Skip to content

Qualifier-Drop Scan

Audit — instantiates Summary-Substance Alignment Audit

Inventories the hedges, scope limits, and conditions in the substance that fix when a claim is true, then flags the material ones the summary silently dropped.

A Qualifier-Drop Scan works omission-first. Instead of checking what the summary asserts, it inventories the qualifiers in the substance — the hedges, scope limits, conditions, exceptions, and eligibility or dosing caveats that fix when a claim is actually true — and then checks which of the material ones the summary silently dropped. Its defining move is to read the body for what must travel with a claim and treat the summary as at fault for any material qualifier missing from it. It says nothing about claims the summary asserted well or added; it hunts only the caveats that fell off in compression, because those are individually deniable — every remaining word is still true.

Example

A cardiology society condenses a 90-page guideline into a one-page pocket card for clinicians. The Qualifier-Drop Scan first indexes the guideline's qualifiers around one drug recommendation: "for patients under 75," "not with stage-4 kidney disease," "start at half dose if on a diuretic," "reassess at four weeks." It then reads the card, which says only "First-line: [drug], titrate to target." Applying its materiality rule — a qualifier is material if acting without it would change what a clinician does — the scan flags three dropped qualifiers as material (the age limit, the renal contraindication, the half-dose rule) and lets "reassess at four weeks" pass as non-material for a quick-reference card. Output: a short register naming the three omissions, ranked by the harm of acting without them, handed back so the card can reinstate them.

How it works

The scan runs in four moves that no wording comparison performs: (1) build the body's qualification index — every hedge, scope limit, and condition, catalogued; (2) apply a materiality rule that separates the qualifiers which must survive compression from those a summary may shed; (3) diff the material set against the summary; (4) log the material omissions in a risk register, ranked by consequence. Because it starts from the substance and looks for absence, it catches the distortion that a summary-side read structurally cannot — the qualifier that was never there to notice.

Tuning parameters

  • Materiality bar — how large an action- or belief-change a qualifier must produce to count as material. Set low, everything flags and the register is noise; set high, real caveats slip.
  • Qualifier taxonomy — which kinds to hunt (scope, temporal, population, confidence, legal). Narrowing speeds the scan but risks a blind spot for the type you excluded.
  • Audience-relative materiality — the same caveat can be material on a clinician card and immaterial on a patient leaflet; the bar is set per audience, not absolutely.
  • Recall vs. precision — flag every candidate omission (catches more, buries the reviewer) or only high-confidence ones (cleaner, misses edge cases).
  • Register threshold — how many or how severe the omissions must be before the summary is blocked rather than merely annotated.

When it helps, and when it misleads

Its strength is catching the most common and most deniable distortion — omission — which no word-to-word comparison catches, because each surviving word is individually true. Its failure modes: over-flagging, where treating every hedge as material buries the two that matter; and the fact that a summary must compress, so a scan tuned too strictly declares an impossible standard and gets ignored. The classic misuse is running it after publication to bless a summary already out — declaring the dropped qualifiers "immaterial" once they can no longer be reinstated. The discipline is to fix the materiality rule and audience profile before scanning, and to tie materiality to a change in action or belief — the legal and accounting test of a material omission — not to the author's comfort.[1]

How it implements the components

Qualifier-Drop Scan fills the omission-detection side of the archetype:

  • body_qualification_index — enumerates every hedge, scope limit, and condition in the substance; the scan's raw input.
  • material_qualifier_rule — the test that separates the qualifiers which must survive compression from those a summary may legitimately shed.
  • summary_surface_risk_register — the output: the ranked list of material qualifiers the summary dropped, with the harm of acting without each.

It does not extract or grade the summary's positive claims — that is Press-Release Claim Review and Certainty & Causality Inflation Check — nor test the omissions on live readers, which is Summary-Only Reader Test.

Notes

The scan presumes compression is legitimate: its target is not zero omission but zero material omission. That boundary is what keeps it usable — a summary that carried every qualifier would just be the body again. It pairs naturally with Executive-Summary Caveat Budget, which allocates how many caveats a summary may carry up front, where this scan detects, after the fact, which material ones a given summary lost.

References

[1] Materiality — the standard, borrowed from law and accounting, that an omission or misstatement matters if it would change the judgment of a reasonable decision-maker. It is what turns "a qualifier is missing" into "a qualifier that should not be missing." ↩