Worked Example Variation¶

Varied worked-example set — instantiates Transfer Scaffolding

Presents the same solution structure across worked examples with systematically varied surface stories, so the learner abstracts the transferable method instead of binding to one cover story.

Teach a method with one example and learners memorize the example; teach it with several that share a structure but nothing else, and they extract the method. Worked Example Variation is a set of fully worked examples that present the same solution structure across systematically varied surface stories. Its defining move is that the variation itself is the teaching device: holding the deep structure constant while deliberately changing the surface forces the learner to abstract the transferable schema rather than bind to any one cover story — which is what makes the method survive contact with a problem it has never seen.

Example¶

A support team must learn a general troubleshooting method: isolate → reproduce → check recent changes → narrow by elimination → verify the fix. Trained on a single example — a login bug — agents pattern-match to login bugs and stall on anything else. Worked Example Variation gives several fully worked cases with identical structure but different surfaces: a login failure, a slow report export, a declined payment, a mobile crash. Each walks the same five moves; only the story changes. Because the surface varies while the steps repeat, agents extract the method instead of memorizing "the login fix." A companion note makes explicit which features changed (product area, symptom, urgency) and which stayed (the five moves).

The output is agents who apply the structure to a ticket type they have never encountered, because the varied examples taught them the schema rather than a single script.

How it works¶

Fix the deep structure — the same solution steps appear in every example.
Vary the surface systematically — swap the cover story, domain, and details across examples.
Flag the varied features as irrelevant and the constant steps as the point.
Fade from worked to independent — full worked solution → partially completed → solve unaided — as the schema takes hold.

Tuning parameters¶

Variation range — how different the surface stories are. Wider variation forces stronger abstraction but raises cognitive load.
Number of variants — enough to force generalization without fatigue.
Worked-to-faded ratio — how quickly examples shift from fully worked to problem-completion to independent solving.
Surface-feature salience — how explicitly the varied, irrelevant features are flagged as such.
Structure highlighting — how strongly the invariant steps are annotated across examples.

When it helps, and when it misleads¶

Its strength is being the direct antidote to surface-binding: it produces abstraction cheaply for novices who would otherwise over-fit to a single example.^[1]

Its failure modes are overload — too much variation too early swamps a novice before the structure is grasped, and worked examples that help novices can bore or hinder experts (the expertise-reversal effect) — and poorly chosen variants that accidentally share an irrelevant surface feature, which then becomes the false lesson. The classic misuse is calling near-identical examples "varied" so no abstraction is actually forced. The discipline is to vary the surface widely while holding the structure exactly constant, and to fade from worked to independent as competence grows.

How it implements the components¶

Worked Example Variation fills the abstraction-through-variation subset of the archetype:

surface_difference_map — the systematic record of which surface features vary across the examples and are therefore irrelevant to the method.
source_abstraction_prompt — varying the surface forces the learner to abstract the source method into a context-free schema; the variation is the prompt.

Where Analogical Mapping Prompt maps one source to one target, this varies surface across many examples of a single method; and it teaches through examples to read, not the graded practice tasks of Cross-Context Practice or Near-to-Far Transfer Sequence. It does not gate applicability or hunt for traps — those stay with their siblings.

Instantiates: Transfer Scaffolding — it is the varied-example set that drives abstraction of a transferable method.
Sibling mechanisms: Bridging Examples · Analogical Mapping Prompt · Case Comparison and Contrast · Cross-Context Practice · Near-to-Far Transfer Sequence · Apprenticeship Reflection · Boundary Condition Checklist · Implementation Coaching · Negative Transfer Red Team · Transfer Debrief · Transfer Prompt

Notes¶

Three example-based siblings differ by what is held fixed. Worked Example Variation holds the structure fixed and varies the surface to force abstraction; Bridging Examples holds the structure fixed and steps the surface from source-like to target-like to carry an intuition; Case Comparison and Contrast varies whole cases to induce a principle and its boundaries. Choosing the wrong one wastes the examples.

References¶

[1] The variability effect in worked examples — studied within cognitive load theory (Sweller and colleagues) — holds that worked examples spanning varied surface features promote transfer better than repeated examples of one type, provided the added variation does not overwhelm a novice's working memory. ↩