Case Library Capture¶

Artifact — instantiates Tacit Knowledge Elicitation

Builds a searchable collection of real cases — each with its cues, the judgment made, the outcome, and its boundaries — so expertise transfers through contrast between examples rather than through a single rule.

Much tacit knowledge resists compression into a rule because the rule's real content lives in the exceptions — this looked like a fraud but wasn't, that looked routine but wasn't. Case Library Capture is the artifact that holds expertise as a curated set of concrete cases instead of a distilled rule: each entry records the situation, the cues that were present, the judgment the expert made, what actually happened, and where the case's lesson stops applying. Its defining move is teaching by contrast — the library's value comes from placing near-identical cases with opposite correct answers side by side, so a learner absorbs the distinction the way experts did, by accumulating examples and counterexamples rather than by memorizing a heuristic that erases the very nuance that matters.

Example¶

A security operations center keeps a case library of past alerts. One entry: an after-hours login from an unusual country, flagged, investigated, and confirmed benign — an engineer on holiday. A neighboring entry looks almost identical — after-hours, unusual geography — but was a real account takeover. Stored beside each other, the two cases carry what no single detection rule could: the cue that separates them (in the malicious case, the login was preceded by a password reset from a different device an hour earlier), and the boundary note that this discriminator only holds for accounts with self-service reset enabled.

A new analyst working a fresh alert doesn't read a rulebook; they pull the two contrasting cases and see immediately which pattern the live alert rhymes with. Over a year the library accumulates the hard, ambiguous, and near-miss cases — precisely the ones a policy document smooths over — and becomes the substrate a team reasons from. What the library does not do is elicit any of this: it is the shelf, not the interview. Each case arrives already surfaced by a debrief, an incident review, or a probe.

How it works¶

Store cases, not rules. The unit is a real (or realistic) episode with its full texture — cues, context, the call made, the outcome — not a generalization abstracted away from it.
Curate for contrast. The highest-value additions are pairs and clusters that look alike but resolve differently; the library is organized so a user can retrieve the near-neighbors of a situation in front of them.
Privilege the hard cases. Successes, failures, near misses, and genuine edge cases carry more discriminating information than routine ones and earn their place first.
Annotate the boundary on each entry. Every case records where its lesson transfers and where it doesn't — the tool, population, or condition under which the same cues would mean something else.

Tuning parameters¶

Case granularity — thin index cards or richly documented dossiers. Richer cases teach more but cost more to capture and to read, and can bury the discriminating cue in detail.
Curation vs. accretion — a tightly curated exemplar set or an accreting archive of everything. Curation keeps signal high; accretion captures more but risks becoming an unsearchable dump.
Ordinary-to-dramatic mix — how many rare critical incidents versus routine cases. Dramatic cases are memorable but skew the sample; a healthy base of ordinary cases keeps the library calibrated.
Retrieval structure — tags, cues, outcomes, or similarity. The indexing decides whether a user in the moment can actually surface the case they need.

When it helps, and when it misleads¶

Its strength is holding knowledge in the one form that survives irreducible nuance: contrasting concrete cases, which transfer the discriminating distinctions that any single rule would flatten.^[1] It is durable, inspectable, and grows richer as practice throws off new hard cases.

Its failure modes come from the library drifting away from live practice. It can rot — cases age, tools change, and a discriminator that once held quietly stops holding, so a stale library confidently teaches yesterday's world. A dramatic-incident bias can make the collection an anthology of war stories that over-weights the rare and under-teaches the routine. And a library is only as good as its retrieval: a case nobody can find in the moment they need it might as well not exist. The discipline that keeps it useful is aggressive curation over accretion, a deliberate base of ordinary cases against the dramatic ones, per-case boundary annotations that flag when a lesson has expired, and treating the library as a provisional representation kept fresh by the mechanisms that feed it — never as a finished rulebook.

How it implements the components¶

Case Library Capture is a storage-and-representation artifact; it fills the components that concern holding elicited knowledge, not producing it:

provisional_articulation — turns elicited know-how into a shareable form: a browsable, contrast-organized collection of cases with cues, rationale, and outcome, explicitly held as provisional rather than as settled rules.
transfer_boundary_annotation — each entry carries its own boundary note, marking the conditions under which its lesson does and does not carry.
critical_incident_case — the library is where high-information successes, failures, and near misses are stored and curated as reusable exemplars.

It does not elicit any of this. The cues (cue_elicitation), the reasoning (decision_rationale_probe), and the surfacing of the incidents themselves (critical_incident_case as an elicitation act) come from mechanisms like Critical Incident Technique, Cognitive Interview, and Expert Debrief; the library only receives and organizes their output. The distilled point-of-work aid is Judgment Aid.

Instantiates: Tacit Knowledge Elicitation — Case Library Capture is the durable artifact where the archetype's elicited knowledge is held and made reusable.
Consumes: Critical Incident Technique, Expert Debrief, and other eliciting mechanisms supply the cases it stores.
Sibling mechanisms: Judgment Aid · Critical Incident Technique · Expert Debrief · Cognitive Interview · Cognitive Task Analysis · Decision Trace Review · Simulation Replay · Think-Aloud Protocol · Shadowing Session · Apprenticeship Observation · Task Analysis

Notes¶

Case Library Capture and Judgment Aid are complementary representations, not rivals. The library is the back-end corpus — comprehensive, for study and contrast; the judgment aid is the front-line distillation — a few cues at the point of work. A common pattern is to derive judgment aids from the library and let both share the same boundary annotations, so the shelf stays authoritative and the aid stays usable.

References¶

[1] The idea that reasoning proceeds by retrieving and adapting past cases rather than applying general rules is the premise of case-based reasoning (Schank, Kolodner). It is why a library organized for contrast between concrete episodes can carry discriminations that no compact rule preserves — and why the near-miss and counterexample cases are the ones worth capturing first. ↩