Skip to content

feat(taxonomy): #498 Fallacies AIF — P1 tranche-1g (8 SUFFIX-ONLY, famille Tricherie, MIXTE, gated, LAST FAMILY) - #784

Merged
jsboige merged 1 commit into
masterfrom
feat/498-p1g-tranche
Jul 11, 2026
Merged

feat(taxonomy): #498 Fallacies AIF — P1 tranche-1g (8 SUFFIX-ONLY, famille Tricherie, MIXTE, gated, LAST FAMILY)#784
jsboige merged 1 commit into
masterfrom
feat/498-p1g-tranche

Conversation

@jsboige

@jsboige jsboige commented Jul 11, 2026

Copy link
Copy Markdown
Contributor

#498 AIF — P1 reconciliation tranche-1g (8 SUFFIX-ONLY, famille « Tricherie », MIXTE, gated) — 🏁 DERNIÈRE FAMILLE

Worker po-2024 · Base cbf9b7c8 · Statut proposition gated (0 write prod CSV).

🎯 Dernière famille SUFFIX-ONLY → write 1g clôturera les 36-row back-fill (5 1b + 7 1c + 4 1d + 5 1e + 7 1f + 8 1g = 36). Suite des tranches 1 + 1b + 1c (prod) + 1d/1e/1f (write batché #783 pending).

Livré : 8 lignes (attackType = jugement neuf)

6/8 tokens ont un precedent typé ; 2/8 novel. Distribution MIXTE 3 undermine / 5 undercut — trois natures de défaut :

pk sophisme scheme → type conf precedent
888 Présentation trompeuse des faits LackOfPTKReliability_Scheme+PositionToKnow_Inference undermine/I MED ⚠️ same-token pk2→undercut ; PositionToKnow pk70
973 Déplacement des critères ConflictingGoals_Conflict undermine/I HIGH ConflictingGoals pk356
1020 Sophisme des coûts irrécupérables Waste_Inference undercut/RA HIGH Waste_Inference pk432
1023 Raisonnement biaisé Bias_Inference+PresumptiveInference_Scheme undermine/I HIGH Bias+Presumptive pk70 ×2
1066 Effet de simple exposition Example_Inference undercut/RA MED token novel ; parallèle pk34/pk595
1087 Preuve anecdotique InductiveInference_Scheme undercut/RA HIGH desc TWIN pk34
1148 Justification de l'effort Waste_Inference undercut/RA HIGH Waste_Inference pk432
1175 Biais d'attribution Example_Inference undercut/RA MED token novel ; parallèle pk34/pk595

Prior de suffixe proscrit (pk804 _Conflict→undercut). Node déterministe (#707§4a).

🔴 Point délicat : 2ᵉ divergence same-token pk888 / pk2

LackOfPTKReliability_Scheme :

  • tranche-1e pk2 (Généralisation hâtive) → undercut (desc inférentielle d'induction).
  • tranche-1g pk888 (Présentation trompeuse) → undermine (desc déférentielle = arrangement trompeur = prémisse biaisée), renforcée par le 2ᵉ token PositionToKnow_Inference→undermine (pk70).

2ᵉ occurrence après 1f pk808/pk33 — renforce le principe desc-driven. À vérifier par ai-01 (si rejeté, pk2/pk888 à aligner).

3 MED flags (888, 1066, 1175) + 5 HIGH

  • 888 — same-token divergence (voir ci-dessus).
  • 1066 mere exposure — undercut (familier→acceptable règle défectueuse) ; alt undermine (prémisse surévaluée).
  • 1175 biais attribution — undercut (attribution over-systematic) ; alt undermine (prémisse causale fausse).
  • HIGH : 973 (ConflictingGoals pk356), 1020 (Waste pk432 sunk cost), 1023 (Bias+Presumptive pk70 ×2), 1087 (desc = pk34), 1148 (Waste pk432 // 1020).

Fichiers (3, 0 write prod)

  • docs/taxonomy/498-reconciliation-p1g.md — proposition (§0 TL;DR, §2 modélisation Walton MIXTE, §3 divergence same-token, §5 baseline dynamique, §6 bornes).
  • docs/taxonomy/498-reconciliation-p1g-annotations.csv — 8 lignes, 12 cols, BOM+CRLF, node/type 8/8.
  • tools/498-p1g-apply.py — apply gated (dry-run par défaut), miroir 498-p1f-apply.py.

