feat(taxonomy): #498 Fallacies AIF — P1 tranche-1f (7 SUFFIX-ONLY, famille Abus de langage, MIXTE, gated) - #781
Merged
Merged
Conversation
…FIX-ONLY, famille Abus de langage, MIXTE gated) P1 reconciliation back-fill, tranche-1f = 7 SUFFIX-ONLY rows (800/808/814/833/839/856/876), the whole "Abus de langage" family (Definition inexacte + Comparaison fallacieuse + Ambiguite). UNLIKE the inference-uniform families (1c/1d/1e all-undercut), this family is HETEROGENEOUS: definitional/classificatory manipulations attack the PREMISE (term application / exhaustivity / comparison-aptitude contestable) -> undermine/I-node (6 rows); only the analogy scheme attacks the INFERENCE rule -> undercut/RA-node (pk839). Distribution: 6 undermine / 1 undercut. 5/7 tokens novel; 2 with undermine precedent (BiasedClassification_Conflict-> pk799 anchors pk833; ConflictingGoals_Conflict->pk356 anchors pk876). attackType = NEW judgment; NODE deterministic (#707§4a); 0 token fabricated. 4 MED flags (800, 808, 856, 876) — honest cost of a heterogeneous family: - pk800 (Acception vague): vagueness = ill-defined premise -> undermine; alt undercut if read as equivocation-by-shifting. - pk808 (Sophisme des correlatifs): DISCONNECT skos/desc (token AlternativeMeans_Conflict suggests means-end undercutter; desc is definitional). desc-driven -> undermine. SAME TOKEN as tranche-1e pk33 (typed UNDERCUT, induction desc). Opposite verdicts justified by desc divergence (definitional vs inferential), NOT by token — flagged as the most delicate point for ai-01 (tests the desc-driven principle). - pk856 (Expression vague): same token as pk800 -> undermine; alt equivocation. - pk876 (Ambiguite narrative): ConflictingGoals precedent -> undermine; alt undercut if ambiguity read as equivocal inference. 3 HIGH (814 false dilemma, 833 fallacious comparison precedent, 839 false analogy scheme undercutter). Clean premise-vs-inference split documented (833 classification-biased vs 839 analogy-transfer). 0 prod write (gated dry-run). Baseline dynamic (write-driven, disjoint PKs): 114->121 (master c1ed77d) / 121->128 (if 1c written) / 125->132 (if 1c+1d) / 130->137 (if 1c+1d+1e). Byte-exact: 0 mismatch, 1409x104, CRLF+BOM, +105 B. - docs/taxonomy/498-reconciliation-p1f.md (proposition, MIXTE) - docs/taxonomy/498-reconciliation-p1f-annotations.csv (7 rows, 12 cols) - tools/498-p1f-apply.py (gated apply, dry-run default, mirrors p1e) Co-authored-by: Claude-Code <noreply@anthropic.com>
jsboige
added a commit
that referenced
this pull request
Jul 11, 2026
…16 SUFFIX-ONLY rows, attack 121->137) (#783) Batched prod write of 3 reviewed+GO'd propositions (master cbf9b7c, baseline 121 post-tranche-1c). ai-01 reviewed the 3 propositions (#778/#780/#781) and concur'd on every verdict including the same-token pk33/pk808 divergence (msg-ftqy51 [GO WRITE]). 16 disjoint PKs, 2 AIF columns cell-filled (empty -> value): - tranche-1d (Erreur mathematique, 4): 595/614/632/677 -> all undercut/RA-node - tranche-1e (Insuffisance, 5): 2/33/34/43/71 -> all undercut/RA-node - tranche-1f (Abus de langage, 7, MIXTE): 800/808/814/833/856/876 -> undermine/I-node + 839 undercut/RA-node Distribution this batch: 10 undercut / 6 undermine. attack layer 121 -> 137 (fully-modeled 46 -> 62). Full attack distribution now: 82 undercut / 50 undermine / 5 rebut. Byte-exact verified 2x: - self-proof in each apply script (0 mismatch, 1409x104, CRLF+BOM) - INDEPENDENT backup-vs-prod (tmp/Fallacies-backup-pre-p1def-batch.csv vs prod): exactly 16 PKs changed in ATI/ANI only = 32 cell changes, 0 stray, 0 value mismatch, pre-state of all 16 was empty (fill not overwrite). Only prod CSV in diff (16 ins/16 del). 0 token fabricated (TYPE rows already skos-vetted). Secret-scan: no credential in diff. Co-authored-by: Your <your.email@example.com> Co-authored-by: Claude-Code <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
#498 AIF — P1 reconciliation tranche-1f (7 SUFFIX-ONLY, famille « Abus de langage », MIXTE, gated)
Worker po-2024 · Base
c1ed77d2· Statut proposition gated (0 write prod CSV).Suite de la réconciliation P1. Tranches 1 + 1b + 1c (#779 pending) + 1d (#778 pending) + 1e (#780 pending) déjà livrées. Ce PR = tranche-1f = 7 SUFFIX-ONLY, famille « Abus de langage » complète.
