eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

How Mathematical Heuristics Work | Why a Good Problem-Solving Move Is Not a Formula

—

by

eduKate Asia
Three students in school uniforms work through open books at a classroom table, with textbooks and stationery nearby and study notes on the whiteboard behind them.

A student can know algebra, fractions, geometry and percentages and still freeze when a problem does not announce which chapter it belongs to. That is one of the strange things about mathematics education. We can teach many correct procedures and still leave a learner with no obvious first move when the surface changes.

Imagine a classroom floor after a lesson. One student has written three equations and crossed out two. Another has drawn a rough rectangle that is not to scale. A third has made a tiny table of values. None of those moves is the final method. Yet one of them may be the action that opens the problem.

That opening move is often a heuristic: a sensible problem-solving strategy that improves the chance of progress without guaranteeing the answer. Common mathematical heuristics include drawing a diagram, making a table, working backwards, simplifying the problem, looking for a pattern, considering an extreme case, introducing useful notation, decomposing the problem, estimating, testing a small case, changing representation, or asking what must be true.

The important word is strategy, not recipe. A heuristic is not an algorithm that always works. It is a way of reducing uncertainty about what to do next.

This distinction matters because modern mathematics education increasingly asks learners to solve unfamiliar, multi-step and non-routine problems rather than merely repeat rehearsed procedures. Current Education Endowment Foundation work on secondary mathematics identifies problem solving, metacognitive thinking and heuristics as promising areas, while its 2026 Thinking Mathematically pilot explicitly trained teachers to model problem-solving approaches, use heuristics and debrief pupils about their methods. The older but still useful What Works Clearinghouse practice guide on mathematical problem solving likewise emphasises monitoring, visual representations and exposure to multiple strategies.

The interesting question, then, is not simply which heuristics should students memorise? It is harder and more useful:

How does a learner recognise when a particular heuristic is worth trying, notice whether it is helping, and abandon it when it is not?

That is the real mechanism.

The 50-second answer

Mathematical heuristics work by giving a learner productive moves under uncertainty.

When a familiar procedure is available, the learner can execute it. When no obvious procedure is available, a heuristic changes the problem into a form that reveals more information. Drawing may expose geometry. A table may expose variation. A smaller case may expose structure. Working backwards may reveal necessary intermediate states. Estimation may expose an impossible answer. A change of representation may turn a verbal tangle into an equation or diagram.

A useful heuristic therefore does four jobs:

  1. reduces the search space — fewer possible directions need to be considered;
  2. makes structure visible — relationships hidden in the original form become easier to inspect;
  3. creates feedback — the learner can see whether the move is producing progress;
  4. supports metacognition — the learner can ask, “Is this still a sensible route?”

But heuristics fail when teachers turn them into a checklist to apply mechanically. “Draw a diagram” is not useful if the learner draws decoration rather than structure. “Look for a pattern” is dangerous if any pattern is accepted without proof. “Work backwards” is unhelpful when the end state does not constrain earlier steps.

The goal is not to give students fifty tricks. It is to build a small repertoire of moves, connect each move to the kinds of information it reveals, model selection aloud, and repeatedly ask learners to compare why one route worked better than another.

A heuristic becomes powerful when the student can say not only what they tried, but why that move was reasonable here.

First, what exactly is a heuristic?

In ordinary speech, heuristic can sound like a sophisticated word for “tip”. In mathematics, it is more precise than that.

An algorithm is a procedure intended to produce a correct result when its conditions are satisfied and its steps are followed correctly. Long division is algorithmic. Solving a linear equation by valid inverse operations can be algorithmic. A standard method for multiplying matrices is algorithmic.

A heuristic does not promise completion. It is a search strategy.

Suppose a learner is asked:

A rectangle has a fixed perimeter. What dimensions make its area as large as possible?

A learner who already knows optimisation may form a function and differentiate. But for a younger learner, several heuristics can enter before any formal theorem:

  • make a table of possible length–width pairs;
  • draw several rectangles;
  • compare extreme cases;
  • look for a pattern as the dimensions become more balanced;
  • test the conjecture that a square is best.

