Elimination Decision Log¶
A register — instantiates Contrapositive Elimination Reasoning
Keeps an append-only record of every candidate ruled out — the rule used, the absent consequence, and the confidence — so the surviving set stays explicit and each elimination is auditable and reversible.
An investigation that eliminates candidates one by one needs a memory, or it will both re-litigate settled cuts and forget that some cuts were only tentative. Elimination Decision Log is that memory. It records each elimination as a dated, justified entry — the rule invoked, the consequence found absent, the evidence, and a confidence label — and it maintains the complementary residual set: what is still in play. It holds no logic and gathers no evidence; its entire value is persistence and reversibility. A confidence-tagged trail is what lets a wrongly-eliminated candidate be cleanly reinstated when a rule's confidence is later downgraded.
Example¶
An aircraft accident investigation works through failure hypotheses over months. The log accumulates entries: "Engine fire — ruled out; no thermal damage on recovered components and no fire-loop warning in the recorder (mandatory signature absent); confidence high." "Bird strike — ruled out; no organic residue on inlets, no relevant radar returns; confidence medium." "Fuel contamination — open." A running header shows the residual set — the causes still standing. Each entry names the rule, the missing consequence, and how firmly that rule holds.
Later, a re-examination of a component downgrades the "no thermal damage" finding — the earlier inspection missed a heat-affected zone. Because the engine-fire elimination was logged with its confidence and its basis, the team reinstates that hypothesis in one move instead of rediscovering it from scratch. The log is the disciplined version of the detective's maxim that once you have eliminated the impossible, whatever remains must be the truth — with the crucial caveat that an "impossible" logged at medium confidence is not yet impossible.[n1]
How it works¶
- Append, never overwrite — each elimination is a permanent entry; nothing is silently erased, so the reasoning history stays intact.
- Record the basis and the confidence — every entry carries the rule used, the absent consequence, and a confidence label, so a shaky elimination is visibly distinct from a solid one.
- Maintain the residual complement — the log always exposes the live set (what remains), computed as the candidates not yet eliminated.
- Support reinstatement — a confidence downgrade or new evidence on any logged elimination reopens that candidate cleanly, without disturbing the others.
Tuning parameters¶
- Confidence granularity — how finely each elimination's certainty is recorded. Finer grades tell you which cuts are safe to treat as final and which are only provisional.
- Reversal policy — what confidence downgrade or new evidence triggers reinstatement of a ruled-out candidate.
- Entry richness — how much justification each entry carries, from a one-line reason to a full evidence trail. Richer entries audit better but cost upkeep.
- Residual-set prominence — whether the surviving set is an always-visible computed header or reconstructed on demand.
When it helps, and when it misleads¶
Its strength is that it defeats both amnesia and lock-in: the team stops re-arguing settled eliminations, yet never forgets which eliminations were merely provisional. The confidence labels make premature closure visible — you can see that the field was narrowed by a run of medium-confidence cuts.
Its failure mode is that a tidy log lends false finality. An elimination logged "high confidence" on a fundamentally weak rule looks every bit as settled as a genuinely strong one, and the residual set can collapse to a single "answer" before the shakiest cuts have been retested. The classic misuse is closing the case on the last-remaining candidate without revisiting the low-confidence eliminations that narrowed the field to it. The discipline is to re-open the lowest-confidence entries before acting on whatever remains.
How it implements the components¶
Elimination Decision Log realizes the bookkeeping side of the archetype — the durable record of what has been cut and what remains:
residual_candidate_register— it maintains the live complement of the eliminated set: the candidates still in play at any moment.rule_confidence_label— every elimination entry carries the confidence of the rule that drove it, which is what makes reversal possible.
It does not perform the deduction (contrapositive_inference_step — that's Modus Tollens Checklist); it does not validate whether the negatives it records were reliable (detectability_and_scope_check, absence_evidence_threshold — that's Negative-Evidence Reliability Review); and it does not enumerate the candidate set at the outset (antecedent_candidate_set, necessary_consequence_signature — that's Required Consequence Table).
Related¶
- Instantiates: Contrapositive Elimination Reasoning — the log is the archetype's durable, reversible record of the elimination process.
- Consumes: Negative-Evidence Reliability Review validates the negatives that the log then records as justified eliminations.
- Sibling mechanisms: Negative-Evidence Reliability Review · Diagnostic Rule-Out Protocol · Rule-to-Observation Matrix · Required Consequence Table · Modus Tollens Checklist · Search-Branch Pruning Test
Editorial Notes¶
Form Classification¶
Form family: Record, Log & Register
Rationale: Elimination Decision Log operates as a durable record, ledger, register, or trace whose value depends on preserving actual state or history because it keeps an append-only record of every candidate ruled out — the rule used, the absent consequence, and the confidence — so the surviving set stays explicit and each elimination is auditable and reversible.
Independent corroboration: The frozen evidence defines Elimination Decision Log as 'Keeps an append-only record of every candidate ruled out — the rule used, the absent consequence, and the confidence — so the surviving set stays explicit and each elimination is auditable and reversible', so its operative form is Record, Log & Register.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Security Studies & Intelligence Analysis
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Structured intelligence analysis cohered explicit candidate sets, confidence-tagged eliminations, and audit trails that preserve residual hypotheses and permit reinstatement.
Related originating lineages:
- Medicine & Healthcare — Differential diagnosis independently eliminates candidate causes while retaining uncertain alternatives.
- Philosophy — Formal logic supplies the contrapositive warrant that absence of a necessary consequence can eliminate a hypothesis.
Review resolution: The current reviewers agree that security_intelligence is primary. For the reported differences (reported_ambiguity, alternate_origin_disagreement), the evidence supports convergent, multi_domain, and medicine_healthcare, philosophy; these choices preserve materially formative origins without conflating later domain reach.
Attribution caveat: The persistent log is a synthesis of a formal inference with investigative and diagnostic practice.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
The log is often confused with Rule-to-Observation Matrix, but they occupy different moments. The matrix is the live working grid where eliminations are worked out; the log is the durable, append-only decision record they are committed to, built for audit and reversal rather than for computing the next cut.
[n1] The detective's dictum — "when you have eliminated the impossible, whatever remains, however improbable, must be the truth" (Conan Doyle's Sherlock Holmes) — is the residual-register idea in one line. The log's contribution is to keep it honest: a candidate eliminated on a low-confidence rule is not yet "impossible," and the residual set should not be treated as the answer until the weak cuts have held up. ↩