Analogical Reasoning

Analogical Reasoning

Transfer structure from a well-understood source domain to a target problem, then explicitly verify the mapping holds where it matters.

Why

Analogy is the fastest way to import a solved problem's machinery into a new one — and, unverified, the fastest way to import wrong conclusions when surface similarity masks a real structural difference. The verification step is what turns analogy from rhetoric into reasoning.

Use when / Don't use when

• Use when: the problem resembles something solved elsewhere; generating approaches for a genuinely novel situation; evaluating someone else's persuasive analogy before accepting its conclusion.
• Don't use when: direct analysis of the target problem is cheap and available — analogy is a bridge to insight, not a destination in itself.

Inputs → Outputs

• Inputs: a target problem plus one or more candidate source domains.
• Outputs: a structure mapping, the transferred insight, and a list of disanalogies with their consequences.

Principles

• Map relations, not surface features — "both involve networks" is a surface match, not a structural one.
• An analogy is an argument only where the mapped structure is causally relevant to the conclusion being drawn.
• Every analogy breaks somewhere; find where before relying on it for anything consequential.

Procedure

1. State the target problem's essential structure — its entities, relations, and constraints.
2. Search for source domains sharing that relational structure, not just surface resemblance; list 2–3 candidates.
3. For the best candidate, build the explicit mapping: what in the source corresponds to what in the target.
4. Transfer the source's solution or lesson through that mapping.
5. Hunt for disanalogies: relations present in the target with no source counterpart, and vice versa.
6. Judge whether any disanalogy touches the causal path of the transferred lesson — if so, discard or repair the transfer.
7. Report the insight plus the specific boundary where the analogy stops working as the residual uncertainty.

Common mistakes

• Matching on surface features instead of relational structure.
• Riding the analogy past its breaking point instead of stopping at the boundary found in step 6.
• Adopting the first analogy that comes to mind instead of comparing several candidates.

Examples

• Using epidemiology's spread models to reason about how misinformation propagates through a network.
• Treating technical debt with financial-debt intuitions, then explicitly finding where the analogy breaks (technical debt doesn't compound at a fixed, knowable rate).
• Porting immune-system self/non-self tolerance concepts to content moderation policy design.

Related

• [[explanatory-analogy]] — the same machinery, aimed at teaching rather than problem-solving.
• [[concept-blending]] — a stronger move merging two frames into a new space rather than just transferring one.
• [[mental-model-extraction]] — often supplies the source domain's structure.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: skyf0xx
• Source: skyf0xx/better-thinking
• License: MIT
• Homepage: https://zenbin.org/p/better-thinking

Install and usage instructions live in the source repository linked above.