Contrairement à 1c/1d/1e (toutes 100% undercut), Abus de langage a une distribution MIXTE 6 undermine / 1 undercut : les manipulations définitionnelles/classificatoires attaquent la prémisse (undermine/I) ; seul le scheme analogique attaque la règle d'inférence (undercut/RA, pk839).
Livré : 7 lignes (attackType = jugement neuf)
5/7 tokens novel ; 2 avec precedent undermine. Modélisation Walton cas-par-cas :
VagueVerbalClassification_InferenceAlternativeMeans_ConflictDilemma_InferenceBiasedClassification_Conflict+ExceptionSimilarityCase_ConflictDifferencesUndermineSimilarity_Conflict+Analogy_InferenceVagueVerbalClassification_InferenceConflictingGoals_ConflictDistribution 6 undermine / 1 undercut / 0 rebut. Prior de suffixe proscrit (pk804
_Conflict→undercut). Node déterministe (#707§4a).🔴 Point le plus délicat : divergence same-token pk808 / pk33
AlternativeMeans_Conflictapparaît 2× avec verdicts opposés :Verdicts opposés justifiés par la desc, pas par le token (mémoire
aif-no-inherit: token seul non-décisif). À vérifier par ai-01 : si le principe desc-driven est rejeté, pk33 et pk808 doivent être alignées (ajustement des 2 tranches).4 MED flags à trancher par ai-01 (800, 808, 856, 876)
Tous documentés avec alternative explicite :
3 HIGH (clean)
Fichiers (3, 0 write prod)
docs/taxonomy/498-reconciliation-p1f.md— proposition (§0 TL;DR, §2 modélisation Walton MIXTE, §3 divergence same-token, §5 baseline dynamique, §6 bornes).docs/taxonomy/498-reconciliation-p1f-annotations.csv— 7 lignes, 12 cols, BOM+CRLF, node/type 7/7.tools/498-p1f-apply.py— apply gated (dry-run par défaut), miroir498-p1e-apply.py.Dry-run (code=truth)
git diff c1ed77d2 -- Cards/…/Taxonomy.csv= vide (0 write prod).Handoff ai-01
tools/498-p1f-apply.py --write(baseline lue dynamiquement ; PK disjoints → ordre indifférent).Gate boundaries
Cards/intact (CSV diff vide) · ❌ [runtime pending] feat(dnn): #649 RulesExplorer i18n — mirror loc() culture cascade #674/[HOLD post-tag] fix(cardpen): #629 convert relative /Cards paths to raw-master URLs (Option 1) #666/chore(dnn): #131 migrate 12 RazorComponent templates to Razor14 (source-level, runtime pending) #596 (HOLD) · ❌ pas de self-merge — verdict QA ai-01.🤖 Worker po-2024 — réconciliation P1 tranche-1f (Abus de langage, MIXTE).