identifies visual representations and monitoring of the problem-solving process as evidence-supported instructional approaches. The principle extends beyond mathematics: better representations often create better routes.

10. Transfer Requires Solving Beyond the Familiar Template

A learner can memorise a solution pattern without becoming a flexible problem solver. Transfer requires changed surface features, mixed problem types, missing cues and sometimes genuinely novel combinations.

The goal is to recognise structure even when the question does not announce which chapter it belongs to.

The Whole Problem-Solving Chain

Represent → define goal → identify constraints → retrieve strategies → choose a route → execute → monitor → use feedback → repair or reroute → verify → generalise what was learned.

A Useful Metaphor: Problem Solving Is Navigation

You know where you are and roughly where you want to go, but the road may be unfamiliar. A map helps represent the territory. Constraints close some routes. Feedback tells you whether you are getting closer. A wrong turn can still teach you something about the landscape.

The best navigator is not the person who never reroutes. It is the person who detects a bad route early and updates without losing the destination.

Problem Solving at Three Zoom Levels

Micro: the next move

What is the smallest action that reduces uncertainty or moves toward the goal?

Meso: the strategy

Is the overall route appropriate, efficient and responsive to feedback?

Macro: adaptive capability

Can the learner carry problem-solving habits into unfamiliar domains where the problem itself must first be discovered and defined?

How Problem Solving Fails

Representation failure: the wrong problem is solved accurately.
Goal ambiguity: activity continues without a clear success condition.
Strategy fixation: one familiar method is repeated despite contradictory evidence.
Working-memory overload: too many elements remain internal and unorganised.
No monitoring: errors accumulate before the solver notices the route has failed.
Answer chasing: the solver looks for an output without understanding the structure.
Template dependence: capability collapses when surface features change.

How Problem Solving Is Repaired

Stop execution and return to representation. Restate the problem. Draw it. Identify what is known, unknown and constrained. Compare with one solved example and one deliberately different example. Ask what evidence would tell you the current route is working.

Often the repair is not “try harder.” It is “build a better model of the problem.”

What Parents and Students Should Notice

Can the student restate the problem in their own words?
Can they identify the goal and constraints?
Do they have more than one possible strategy?
Can they explain why they selected this route?
Do they monitor intermediate results?
Can they change approach without treating rerouting as failure?
Can they solve a structurally similar problem with different surface features?

Some Problems Must First Be Discovered Before They Can Be Solved

Textbook problems usually announce themselves. Real-world problems often do not. A student may say, “I need to study more,” when the actual problem is poor method selection. A school may say, “attendance is the problem,” when the deeper problem is unsafe climate, transport or curriculum disengagement. An organisation may optimise a visible symptom while the causal bottleneck sits elsewhere.

This creates a layer before ordinary problem solving: problem finding and problem framing. The solver has to decide what state is undesirable, what evidence shows a gap, and whether the first description names a cause, a symptom or only a consequence.

Signal → define the actual gap → separate symptom from cause → identify the controllable problem → then solve.

A Representation Is a Model of the Problem — and Models Can Be Wrong

Representations are not neutral containers. A diagram, equation, table or verbal restatement emphasises some relationships and suppresses others. The representation can therefore make a route visible or hide it.

A strong solver asks two questions before committing:

What structure does this representation make easier to see?
What important feature might it be hiding or simplifying?

Changing representation is not cosmetic. In Mathematics, a word problem may become solvable after drawing a diagram or expressing a relationship algebraically. In Science, a process diagram can expose a missing causal step. In writing, an argument map can reveal that two paragraphs are trying to do the same job.

Problem Spaces Contain States, Moves and Constraints

A useful formal way to think about many problems is as a problem space. There is a starting state, a goal state, possible moves, and constraints on which moves are allowed or affordable.

This matters because expertise often means shrinking the search space. A novice sees many possible moves and has little basis for choosing. An expert recognises structure that rules out large parts of the space before calculation begins.

Better knowledge does not merely produce faster execution; it changes which routes the solver considers worth searching.

Heuristics Help Search — but Every Heuristic Has Boundary Conditions

Problem solving uses heuristics such as work backwards, draw a simpler case, decompose the task, search for invariants, compare with a solved example, estimate first, or test an extreme case. These are not guaranteed algorithms. They are search tools.

A heuristic is valuable when it reduces the space enough to reveal useful structure. It becomes dangerous when learners memorise the heuristic itself without recognising when the problem does not fit.

This is why multiple-strategy exposure matters. The US Institute of Education Sciences’ mathematical problem-solving guide supports teaching visual representations, monitoring and reflection, and exposing students to multiple strategies rather than training one universal recipe.

Worked Examples and Independent Problem Solving Solve Different Learning Problems

Beginners often benefit from worked examples because a complete route allows limited working memory to focus on the relationships that make the solution work. Independent problem solving asks the learner to generate and select the route themselves.

The stronger progression is therefore not “examples or problem solving?” but:

Study an intelligible example → explain why each move is legal → complete partially worked cases → solve independently → compare alternative strategies → solve changed and mixed problems.

Removing support too early wastes cognitive resources on blind search. Keeping support too long creates imitation without independent selection.

Strategy Switching Is a Skill, Not an Admission of Failure

Good problem solving requires a threshold for abandoning or modifying a route. Switch too quickly and the learner never persists long enough to discover the structure. Switch too late and persistence becomes sunk-cost repetition.

A useful monitoring question is:

What evidence should I have seen by now if this strategy were working?

If the predicted intermediate signal does not appear, the solver can revisit the representation, check execution, change the strategy, or gather information that discriminates among those explanations.

