Case Library¶
Case repository — instantiates Archetype Pattern Indexing
Indexes concrete, cited precedents by their case features so a new situation retrieves the closest prior case rather than an abstract rule.
A Case Library indexes recurring patterns through their concrete precedents: the unit of retrieval is a whole, cited case, not a distilled signature. Its defining move is reasoning by precedent — you find the pattern by finding the prior case most like the one in front of you, then transfer from it, keeping the original in view so you can judge how far the analogy actually carries. Where an abstract catalog answers "what is the general rule here," a case library answers "what is the most similar thing we have seen before, and what happened." The cases stay primary and their provenance stays attached, so a user can always inspect the source rather than trusting a summary.
Example¶
A litigation team maintains a case library of employment-dispute precedents. Each entry is an actual decided case, tagged by its features: jurisdiction, whether the worker was classified as a contractor, the presence of a written arbitration clause, the remedy awarded, and the governing statute. An associate handed a new misclassification matter starts not from a treatise but from the facts — a gig-economy driver, an arbitration clause, a state with a strict ABC test — and queries the library on those features. It returns three close precedents, each linked straight to the published opinion and docket so nothing rides on the paraphrase. Two of the three turned on whether the clause was signed at onboarding; that becomes the associate's first question to the client. The library did not tell her the rule of law; it surfaced the nearest decided cases, cited to source, so she could reason from precedent to her situation — the ordinary logic of stare decisis[1] made searchable.
How it works¶
The atom is the case, recorded richly enough that its features can be retrieved on. Each entry carries: the concrete facts (the exemplar itself), a link to its authoritative source, and a set of indexed features — the case descriptors a new situation is matched against. Retrieval is similarity-first: given the features of a live situation, the library ranks stored cases by how closely they match and returns the nearest with their outcomes. What distinguishes it from an abstract catalog is that it does not commit to a single generalized signature; it preserves the messy particulars and lets the user do the generalizing, case by case, which is exactly what makes it robust when the situation is novel or the abstraction is contested.
Tuning parameters¶
- Feature granularity — how many descriptors each case is indexed on. Rich features sharpen similarity matching but make consistent tagging expensive and cases harder to compare.
- Similarity metric — which features dominate the "nearest case" ranking, and how they are weighted. Weighting the wrong feature retrieves superficially similar but structurally irrelevant precedents.
- Corpus curation bar — whether every case is admitted or only landmark ones. A complete corpus captures edge cases; a curated one is faster to search but can bury an unusual-but-decisive precedent.
- Provenance depth — how much of the original source travels with each entry, from a bare citation to the full document. Deeper provenance guards against paraphrase drift at the cost of storage and upkeep.
When it helps, and when it misleads¶
Its strength is concreteness: precedents are memorable, defensible, and honest about particulars, and similarity retrieval works even where no one has yet agreed on the general rule — which is why case-based reasoning outperforms rule-based systems in domains that resist tidy abstraction.
Its failure mode is superficial similarity — retrieving the case that looks alike on salient surface features while differing on the structural feature that actually decided it (the arbitration clause, not the industry). A library that indexes on vivid-but-irrelevant descriptors will confidently return the wrong precedent. The classic misuse is treating a single retrieved case as authority rather than as an analogy to be tested. The guarding discipline is to weight the decision-relevant features in the similarity metric and to keep each case linked to its source, so a user can read the original and check whether the resemblance is more than skin-deep before relying on it.
How it implements the components¶
example_set— the cases are the library; each entry is a concrete, richly recorded precedent rather than an illustration of a separate abstract rule.retrieval_index— cases are indexed on their descriptive features so a live situation retrieves the nearest prior cases by similarity, not by remembered name.source_trace— every entry links to its authoritative origin (opinion, docket, report), so a user reasons from the record and not from a paraphrase.
It distills no pattern_signature and issues no response_guidance: a case library hands you the nearest precedent and leaves the generalizing to you — the abstracted signature belongs to Design Pattern Catalog and the prescribed next move to System Archetype Index.
Related¶
- Instantiates: Archetype Pattern Indexing — it supplies the precedent-anchored side of the index, where retrieval starts from concrete cases rather than named abstractions.
- Consumes: Tagging Schema provides the controlled feature vocabulary its cases are indexed on.
- Sibling mechanisms: Design Pattern Catalog · Anti-Pattern Catalog · Diagnostic Atlas · System Archetype Index · Solution Archetype Archive · Pattern Card Template · Tagging Schema · Pattern Library (which collects approved exemplar solutions to reuse, where this collects decided cases to reason from).
Editorial Notes¶
Form Classification¶
Form family: Record, Log & Register
Rationale: The mechanism preserves concrete sourced precedents with indexed facts and outcomes so current situations can retrieve the closest prior case, making it a case record repository.
Nearest alternative: Representation, Specification & Plan — Indexing represents case features, but the library's substance is accumulated historical precedents rather than a prospective model.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Law & Governance
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Common-law practice built indexed bodies of precedents whose fact patterns and holdings are retrieved to reason about a new matter.
Related originating lineages:
- Computer Science & Software Engineering — Case-based reasoning formalized feature-indexed retrieval and adaptation of similar prior cases.
- Library & Information Science — Indexing, citation, controlled description, and retrieval make the precedent collection usable as a library.
Review resolution: Law is primary because precedent repositories organized prior cases for retrieval and analogy long before computational case libraries. Computer science contributes case-based-reasoning retrieval, and library science contributes indexing and controlled access, yielding convergent multi-domain provenance.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Case-based reasoning (Roger Schank; Janet Kolodner, 1993) solves a new problem by retrieving and adapting the most similar past case rather than applying general rules — the computational formalization of the common-law habit of arguing from precedent (stare decisis). registry ↩