Each move changes what the learner can see. None, by itself, is a proof.

This is why heuristics sit between knowledge and invention. They are not random creativity. They are disciplined ways of searching.

That boundary also separates this article from broader problem-solving guidance. eduKateSG already has a general owner for How Problem Solving Works and related pages about method selection and changing representation when a problem will not open. The narrower job here is the heuristic layer: the reusable, non-guaranteed moves that help a learner search an unfamiliar mathematical problem.

Stage 1: the learner notices that routine execution has ended

Heuristics are most useful at a boundary.

A routine exercise says, in effect, “Use the method you have practised.” An unfamiliar problem says, “First decide what kind of thing this is.”

That change creates a common failure mode: a student keeps searching memory for a matching template even when no close template exists. The learner may ask, “Which formula?” before asking, “What is happening here?”

Good problem solvers detect that shift earlier. They recognise a method gap.

This is a metacognitive act. The learner notices:

  • I understand the words, but no procedure is obvious.
  • I have some relevant knowledge, but I do not yet know how to organise it.
  • The first job is exploration, not execution.

That diagnosis matters because heuristics are not substitutes for basic knowledge. A student who cannot manipulate fractions is not helped much by being told to “think creatively” about a fraction problem. Heuristics operate on knowledge. They help select, connect and reorganise what the learner already has.

A useful teaching sentence is:

“We do not know the whole route yet, but we can choose a move that gives us more information.”

That sentence changes the emotional meaning of uncertainty. Not knowing the complete method is no longer evidence that the learner is stuck forever. It is the normal starting condition for problem solving.

Stage 2: a heuristic changes the representation or scale of the problem

Many strong heuristics work because they transform the problem.

Take a simple example:

There are 100 lockers, all closed. On pass 1, every locker is toggled. On pass 2, every second locker is toggled. On pass 3, every third locker is toggled, and so on. Which lockers are open at the end?

Trying to simulate all 100 lockers by brute force is possible but unpleasant. A useful heuristic is solve a smaller case.

Try 10 lockers. Now patterns become inspectable. A locker changes state once for each divisor it has. Most divisors occur in pairs. Perfect squares have one unpaired divisor. That smaller case can reveal the underlying structure: square-numbered lockers remain open.

The smaller case did not merely save time. It changed the cognitive scale so a relationship became visible.

Other heuristics transform representation:

  • a word problem becomes a diagram;
  • a geometric relation becomes coordinates;
  • repeated arithmetic becomes a table;
  • a sequence becomes differences between terms;
  • a process becomes a state diagram;
  • an unknown quantity becomes a variable;
  • a complicated expression becomes a simpler analogous expression.

This connects naturally with eduKateSG’s page on representational competence. Heuristics often work precisely because the learner is able to choose or create a form that makes hidden structure easier to inspect.

Stage 3: the learner uses the new form to generate information

A heuristic is useful only if it changes the learner’s information state.

Drawing a diagram is not automatically progress. The diagram must expose something. Making a table is not automatically progress. The table must organise variation. Trying a small case is not automatically progress. The case must reveal a structure that can inform the general problem.

This gives us a powerful test:

What did the heuristic reveal that I could not see before?

Suppose a student is told to find how many handshakes occur if every person in a group shakes hands with every other person exactly once.

The learner can list cases:

  • 2 people → 1 handshake;
  • 3 people → 3 handshakes;
  • 4 people → 6 handshakes;
  • 5 people → 10 handshakes.

Now the pattern of differences appears: +2, +3, +4. The learner may conjecture a triangular-number pattern or notice that each new person adds one handshake with every person already present.

The small cases have generated a rule candidate.

But this is where mathematics becomes more than pattern spotting. The learner must still justify the rule. The heuristic generated information; proof or argument must establish the claim.

Stage 4: the learner monitors whether the heuristic is paying rent

Bad problem solving often contains a long period in which the learner continues a route because they have already invested time in it.

Three pages of algebra appear. The student feels that stopping would waste the effort. So the algebra continues.

Heuristics need exit conditions.

If you draw a diagram, what will make it useful? If it adds no new relation, perhaps switch.

If you build a table, how many rows are enough before you ask what pattern is emerging?

If you work backwards, are the predecessor states becoming more constrained or simply multiplying?

If you test examples, are the examples discriminating among hypotheses or merely confirming what you already expect?

This is a metacognitive skill: the learner evaluates the strategy, not just the final answer.

Teachers can model this aloud:

“I have tried algebra for two minutes. It is getting longer without exposing the relationship, so I am changing representation.”

That sentence teaches something that polished worked solutions often hide: abandoning a valid but unproductive route can be mathematically intelligent.

Stage 5: after solving, the learner names what made the heuristic fit

The final answer is not the end of heuristic learning.

After the problem is solved, ask:

Why did this move help here?

Suppose working backwards solved a problem. The reusable lesson is not “always work backwards.” It might be: “The final state contained strong constraints, while the starting possibilities were numerous.”

Suppose a diagram helped. The reusable lesson might be: “The relationships were spatial and difficult to hold verbally.”

Suppose smaller cases helped. The lesson might be: “The general problem was too large to inspect, but the structure survived at small scale.”

This reflection converts a one-off trick into conditional strategic knowledge.

Over time the learner builds a library that looks less like:

“Here are ten strategies.”

and more like:

“When I see this kind of uncertainty, these moves are promising.”

That is the mature form of heuristic competence.

Heuristic 1: draw a diagram when relationships are spatial or relational

“Draw a diagram” is probably one of the most common problem-solving suggestions in mathematics.

It is also one of the most commonly wasted.

A useful diagram is not an illustration. It is a representation of relationships.

Suppose a problem describes two cyclists leaving different points, travelling at different speeds and meeting after some time. A decorative picture of two bicycles contributes little. A line segment with labelled starting points, directions, distances and time variables can expose the relationship.

Good diagrams:

  • label known and unknown quantities;
  • show adjacency, direction or containment;
  • make equality or proportion visible;
  • externalise information that is hard to hold mentally;
  • permit new deductions.

Poor diagrams merely redraw the story.

A teacher can improve the heuristic by asking: “Which relationship is the diagram supposed to show?”

That question makes the representational job explicit.

Heuristic 2: make a table when quantities vary together

Tables are especially powerful when a learner needs to see how one quantity changes as another changes.

Imagine a problem about the number of diagonals in polygons.

A table can organise:

  • 3 vertices → 0 diagonals;
  • 4 → 2;
  • 5 → 5;
  • 6 → 9.

The table invites comparison. Differences are visible. The learner may then reason structurally: each vertex connects to all non-adjacent vertices, but every diagonal is counted twice.

The heuristic does not eliminate reasoning. It creates a surface on which reasoning can happen.

Tables are particularly useful when:

  • there are multiple cases;
  • two or more quantities co-vary;
  • patterns are hard to see in prose;
  • systematic enumeration matters;
  • the learner risks losing cases.

The danger is collecting data without interpreting it. After making the table, the next question must be: What changed, and what stayed structurally the same?

Heuristic 3: solve a smaller or simpler case

Large problems can hide structure in volume.

Simplifying does not mean changing the problem into an unrelated one. It means preserving the mechanism while reducing scale or complexity.

If a combinatorial problem has 20 objects, try 3 or 4.

If a recursive process is difficult, run the first few iterations.

If a geometry problem has many lines, temporarily remove one condition and see what becomes clear.

If numbers are ugly, test a cleaner numerical analogue before returning to the original values.

The critical question is whether the simplification preserves the relevant structure.

A smaller case can mislead if the behaviour changes at larger scale. So the learner must eventually return to the original problem and justify generalisation.

Heuristic 4: work backwards when the endpoint strongly constrains the route

Working backwards is powerful when the final state contains more useful information than the initial state.

Consider a number puzzle:

I think of a number, multiply it by 3, add 5 and obtain 26. What was the number?

Working backwards is almost algorithmic here: subtract 5, divide by 3.

But in richer problems the heuristic helps because the endpoint narrows possible predecessor states.

It is useful in:

  • multi-step transformations;
  • games where the winning state is known;
  • construction problems;
  • proof planning;
  • path problems;
  • equations and inverse operations.

The learner must be careful that backward steps are reversible. Some operations lose information. Squaring, for example, can make two different values produce the same result. Working backwards then requires branching or extra constraints.

Heuristic 5: look for an invariant

Some problems are full of change. A good solver asks whether something does not change.

An invariant might be:

  • parity;
  • total sum;
  • area;
  • a remainder modulo a number;
  • colour balance;
  • connectivity;
  • distance under a transformation.

Suppose a game repeatedly changes two numbers but preserves their sum. The final state is constrained by that invariant no matter how many moves occur.

This heuristic is powerful because it reduces a dynamic problem to a stable property.

For school learners, invariants often need explicit modelling. They are less visually obvious than diagrams or tables. A teacher can ask, “Everything seems to be changing. What quantity might remain conserved?”

That question opens an advanced but widely useful line of mathematical thought.

Heuristic 6: test an extreme case

Extreme cases stress a claim.

If a formula or conjecture is true, what happens when a value becomes very small, very large, zero, one, equal to another variable, or close to a boundary?

Suppose a student claims that increasing the length of a rectangle always increases its area. If the perimeter is fixed, push the length toward the full perimeter. The width approaches zero and the area collapses. The extreme case exposes the missing condition.

Extreme cases are useful for:

  • testing conjectures;
  • checking formulas;
  • revealing hidden constraints;
  • understanding limits;
  • detecting impossible generalisations.

They are not proof by themselves, but they can quickly destroy a bad claim or expose the shape of a better one.

Heuristic 7: estimate before calculating exactly

Exact calculation can create an illusion of correctness.

A calculator produces 0.0376. The student trusts it because many digits appear.

Estimation creates an independent expectation.

If 19.8 × 4.9 is approximately 20 × 5 = 100, an answer of 9.702 should immediately look suspicious.

Estimation can also guide method choice. If only an order of magnitude matters, exact computation may be unnecessary. If two answer options differ widely, a rough estimate may resolve the problem faster.

As a heuristic, estimation does three jobs:

  • creates a plausibility range;
  • reduces dependence on exact arithmetic;
  • helps detect transcription or calculator errors.

It is especially valuable in examination conditions because it creates a second route to error detection.

Heuristic 8: introduce notation to compress repetition

Long verbal descriptions consume working memory.

Suppose a problem repeatedly refers to “the number of red counters after the nth move.” Naming that quantity \(R_n\) creates a handle.

Good notation is cognitive compression.

It can expose relationships such as:

\(R_{n+1} = R_n + 2\)

or:

\(A = l(20-l)\)

Once a relationship is written symbolically, the learner can manipulate structure rather than repeatedly parse prose.

The danger is premature symbolism. If notation is introduced before the learner understands what the symbols represent, the problem becomes more opaque. Notation is a heuristic only when it compresses known meaning.

Heuristic 9: decompose the problem

A hard problem may contain several easier subproblems.

Decomposition asks:

What would I need to know first in order to know the final answer?

Suppose a problem asks for the cost of tiling a border around a rectangular pool. The learner may need to find:

  1. the area of the outer rectangle;
  2. the area of the pool;
  3. the difference;
  4. the number of tiles;
  5. the cost.

The value of decomposition is not just organisational. It identifies dependencies. If the final answer depends on a quantity that is not yet known, that quantity becomes an intermediate goal.

This heuristic is closely related to planning in programming, proofs, engineering and complex writing. Mathematics is one training ground for it.

Heuristic 10: change the question slightly

Sometimes a direct question is hard because the learner cannot yet see which features matter.

Change the question while preserving the structure.

Instead of “Find the maximum,” ask, “What happens near the extremes?”

Instead of “Prove this for all n,” ask, “Why does it work for n = 1, 2, 3, and what changes when n increases by 1?”

Instead of “Which strategy wins?”, ask, “What would have to be true for Strategy A to be better?”

This heuristic can expose hidden assumptions and reduce an all-or-nothing demand into a sequence of tractable questions.

But the learner must eventually return to the original question. A transformed question is a bridge, not an escape.

Why lists of heuristics often disappoint

It seems easy to teach heuristics. Put ten on a poster.

  • Draw a picture.
  • Make a table.
  • Look for a pattern.
  • Work backwards.
  • Guess and check.
  • Simplify.

Now the class “knows problem-solving strategies.”

But naming a heuristic is not the same as using it intelligently.

The missing knowledge is conditional knowledge:

When is this move promising?

What information should it reveal?

How long should I persist?

What signs tell me to switch?

What makes the output valid?

Without these conditions, a heuristic list creates new rituals.

The student sees a hard problem, draws a picture because “draw a picture” is on the poster, learns nothing from the picture, then declares that the strategy did not work.

That is why modelling and debriefing matter so much. The teacher has to make selection logic visible.

The teacher think-aloud: model uncertainty, not theatre

A worked example usually shows a clean route. That is useful for learning a procedure.

Heuristic learning sometimes needs a different model: a teacher who does not pretend the route was obvious.

For example:

“I do not yet see an equation. The problem describes distances between several points, so I’m going to draw a line diagram. I’ll label only the information the question gives. Now I can see two overlapping intervals. That suggests subtraction rather than adding everything.”

The useful parts are:

  • the teacher names the uncertainty;
  • the teacher names the feature that triggered the heuristic;
  • the teacher explains what the representation revealed;
  • the teacher moves on.

Do not overperform “thinking aloud” until every simple problem becomes a drama. The model should expose selection that students cannot yet see.

This links to eduKateSG’s broader owner on Teacher Modelling. In mathematical heuristics, the special modelling job is to show why a search move is chosen under uncertainty.

Why comparison is one of the best heuristic teachers

Take one problem and solve it two ways.

Method A: algebra.

Method B: diagram plus reasoning.

Then ask:

  • Which method exposed the structure sooner?
  • Which required more prior knowledge?
  • Which is easier to check?
  • Which would scale better if the numbers changed?
  • Which method is more general?

This teaches strategy choice more directly than merely showing more examples.

Students often develop the false belief that one method is “the method” because textbooks present one polished solution. Comparing valid routes teaches that mathematical quality includes efficiency, transparency, generality and checkability.

It also teaches humility. A clever-looking route is not automatically the best one.

Common failure mode: using heuristics as excuses for weak knowledge

Heuristics cannot substitute for multiplication facts, algebraic manipulation, geometric properties, probability rules or mathematical vocabulary.

A learner needs material to think with.

There is an important instructional sequence:

knowledge → guided use → strategy comparison → independent selection

If the learner has no stable knowledge base, heuristic search becomes expensive and unreliable.

This is one reason current EEF mathematics work combines problem-solving approaches with strong subject teaching rather than framing heuristics as an alternative to explicit instruction.

A diagram is powerful because the learner understands what the labels mean.

A small case is powerful because the learner can calculate it.

An invariant is powerful because the learner has the conceptual language to recognise it.

Knowledge and heuristics are partners.

Common failure mode: “look for a pattern” without proof

Pattern recognition is seductive.

Try three cases. Notice 2, 4, 6. Predict 8.

Often correct. Not necessarily proven.

A mathematical heuristic can generate a conjecture. It cannot always justify one.

Students need a two-stage habit:

  1. discover: what pattern seems to appear?
  2. justify: why must that pattern continue?

The second stage protects mathematics from becoming numerology.

Ask, “What feature of the structure forces the pattern?” rather than “What comes next?”