Verification Is a Separate Stage From Reaching an Answer

Producing a candidate answer is not the same as solving the problem. The output still has to satisfy the goal and constraints.

Verification can ask:

Does the answer satisfy the original conditions?
Are the units, scale and direction plausible?
Can the result be reached by a different route?
Does substituting the result back into the original relationship work?
What happens in an extreme or boundary case?
Did the solution create a new problem elsewhere?

In real systems, verification often includes consequences after implementation. A policy can be internally coherent yet fail at delivery. A study plan can look efficient but produce weak retention. The world return is part of the solution test.

Optimization Is Different From Satisficing

Some problems require the best possible solution under a defined objective. Others only require a solution that is safe, legal, affordable and good enough before a deadline. Searching indefinitely for the theoretical optimum can itself become a failure when delay has a cost.

This distinction matters in student work. A one-mark examination item should not consume ten minutes of elegant exploration. A high-stakes design problem may justify a much deeper search. Problem-solving quality includes matching search effort to the consequence and value of improvement.

Information-Gathering Actions Can Be Better Than Solution Actions

When uncertainty is the dominant obstacle, the best next move may be to learn rather than to solve directly. Ask a discriminating question, take a measurement, test one small case, inspect a failed example or request missing information.

This is especially valuable when a cheap observation can prevent a large wrong intervention. In learner diagnosis, one marked-paper sample can sometimes separate a knowledge failure from a time-pressure failure more effectively than hours of generic practice.

Constraint Violations Can Make a Clever Solution Invalid

Solutions are admissible only inside the actual constraints. An examination answer must obey the question and available time. An engineering solution must satisfy safety and material limits. A school solution must respect student welfare, legal duties, resources and curriculum. A life solution must fit the person’s real values and obligations.

This creates a useful distinction:

Possible solution ≠ admissible solution ≠ preferred solution.

Transfer Depends on Recognising Deep Structure, Not Copying Surface Form

Students often appear proficient because practice problems repeat familiar cues. True problem-solving transfer requires recognising the same underlying relationship when the numbers, context, diagram, wording or sequence changes.

One powerful teaching method is contrast: place two superficially similar problems side by side where different methods apply, or two superficially different problems where the same deep structure applies. Ask the learner to explain the feature that changes the method choice.

This turns strategy selection from chapter recognition into structural discrimination.

Complex Problems Need Layered Ownership

Many adult problems are too large for one person. Teams decompose the problem, assign subproblems to specialists, integrate outputs and maintain responsibility for the whole. This creates a new failure mode: each local piece can be solved correctly while the integrated system still fails.

Good collaborative problem solving therefore needs interfaces: shared definitions, constraints, assumptions, handoff conditions and a final system-level verification.

A High-Resolution Problem-Solving Audit

Signal: What observation tells us a problem exists?
Framing: Are we naming the cause, the symptom or only the consequence?
Current state: What is true now?
Goal: What state would count as sufficiently solved?
Representation: Which model best exposes the important relationships?
Constraints: What moves are prohibited, costly or irreversible?
Resources: Which knowledge, tools, people and information are available?
Search space: What families of routes are plausible?
Strategy: Why is this next move better than the alternatives?
Information value: Is the next best move actually a measurement or question?
Monitoring: What intermediate signal should appear if the route is working?
Switch threshold: What evidence should trigger repair or rerouting?
Verification: Does the candidate solution satisfy the original goal and constraints?
Efficiency: Is further optimisation worth its time and cost?
Transfer: What deep structure should be reusable in a changed problem?
World return: After implementation, did reality behave as the solution predicted?

Evidence Boundary: Problem Solving Is Teachable, but It Is Not Content-Free

The US Institute of Education Sciences’ Improving Mathematical Problem Solving in Grades 4 Through 8 rates evidence as strong for helping students monitor and reflect on the problem-solving process and for teaching visual representations; it also recommends exposure to multiple strategies. Its recommendations are mathematics-specific, but they illustrate a wider principle: representation, strategy repertoire and monitoring can be taught explicitly.

At the same time, problem solving is not a generic mental muscle independent of knowledge. Expertise changes what the solver notices, how the problem is represented and which routes are even visible. Strong problem solving therefore develops through domain knowledge + strategy knowledge + monitoring + varied transfer.

Connect Problem Solving to the Wider eduKateSG Mechanism Estate

How Thinking Works — the attention, representation and working-memory system underneath problem solving.
How Knowledge Works — why expertise changes the search space.
How Reasoning Works — how candidate routes and explanations are evaluated.
How Decision-Making Works — how one admissible route is chosen when consequences and uncertainty matter.
How Feedback Works — how intermediate results trigger continuation, repair or rerouting.

Continue Through eduKateSG

Evidence and Further Reading

The US Institute of Education Sciences Improving Mathematical Problem Solving in Grades 4 Through 8 recommends helping students monitor and reflect on their problem-solving process and teaching them to use visual representations. The guide rates evidence for both recommendations as strong, providing a useful research anchor for two mechanisms described here.

Frequently Asked Questions

Is problem solving just intelligence?

No. Problem solving depends on knowledge, representation, strategy repertoire, working-memory management, monitoring and experience. Many of these components can be taught and practised.

Should students struggle without help?

Not indefinitely. Productive struggle should generate thinking and information. When the learner lacks a prerequisite or has built the wrong representation, targeted support can make later independent problem solving more likely.

Why can students solve practice questions but fail unfamiliar ones?

Practice may have trained execution of a known method without training problem representation and strategy selection. Mixed and varied problems are needed to develop recognition of deeper structure.

Final compression: Problem solving works when uncertainty is converted into a usable representation, a route is selected and tested, and feedback keeps the solver willing to repair or reroute until the goal is reached.