Analogy generation

Generate a handful of candidate analogies for one idea, keep the one whose structure actually maps, then restate the target without the metaphor so the story becomes a bridge, not a substitute.

When to use this: First time through the material
Learning impact: Clear help, modest gain
Learning evidence: Credible with caveats
Where you are with the topic: Meeting the idea
How often to use it: Useful once for a topic

When to use it

You are meeting this idea for the first time and need a way in.
The reading feels dense and you want a first map before the details.
You asked for an analogy once, liked it, and now cannot write the real definition.
A friend asks you to explain the idea and you only have a story, not the mechanism.

What to ask AI

Type something like: “Give me THREE analogies for why adaptive immunity remembers a pathogen after”.

Ask it to generate material for you to attempt, problems, cues, or a schedule, not finished answers.

Keep the hard part on your side: you retrieve, attempt, explain, or decide before you ask for help.

How it helps

"Explain it like I'm five" and "give me an analogy" are among the most-used AI prompts in education.

They often produce something memorable, vivid, and slightly wrong, and the memorability makes the wrongness stick.

An analogy that is never checked becomes the concept.

You leave with a story about pipes and water pressure and cannot write Ohm's law.

The analogy did its job as entertainment and failed its job as teaching.

Analogy generation is still worth doing, and with AI it gets weirdly fun: you can ask for five maps in thirty seconds, throw out the cute ones, and keep the one that makes the structure click.

The repair is in what happens after the analogy arrives: you must restate the target idea without the analogy, and you must know where the mapping holds.

Analogy stress-testing is the stress-test half; this article is the generation-and-restatement half.

How to run it

Name the target precisely. "Photosynthesis" invites a cartoon. "Why the light-dependent reactions must come before the Calvin cycle" invites a real map.
Ask for several analogies plus mapping tables. One story is a lottery ticket. Three is a shortlist.
Reject surface similarity. If the analogy only shares a vibe ("it's like a team working together"), demand another.
Pick the one that maps onto something you already know well. Your prior knowledge is the battery; the analogy is the cable.
Restate the target unaided. Close the analogy. Explain the real idea in three sentences with no metaphor. This is the learning event.
Flag one place the analogy lies. Even a good analogy misleads somewhere. Naming it prevents the story from becoming the concept.
Hand off to Analogy stress-testing when the stakes are high: exam definitions, mechanisms, anything where a cute wrong model will cost you.

Example exchange

You

Give me three structural analogies for why a feature branch can diverge from main and later merge.

Here are three. First, a bookmark in a choose-your-own-adventure book that you rewrite from mid-story. Second, a side-prep ticket on a kitchen rail while the main ticket keeps moving. Third, cousins on a family tree who share grandparents but grow apart.

You

Rank those for someone who cooks, not someone who codes.

The kitchen ticket fits you best. Before we keep it: close the metaphor. What is a branch, in git terms, with no kitchen words?

You

A branch is a movable pointer to a commit. Diverging means new commits on that line. Merge reconciles two histories into one.

Good. One place the kitchen picture lies: food gets consumed; commits don't. Want to name one more lie before we stop?

Copyable prompt

Give me THREE analogies for why adaptive immunity remembers a pathogen after
first exposure. For each: a two-column mapping table (analogy part matched to the real part).
Prefer structural similarity over vividness. Rank them by how well causal roles
match. If one is weak, say so.

The Tell

Here is how you know this method has flipped: you can recite the analogy fluently and freeze when asked for the definition, equation, or mechanism in the subject's own terms.

In that moment the AI (or the schedule, or the story) did the thinking, and you only recognised a finished product.

Tighten the prompt for analogy generation so you attempt, decide, or retrieve before anything is handed to you.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Analogies as learning tools are well studied in cognitive science (structure-mapping: Gentner; classroom analogy research). Benefits depend on structural alignment and on learners mapping relations, not surface features. Generating an analogy alone is a weaker move than stress-testing it (Analogy stress-testing).

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

Principle: moderate; AI-invented analogies often mislead, delivery speculative and high-risk without critique.