Coarse Screening Checklist¶
Checklist — instantiates Cascaded Hierarchical Recognition
Gives human reviewers a short high-recall set of initial cues before deeper review.
Coarse Screening Checklist is a short, fixed list of cheap cues a human runs first, before any deep inspection, deliberately tuned so that almost no plausible case is dropped. Its defining idea is that passing the checklist means only "worth a closer look," never "confirmed" — it is a human-operated coarse gate optimized for recall, not a decision. Where a full review asks "is this the right one?", the checklist asks the much cheaper question "could this plausibly be one at all?", and errs consistently toward keeping candidates alive. Its whole value is replacing a reviewer's idiosyncratic, mood-dependent first impression with a repeatable high-recall screen that every reviewer applies the same way.
Example¶
A recruiter faces 600 applications for a specialized instrumentation-engineer role and cannot deep-read them all. Before any careful review, they run a five-item Coarse Screening Checklist, each item written broadly on purpose: does the résumé mention any form of the required instrumentation background (including adjacent terms and older tool names), any relevant regulated-industry exposure, any hands-on rather than purely managerial signal, and so on. The bar for each item is "some plausible signal," not "strong match" — a candidate who trips a single item survives to real review. The screen cuts the 600 to roughly 180 without a recruiter ever having to make a judgment call about fit; that judgment is deferred to the deeper stages. Crucially, the checklist is written so that a strong-but-oddly-worded candidate is not silently lost at step one.
How it works¶
- Fix the target set first. Name the roles or classes the screen must never drop; the checklist exists to protect them.
- Draw items from a domain dictionary. Each item is a named coarse cue with its accepted variants, spellings, and synonyms, so breadth is explicit rather than left to the reviewer.
- Order cheapest cues first and set each item's bar to "any plausible signal."
- Pass on any hit. Survivors go to deeper review; the checklist itself decides nothing final.
Tuning parameters¶
- Recall target — how loose each item is; looser keeps more true cases but forwards more noise to expensive later stages.
- Number of items — a longer list screens harder but slows the screen and tempts reviewers to treat a miss as a rejection.
- Cue breadth — how many variants and synonyms each item admits; wider catches oddly-worded true cases but lets more irrelevant ones through.
- Reviewer training — how much interpretation each item demands; simpler items are more consistent across reviewers.
- Scoring rule — stop-after-first-fail versus tally-and-forward; tallying preserves borderline cases better.
When it helps, and when it misleads¶
Its strength is consistency and speed: it stops early rejection from depending on which reviewer opened the file, and it keeps the true case alive for the discriminating stages that follow. It is the human analogue of a high-sensitivity screen — sized to catch nearly all true cases at the cost of forwarding false alarms.[1]
Its failure mode is quiet role-creep: a checklist meant to screen in gets treated as a decision, and "failed the checklist" hardens into "rejected" with no one sampling the discard pile. The classic misuse is tightening the items to cut downstream volume — trading away the recall that was the checklist's entire reason for existing. The guarding discipline is to keep the screen deliberately loose, spot-check a handful of screened-out cases as an informal reality test, and never let a checklist miss stand as a final reject without a real guardrail behind it.
How it implements the components¶
coarse_gate— the checklist is a coarse gate operated by a human: broad, cheap, high-recall, removing only the obviously irrelevant.recognition_target_set— it is built from an explicit list of the classes it must not drop, which is what its recall is measured against.domain_feature_dictionary— each checklist item is a dictionary entry: a named coarse cue together with its accepted variants and synonyms.
It deliberately stops short of the fine test that separates close candidates (discriminative_feature_test — that's feature_pyramid_or_hierarchical_model) and it does not systematically sample its own rejects (false_negative_guardrail — that's cascade_error_audit); the spot-check above is an informal self-check, not that guardrail.
Related¶
- Instantiates: Cascaded Hierarchical Recognition — the human-run first gate that keeps plausible candidates alive.
- Sibling mechanisms: Cascade Error Audit · Drilldown Recognition Interface · Expert Review Checkpoint · Feature Pyramid or Hierarchical Model · Multi-Stage Classifier Pipeline · Salience Map or Attention Heatmap · Stage Transition Log · Triage Queue with Escalation Rules
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Gives human reviewers a short high-recall set of initial cues before deeper review, making its operative form a bounded evaluation of existing evidence or work that produces a finding or disposition.
Independent corroboration: The frozen evidence defines Coarse Screening Checklist as 'Gives human reviewers a short high-recall set of initial cues before deeper review', so its operative form is Assessment, Review & Assurance.
Nearest alternative: Decision, Gate & Allocation — Its short cues provide a high-recall initial assessment whose survivors are routed onward, rather than making the final allocation itself.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Medicine & Healthcare
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Clinical screening established short, high-recall first-pass checklists that retain plausible cases for deeper assessment.
Related originating lineages:
- Aviation & Aeronautics — Operational safety screening contributes standardized human first-pass cues.
Review resolution: Both reviewers agree on medicine_healthcare as primary. Reading the mechanism confirms that its defining operation belongs to that lineage; the final record retains aviation_aeronautics only as materially formative origin and keeps present-day application breadth separate from provenance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Wilson, J. M. G., & Jungner, G. Principles and Practice of Screening for Disease. Public Health Papers No. 34, World Health Organization (1968). Wilson and Jungner describe a high-sensitivity screen as catching nearly all true cases while accepting more false positives. They do not identify the mechanism's checklist as a human analogue of such a screen. registry ↩