Dry-run (code=truth)

annotation CSV re-verified 8/8 vs P1G_MAP: OK
pre-state: all 8 target PKs empty + carry skos (skos-only back-fill): OK
apply_set: 8 PKs -> distribution: {'undermine': 3, 'undercut': 5}
byte-preservation mismatches: 0
well-formedness: 1409 rows x 104 cols, CRLF(True)+BOM(True) preserved
CSV attack-typed total: 121 -> 129   [LAST SUFFIX-ONLY family -> closes 36-row back-fill]

git diff cbf9b7c8 -- Cards/…/Taxonomy.csv = vide (0 write prod).

Handoff ai-01

  • Review byte/discipline/modélisation. 3 MED (888, 1066, 1175) + same-token divergence pk888/pk2 à trancher.
  • Pour sérialiser : tools/498-p1g-apply.py --write (baseline lue dynamiquement : 121→129 ou 137→145 si feat(taxonomy): #498 AIF — serialize P1 tranche-1d+1e+1f prod write (16 SUFFIX-ONLY rows, attack 121->137) #783 mergeé ; PK disjoints).
  • Après write 1g : les 36 SUFFIX-ONLY sont clos → la couche attack couvre les 52 skos-only. Prochaine phase (track séparé, hors-scope tranche) = passe générative des ~866 feuilles restantes (Layer C) — nécessite GO dédié.

Gate boundaries

🤖 Worker po-2024 — réconciliation P1 tranche-1g (Tricherie, MIXTE, dernière famille).

…FIX-ONLY, famille Tricherie, MIXTE gated, LAST FAMILY)

P1 reconciliation back-fill, tranche-1g = 8 SUFFIX-ONLY rows
(888/973/1020/1023/1066/1087/1148/1175), the whole "Tricherie" family
(Arranger les faits + Changement de cap + Pensee biaisee). LAST SUFFIX-ONLY
family -> closes the 36-row back-fill
(5 1b + 7 1c + 4 1d + 5 1e + 7 1f + 8 1g = 36).

HETEROGENEOUS (MIXTE): 3 undermine / 5 undercut. Three defect natures:
  - relational manipulation (deceptive arrangement, goalpost-moving, biased
    reasoning) -> premise-level (undermine/I): 888, 973, 1023 (precedents
    PositionToKnow pk70, ConflictingGoals pk356, Bias pk70).
  - sunk-cost / effort-justification -> inference rule undercut via Waste
    scheme (undercut/RA): 1020, 1148 (precedent Waste_Inference pk432).
  - cognitive induction biases (mere exposure, anecdotal, attribution) ->
    weak induction (undercut/RA): 1066, 1087, 1175 (1066/1175 Example_Inference
    novel; 1087 desc-twin of tranche-1e pk34).

6/8 tokens have a typed precedent; 2/8 novel. attackType = NEW judgment;
NODE deterministic (#707§4a); 0 token fabricated.

3 MED flags (888, 1066, 1175):
  - pk888 (Presentation trompeuse): SAME-TOKEN DIVERGENCE with tranche-1e pk2
    (LackOfPTKReliability_Scheme). pk2->undercut (induction); pk888->undermine
    (deceptive arrangement = biased premise, reinforced by PositionToKnow pk70).
    2nd such divergence (after 1f pk808/pk33) — reinforces desc-driven principle.
  - pk1066 (Effet de simple exposition): Example_Inference novel; undercut
    (familiarity->acceptable defective rule); alt undermine.
  - pk1175 (Biais d'attribution): Example_Inference novel; undercut (attribution
    over-systematic); alt undermine.

5 HIGH: 973 (ConflictingGoals precedent), 1020 (Waste precedent sunk cost),
1023 (Bias+Presumptive precedent), 1087 (desc twin pk34), 1148 (Waste precedent).

0 prod write (gated dry-run). Baseline dynamic (write-driven, disjoint PKs):
121->129 (master cbf9b7c) / 137->145 (if 1d+1e+1f #783 merged). Byte-exact:
0 mismatch, 1409x104, CRLF+BOM, +120 B.

- docs/taxonomy/498-reconciliation-p1g.md (proposition, MIXTE, last family)
- docs/taxonomy/498-reconciliation-p1g-annotations.csv (8 rows, 12 cols)
- tools/498-p1g-apply.py (gated apply, dry-run default, mirrors p1f)

Co-authored-by: Claude-Code <noreply@anthropic.com>
@jsboige
jsboige merged commit 7c2c963 into master Jul 11, 2026
3 checks passed
@jsboige
jsboige deleted the feat/498-p1g-tranche branch July 11, 2026 19:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant