MindOS · Learning Operation · Observe → Explain → Predict → Complete → Fade → Solve → Transfer
Wait, What? A Perfect Worked Solution Can Teach Almost Nothing if the Student Only Watches It
A tutor solves a difficult problem beautifully. Every step is correct. The explanation is elegant. The student nods.
Then the tutor removes the page and gives a nearly identical problem.
The student cannot begin.
The worked example was not useless. It simply remained an external performance. The learner saw a route without yet owning the route.
Quick Answer
Worked examples are step-by-step models of how a problem is solved. They can be especially useful during initial skill acquisition because they reduce unproductive search and let novices study the structure of a successful solution. But their value depends on what the learner does with the example and on when the example is withdrawn.
Recent research continues to support a conditional view. A 2025 study in Contemporary Educational Psychology found that the choice of example itself matters: ambiguous examples can produce worse learning, misconceptions and overconfidence even when explanations are supplied. See Teaching with worked examples – Why the selection of problems for exemplification is critical. A 2025 systematic review of erroneous and contrasting examples likewise concludes that benefits depend on prompts, prior knowledge, cognitive demands and how errors are explained. See Conditions for Effective Learning from Erroneous Examples.
The Owned Job of This Page
This page owns the modelled-route acquisition job: when to show a complete or partial expert solution so a learner can acquire the structure of a new procedure. Working Memory Load owns overload discrimination. The later Scaffold Fading page owns how assistance is progressively removed across support types. This page owns the worked example itself: selection, processing and transition into learner performance.
CLEAR MODEL ↓ LEARNER EXPLAINS WHY ↓ LEARNER PREDICTS NEXT STEP ↓ PARTIAL EXAMPLE ↓ LEARNER COMPLETES MISSING STEPS ↓ INDEPENDENT PROBLEM ↓ NEW SURFACE / NEW CONTEXT ↓ TRANSFER
When a Worked Example Is Most Useful
- The learner is new to a procedure and has no workable route.
- Unguided problem solving produces random search rather than meaningful hypothesis testing.
- The task contains many steps and the learner needs to see their organisation.
- The learner understands the concepts but not how they are assembled into a solution.
- A new representation or notation makes a familiar idea look inaccessible.
Discrimination 1: Does the Learner Need a Model or a Missing Prerequisite?
A worked example cannot compensate for a concept the learner does not understand. If every line contains unfamiliar vocabulary, symbols or relationships, first repair the prerequisite. Otherwise the example becomes a sequence to copy rather than a structure to understand.
Discrimination 2: Is the Learner Studying or Merely Following?
Cover the next step and ask: “What should happen now?” Ask why the current step is legal, what quantity changed, what assumption is being used, and what would happen if the problem were slightly different. If the learner can only continue while seeing the completed line, the example is still carrying the process.
Discrimination 3: Is the Example Itself Misleading?
Examples are not neutral. A mathematically correct example can still be pedagogically poor if it uses an unusual case, hides the reason a step works, introduces unnecessary ambiguity, or accidentally teaches a shortcut that fails in normal cases.
Choose examples whose structure makes the intended principle visible. Later, deliberately introduce variation and edge cases after the learner has a stable base.
Make the Example Generative
The learner should do cognitive work while studying the model:
- predict the next step before revealing it;
- self-explain why each step follows;
- label the purpose of a line, not only the operation;
- identify which information in the problem triggered the method;
- compare two worked examples and locate what stays invariant;
- find and repair a carefully designed error when prior knowledge is sufficient;
- reconstruct the route from memory after the example is removed.
Recent research on stepwise examples also suggests that retrieval can be inserted into the example itself: asking learners to retrieve or execute upcoming steps before revealing them can improve later learning compared with simply reading the steps.
Worked Examples in Mathematics
Do not present only the algebra. Show the decision structure: what was noticed, why a representation was chosen, why one method is efficient here, where an error is likely, and how the final answer is checked. Then change the numbers or surface wording so the learner must reconstruct the same route without copying the surface.
Worked Examples in English
A worked example can model the construction of an inference answer, a composition paragraph or evidence selection. The useful object is not the polished prose alone. Make the hidden decisions visible: what the writer noticed, why this evidence was selected, how the explanation connects evidence to claim, and what alternatives were rejected.
Worked Examples in Science
For a Science explanation, model the causal chain rather than a memorised answer pattern. Ask the learner to identify observation, mechanism, relationship, evidence and expression. Then change the experiment and test whether the causal structure survives.
The Expertise Reversal Problem
Support that is useful for a novice can become redundant for a more expert learner. Once the route is stable, forcing the student to read complete solutions can waste attention and reduce opportunities for independent selection and execution.
So the worked example has an expiry condition: when the learner can predict, explain and execute the route, the model should begin to disappear.
Scaffold-Fade Logic
FULLY WORKED EXAMPLE ↓ FULL EXAMPLE + PREDICTION PAUSES ↓ PARTIAL EXAMPLE ↓ LATER STEPS REMOVED ↓ MORE STEPS REMOVED ↓ PROBLEM ONLY ↓ UNSEEN PROBLEM ↓ MIXED / EXAMINATION CONDITIONS
Transfer Test
- Remove the completed solution.
- Change the numbers or wording.
- Change the representation.
- Ask the learner to identify which earlier example is structurally relevant.
- Insert a plausible but incorrect route and ask for critique.
- Mix the problem with other methods so selection is required.
Return Test: What Came Back?
- Can the learner begin without seeing the model?
- Can they explain why the first step is appropriate?
- Can they recover after a deliberate interruption?
- Can they detect an error in a worked solution?
- Can they solve a new problem that preserves the deep structure?
- Does the learner still request the full model, or only a small cue?
Common Misconceptions
- “Worked examples are spoon-feeding.” They can reduce unproductive search during early acquisition when followed by active processing and fading.
- “If the example is correct, it is good.” Example selection and clarity matter.
- “The student understood because they followed every step.” Following is not independent reconstruction.
- “More examples are always better.” Once the route is stable, independent problems may provide greater learning value.
- “Errors should never appear in examples.” Erroneous examples can be useful under the right conditions, but they require careful design and sufficient learner support.
Parent and Tutor Teaching Guide
When showing a solution, do not let the learner become an audience. Stop. Cover a line. Ask what comes next. Ask why. Then remove the model and make the child reconstruct the route. A good worked example should eventually make itself unnecessary.
MindOS Direction Graph
WORKED EXAMPLE STATE ├── Missing prerequisite? → CONCEPT REPAIR ├── No route at all? → FULL MODEL ├── Following without explanation? → SELF-EXPLAIN / PREDICT ├── Model ambiguous? → SELECT CLEARER EXAMPLE ├── Route becoming stable? → PARTIAL EXAMPLE ├── Model now redundant? → FADE ├── Can execute but not select? → INTERLEAVING / STRATEGY SELECTION └── New problem survives? → TRANSFER
Continue Through MindOS
MindOS boundary: Worked examples are instructional tools, not measures of intelligence or clinical cognition. Their usefulness depends on prior knowledge, task design and the learner’s actual return performance.
eduKate Sengkang 