Negative-Case Conservation¶
Evidence retention — instantiates Counterexample Boundary-Shift Audit
Keeps every disconfirming case on a durable ledger and logs each boundary change against the cases it would drop, so counterexamples can't be quietly deleted.
Negative-Case Conservation is the archive that makes the boundary-shift move un-erasable. The dodge's quiet accomplice is memory loss: once a counterexample is redefined out of the category, it tends to vanish from the discussion, and with it the evidence that the boundary ever moved. Conservation refuses that erasure. It maintains a durable ledger of every disconfirming case — each one entered once and never deleted, only annotated — and it logs every boundary change against the specific cases that change would exclude. The result is a running tally the audit can always consult: not just "here is the current definition," but "here are the seven cases that were dropped to reach it, and when each was dropped." Its defining trait is retention, not judgment — it decides nothing, but it guarantees that whatever the audit decides, it decides in full view of every case it is tempted to forget.
Example¶
A pharmaceutical safety review tracks the claim: "This drug has no serious cardiac side effects." Over eighteen months, adverse-event reports arrive: a 58-year-old with arrhythmia, a hospitalization for tachycardia, two more. Each time, there is pressure to reclassify — "that patient had an undisclosed pre-existing condition," "that one's dose was off-label," "borderline, probably unrelated." Conservation enters every one of these cases on a permanent ledger the moment it appears, tagged with its raw facts before any reclassification argument is applied.
When the safety definition is subsequently narrowed — "serious cardiac side effect" quietly redefined to require hospitalization and no confounding condition — conservation logs that boundary change against the four cases it would now exclude, with dates. Six months on, when the team reports "still no serious cardiac events," anyone can pull the ledger and see the four conserved cases and the two definitional narrowings that made them disappear. Conservation renders no verdict on causation; it simply ensures the negative cases and the boundary moves are both still on the books, side by side, where a reviewer or regulator can weigh them. This is the file-drawer problem[n1] defeated by refusing to let the drawer close.
How it works¶
- Enter each disconfirming case once, permanently — a write-only ledger; entries are annotated, never removed or overwritten.
- Snapshot raw facts at intake — record the case as it stood before any reclassification argument, so later reframing is visible as an annotation rather than a silent edit.
- Log every boundary change against affected cases — when a definition narrows, record which conserved cases it newly excludes and when.
- Expose the running tally — the current claim always carries its trailing list of dropped cases and the diffs that dropped them.
Tuning parameters¶
- Retention scope — every raised case, or only those passing a minimal relevance bar. Broad retention is safest against erasure but accumulates noise.
- Annotation vs. deletion policy — strictly append-only, or allowing supervised redaction. Append-only maximizes auditability at the cost of a growing ledger.
- Diff granularity — whether each boundary change is logged per-case or in bulk. Per-case is more informative and more expensive.
- Surfacing cadence — how prominently the trailing dropped-case list travels with the claim: always attached, or available on request.
When it helps, and when it misleads¶
Its strength is memory: it makes the cumulative pattern visible — the difference between one honest reclassification and a claim propped up by ten of them — which no single-case check can see. It is the mechanism that turns "repeated boundary contraction" from an invisible drift into a countable record.
Its failure mode is ledger overload and false equivalence: a conservation archive that retains everything without weighting can bury a decisive counterexample among trivia, or lend a long-refuted case the same standing as a live one. The classic misuse is citing raw ledger length ("look how many exceptions!") as if quantity alone refutes a claim, ignoring that some entries were legitimately resolved. The guarding discipline is to keep the ledger append-only and annotated with each case's disposition, so the record is complete without being undifferentiated.
How it implements the components¶
Conservation fills the archetype's retention slots:
counterexample_candidate_record— it holds the durable, write-only record of every disconfirming case, snapshotted at intake.boundary_change_diff_log— it logs each boundary change against the specific cases that change excludes, building the cumulative trail.
It does not run the circularity-and-timing analysis that judges a shift illegitimate (circularity_and_timing_test) — that is its nearest twin Ad Hoc Boundary-Shift Probe, which reads the log to fire a flag; conservation only preserves the cases and diffs, whereas the probe evaluates them. Nor does it label a change legitimate or record the revised claim (legitimate_scope_revision_rule) — that is Scope-Revision Memo.
Related¶
- Instantiates: Counterexample Boundary-Shift Audit — conservation is the audit's durable memory of dropped cases and boundary moves.
- Sibling mechanisms: Ad Hoc Boundary-Shift Probe · Category-Predicate Separation · Claim Scope Freeze · Counterexample Admissibility Test · Independent-Criterion Challenge · Scope-Revision Memo · Symmetric-Case Application · True-Member Language Flag
Editorial Notes¶
Form Classification¶
Form family: Record, Log & Register
Rationale: Negative-Case Conservation operates as a persistent ledger, log, register, or case record that preserves history and traceability because it keeps every disconfirming case on a durable ledger and logs each boundary change against the cases it would drop, so counterexamples can't be quietly deleted.
Independent corroboration: The frozen evidence defines Negative-Case Conservation as 'Keeps every disconfirming case on a durable ledger and logs each boundary change against the cases it would drop, so counterexamples can't be quietly deleted', so its operative form is Record, Log & Register.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: Refusing to discard counterexamples and making theory changes answer to them follows falsificationist philosophy of science.
Related originating lineages:
- Ethnography & Qualitative Methods — Negative-case analysis supplies the durable handling of disconfirming cases during theory revision.
- History & Historiography — Archival provenance and revision history contribute the append-only evidence ledger.
- Medicine & Healthcare — Drug-safety surveillance and adverse-event review supplied the durable case-ledger practice and the high stakes of reclassification.
- Statistics & Experimental Design — The mechanism responds to publication and selection bias by preserving disconfirming observations before later boundary changes can erase them.
Review resolution: Authoritative-source research resolves the primary-origin disagreement. The obligation to preserve potential counterinstances is rooted in falsificationist philosophy of science; append-only conservation is a synthesis with audit-trail and bias-control practices. Origin breadth is limited to formative lineages; present-day applicability is recorded separately as domain_reach=universal.
Attribution caveat: The conservation label and append-only formulation appear synthesized from open-science and safety-record traditions. The conservation ledger is encyclopedia-authored; its normative basis is philosophical while its practice resembles qualitative audit trails.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; medium confidence.
Sources consulted:
Notes¶
[n1] The file-drawer problem (Robert Rosenthal) — disconfirming or null results disappear into a drawer, biasing the surviving record toward confirmation. Conservation is the direct antidote: a write-only ledger keeps the negative cases where they cannot be filed away. ↩