Common failure mode: overusing guess-and-check

Guess-and-check has a poor reputation because random guessing is inefficient.

Systematic trial can be excellent.

The difference lies in information.

A useful guess changes the next guess.

Suppose a learner is solving \(x^2 = 50\) without a calculator. Testing 7 gives 49. Now the answer is known to be slightly above 7. That check narrowed the interval dramatically.

In optimisation, trying feasible cases may reveal a trend. In logic puzzles, a temporary assumption can expose contradiction. In number problems, modular cases can eliminate large classes of possibilities.

Teach students to ask: “What will I learn if this guess is wrong?”

If the answer is “almost nothing,” choose a better trial.

Common failure mode: stopping at the heuristic

A table reveals a pattern. A diagram suggests equality. A small case suggests a formula.

Then the learner stops.

Heuristics are often bridges to mathematics, not substitutes for mathematical completion.

The final solution may require algebra, proof, calculation, formal explanation or a checked argument.

Teach the transition explicitly:

explore → conjecture → formalise → verify.

That cycle respects both creativity and rigour.

Common failure mode: assessing only the final answer

Two students get the same problem wrong.

Student A chose a sensible diagram, identified the right relationship, then made an arithmetic error.

Student B never found a relevant representation and guessed a formula.

Both answers are wrong. Their learning needs are very different.

When teachers inspect only the final answer, heuristic competence is invisible.

Useful feedback can target:

  • problem interpretation;
  • strategy selection;
  • representation quality;
  • monitoring;
  • execution;
  • checking.

This does not mean every exam should award marks for any attempt. It means instruction should diagnose the layer that failed.

Common failure mode: the “creative student” label

Some students appear naturally inventive. Others appear procedural.

Those differences are real, but strategy repertoires are learnable.

Do not reserve heuristics for students who already enjoy puzzle solving. In fact, learners who freeze at unfamiliar problems may need explicit exposure to search moves most.

Likewise, do not assume a student who produces many ideas is necessarily strong at heuristic problem solving. Generating routes is only one part. The learner also needs to evaluate, select, justify and abandon.

Heuristic competence is disciplined flexibility.

How heuristics interact with metacognition

Heuristics answer: What might I try?

Metacognition asks: Why this? Is it working? What should I do next?

The two are tightly linked.

A learner with many heuristics but poor monitoring may switch constantly or persist too long.

A learner with good monitoring but no heuristic repertoire may correctly recognise being stuck and still have no move to make.

Instruction therefore needs both repertoire and regulation.

Current EEF research on secondary mathematics explicitly links heuristics, problem solving and metacognitive approaches. That makes sense at the mechanism level: unfamiliar mathematics is partly a problem of managing search.

For the broader learning-control mechanism, see How Metacognition Works | Thinking About Thinking So You Can Learn Better.

How heuristics interact with transfer

A formula often transfers poorly when the surface changes because the learner has memorised where it appeared rather than why it applies.

Heuristics can help transfer because they are more abstract.

“Draw a diagram” can apply to rate problems, geometry, probability and combinatorics.

“Test an extreme case” can apply to algebraic conjectures, geometry, optimisation and functions.

“Look for an invariant” can apply to games, transformations, number theory and conservation problems.

But this transfer does not happen automatically. Learners need examples across contexts and explicit comparison of what made the heuristic useful.

A good curriculum therefore revisits the same heuristic in different mathematical territories.

The student gradually learns the trigger, not just the chapter.

How heuristics interact with adaptive expertise

Routine expertise is the ability to perform known procedures accurately and efficiently.

Adaptive expertise adds the capacity to modify, combine or invent approaches when the situation changes.

Heuristics sit at that bridge.

A learner who knows one efficient routine should use it when it fits. Reinventing multiplication every time would be absurd.

But when the routine fails to open the problem, the learner needs search moves.

The strongest mathematics education therefore produces students who know when not to be creative and when creativity becomes necessary.

That is a subtler goal than “problem solving”. It is the ability to move between efficiency and exploration.

Evidence, caveats and what we should not claim

Current evidence supports the value of teaching mathematical problem solving through explicit strategy work, visual representations, monitoring and multiple approaches. The What Works Clearinghouse practice guide for Grades 4–8 gives strong evidence to monitoring and reflecting on the problem-solving process, and moderate evidence to visual representations and exposure to multiple strategies. EEF’s current secondary-mathematics research agenda likewise identifies heuristics and metacognition as promising areas requiring further high-quality study, and the 2026 Thinking Mathematically pilot tests professional development that includes modelling problem solving, heuristics and pupil debrief.

Several caveats are important.

Heuristic instruction is heterogeneous. “Teach heuristics” can mean a poster, teacher think-alouds, extended problem-solving programmes, metacognitive questioning or curriculum-specific routines. Effects from one implementation should not be assumed for all.

Problem-solving performance is domain-dependent. General strategies do not float free of mathematical knowledge. A learner cannot compensate indefinitely for missing concepts with generic problem-solving language.

Transfer is difficult. Students may use a heuristic when cued in the lesson and fail to use it later in an unfamiliar context.

Non-routine problems vary greatly. Some are difficult because of deep mathematics; others because of language, unfamiliar representation or poor task design.

Pilot evidence is not efficacy evidence. The EEF Thinking Mathematically study referenced here is a pilot focused on feasibility and implementation; it should not be described as proof of attainment impact.

The responsible conclusion is not “heuristics add X months of learning”. It is narrower and stronger: problem solving requires strategic search as well as mathematical knowledge, and students can be taught to make that search more explicit, varied and monitored.

If you are a learner

When you meet an unfamiliar problem, resist the urge to ask for the whole method immediately.

First, write what you know and what is unknown. This turns uncertainty from a feeling into a structure.

Second, ask what kind of information is hidden. Is it spatial? Does something vary? Is there a repeated process? Is the final state known? Are there too many cases? Is there a quantity that may stay unchanged?

Third, choose one heuristic because of that structure.

You might say:

  • “I will draw it because the relationship is spatial.”
  • “I will try small values because the general case is too large to inspect.”
  • “I will work backwards because the end state is fixed.”
  • “I will estimate because I need to know whether my exact answer is even plausible.”

Fourth, set a checkpoint. After a few lines or a minute or two, ask whether the move has produced new information.

Fifth, keep useful debris. A failed route may have generated a constraint, example or pattern that helps the next attempt.

Finally, after solving, write one sentence about the search rather than the answer: “The key move was ___ because ___.” That sentence is how a one-off solution starts becoming reusable knowledge.

If you are a parent

A child who says “I don’t know how to start” may not need the answer. They may need help identifying the kind of uncertainty they are facing.

Instead of saying, “Use algebra,” try questions such as:

  • What do you know for certain?
  • Could you draw the situation?
  • What would happen with smaller numbers?
  • Can you make a table?
  • What does the answer roughly have to look like?
  • If the final condition is true, what must have happened just before it?
  • Is there another way to represent the same information?

Do not fire all the questions at once. Pick one that fits the problem.

Also resist the urge to reward speed exclusively. Heuristic search can look messy. Crossing out, switching approaches and testing cases may be evidence of productive thinking, not confusion.

The parent’s best role is often to protect the learner’s ownership of the search while preventing complete paralysis.

If you are a teacher

Teach fewer heuristics more deeply.

For each one, specify four things:

  1. Trigger: what kind of problem feature makes the heuristic plausible?
  2. Action: what does the learner actually do?
  3. Information: what should the move reveal?
  4. Exit: how will the learner know whether to continue or switch?

Then vary the examples. If “draw a diagram” only appears in geometry, students may never use it in ratio, probability or algebra. If “make a table” only appears in sequences, its wider usefulness remains hidden.

Model strategy selection aloud, but avoid narrating every tiny thought forever. Support should fade. Learners eventually need to generate their own candidate moves.

Use board work to preserve selected false starts. A clean final solution is useful for consolidation; a carefully chosen messy route is useful for learning how mathematical search actually behaves.

When assessing problem solving, separate at least three things:

  • mathematical knowledge;
  • strategy selection and monitoring;
  • execution accuracy.

Otherwise “wrong answer” hides too much diagnostic information.

And connect heuristic teaching to genuine mathematics. A heuristic lesson should still end with valid claims, correct procedures, proof where appropriate, and answers that survive checking.

A Clementi-floor test: can the strategy survive a different chapter?

A good heuristic should not depend entirely on today’s topic heading.

Try this transfer test.

English-style data question: Make a table to organise evidence categories before comparing them.

Science investigation: Test an extreme condition to challenge a proposed relationship.

Geometry: Draw an auxiliary representation or decompose the figure.

Algebra: Substitute a simpler case to inspect structure.

Probability: Work backwards from a required outcome.

Number theory: Look for an invariant such as parity or remainder class.

The mathematical content changes. The search move survives.

That is the point.

Frequently asked questions

Are heuristics just tricks?

No. A trick is often narrow and surface-bound. A heuristic is a general search move connected to a type of structure or uncertainty. It still requires mathematical justification.

Should students memorise a list of heuristics?

A small list can help learners name strategies, but naming is not enough. Students need repeated practice selecting, using, monitoring and comparing the strategies in varied problems.

Is “guess and check” a legitimate heuristic?

Yes, if the guesses are informed and the checks generate information. Random guessing is inefficient. Systematic trial, bounded search and using feedback from each attempt can be mathematically powerful.

When should a student abandon a heuristic?

When it is no longer producing useful information, when the resulting work is becoming much more complex than the original problem, or when another representation now exposes the structure more directly.

Do heuristics replace worked examples?

No. Worked examples help learners acquire procedures and recognise structures. Heuristic problem solving becomes especially important when a ready-made procedure is not obvious. The two can support each other.

Are heuristics only for advanced students?

No. Young learners already use simple heuristics such as drawing, acting out, using smaller numbers, making organised lists and checking extremes. The repertoire and sophistication can grow with mathematical knowledge.

Can teachers assess heuristic use fairly?

They can assess observable reasoning: the relevance of a chosen representation, whether cases are organised, whether the learner monitors progress, and whether claims are checked. But avoid rewarding a particular named strategy when another valid route works better.

What if a student finds a completely different method?

That is often valuable. Ask the student to explain why the method works, what information it uses, and whether it generalises. Mathematics values valid structure, not obedience to a preferred classroom route.

Why do strong students sometimes fail on unfamiliar problems?

High procedural fluency can coexist with weak strategy selection. If most practice has signalled the method in advance, the learner may have had little opportunity to diagnose a new problem and choose a route independently.

What is the simplest habit to start with?

When stuck, ask: “What move would give me more information?” It shifts the goal from instantly knowing the solution to making the next intelligent step.

The deeper idea: mathematics is not only a library of methods

School mathematics can accidentally teach a false picture of expertise.

Chapter title. Worked example. Exercise type. Correct method. Answer.

Real mathematical thinking is less tidy. The solver often does not know the route at the beginning. They transform the question, inspect a smaller case, draw something, reject an approach, rename a quantity, compare examples, test a conjecture and only later discover the clean argument.

Heuristics make that middle visible.

They do not replace knowledge. They make knowledge mobile.

They do not guarantee success. They make uncertainty navigable.

They do not eliminate mistakes. They make mistakes informative.

And they do not turn every learner into a mathematician by handing over a poster of strategies. The real educational job is harder: help the learner build enough mathematical knowledge to have options, enough strategic knowledge to choose among them, and enough metacognitive control to notice when a route should change.

The best evidence that heuristic teaching is working is therefore not that a student can recite “draw a diagram, make a table, work backwards”. It is that, on an unfamiliar Tuesday-afternoon problem with no chapter label and no hint, the student begins doing something sensible—and can explain why.

Sources and further reading