docs(taxonomy): #499 Virtues AIF PR-8 — Argument pertinent 1.2 Prémisses fiables + valide 3e scheme majeur Position to Know (8 nœuds) + confirme credibility-attack 942

[jsboige]

Summary

8th cluster of Virtues AIF I/RA/CA decomposition (schema #707§4 Option A ratified). Covers family 1 Argument pertinent / subfamily 1.2 Prémisses fiables (3 leaves pk 13/25/26, anchor pk 12). 2 schemes: Argument from Position to Know (13, 25) + Argument from Bias (26).

Follows PR #741-#747 (7 clusters + back-fill plan; #741-#746 merged at e02857c6, #747 OPEN). Same method. Docs-only, no code/CSV/DB/OWL writes.

🔑 KEY CONTRIBUTION 1 — validates the 3rd major scheme Argument from Position to Know (first coverage, 8 nodes)

Argument from Position to Know was the 3rd major scheme remaining uncovered (8 nodes / 222 = ~4%). PR-8 covers it on the paradigmatic case:

• PK 13 Sources crédibles + PK 25 Sources indépendantes oppose {71 Argument d'autorité, 942 Fausse attribution} (leaf 13) / {953, 596} (leaf 25) — the archetypes of mis-invokeable expertise (authority cited instead of facts / falsified source).
• Verdict: undercut / RA-node. The fallacies defeat the Position to Know Application node — the source is not genuinely in a position to know what it claims.

Confirms (with Rule PR-5, Sign PR-7): Argument from Position to Know → default undercut/RA-node, NO override expected. The override comes from the opposed fallacy's mechanism (889/804 → mine, 340 → rebut), NOT from the virtue's scheme. The 8 Position to Know nodes = pure mechanical default in #744. "Scheme is NOT an override axis" now validated on top-3 schemes: Rule 50 + Sign 26 + Position to Know 8 = 84 nodes / ~38% of taxonomy.

🔑 KEY CONTRIBUTION 2 — confirms the credibility-attack classification of 942 Fausse attribution (re-examined in Position to Know context)

The back-fill plan #744 §2a explicitly excluded 942 Fausse attribution from the undermine set ("source = credibility, cf. PR-2 → undercut/RA, NOT undermine"). PR-8 re-examines 942 in its Position to Know context (leaf 13) to confirm and document the distinction:

942 Fausse attribution — « vous donnez du poids à votre argument en citant une source fausse, mal identifiée, incompétente ou inventée ».

• The source is false/invented. But the fabrication does NOT bear on the underlying proposition P — it bears on the epistemic qualification of the source ("this source is competent/identifiable").
• The arguer does not assert "P is true" as a fabricated premise (889-style); he asserts "this source knows P" and derives P from that. The falsified premise is the source's reliability, not P itself.
• → the attack defeats the inference (reliable source → P), not the conclusion P directly.
• → undercut / RA-node, NOT undermine.

Confirmation: 942 is a credibility-attack (as established PR-2). The discriminator holds: undermine/I is reserved for fabricated/false underlying premises (889, 804); the falsification of a source qualification (942) defeats the inference, not the underlying premise. Strengthens #744 §2a (942 stays undercut/RA).

Distribution cluster 1.2

6/6 undercut/RA-node (uniform like PR-1 4.1, PR-3 7.3, PR-5 3.1, PR-7 1.1). 0 undermine, 0 rebut. Un-reliable-source fallacies (authority, false attribution), biased-selection fallacies (953, 596) and cognitive-bias fallacies (1024) are all inference failures — none asserts a fabricated underlying premise, none presents a counter-conclusion.

Leaf	Fallacy	attackType	attackedNode
13	71 Argument d'autorité	undercut	RA-node
13	942 Fausse attribution	undercut	RA-node
25	953 Attention sélective	undercut	RA-node
25	596 Échantillon biaisé	undercut	RA-node
26	953 Attention sélective	undercut	RA-node
26	1024 Biais naturels	undercut	RA-node

Cumulative rebut-rarity (8 clusters, load-bearing)

Cluster	Subfamily	Family	Schemes	mine	undercut	rebut
PR-1	4.1	Raisonnement	Cause to Effect	0	6	0
PR-2	6.1	Honnêteté	Witness/Bias	1	2	0
PR-3	7.3	Échange	Commitment	0	6	0
PR-4	5.1	Langage	Verbal Cl./Expert	3	7	0
PR-5	3.1	Rigueur math	Example/Rule/Analogy	0	6	0
PR-6	2.2	Présentation	Bias/Consequences	0	4	2
PR-7	1.1	Argument pertinent	Sign/Bias	0	6	0
PR-8	1.2	Argument pertinent	Position to Know/Bias	0	6	0
Cumul	8 subfamilies	7 families / 11 schemes		4	43	2

rebut/CA-node = 2 across 8 families / 11 scheme-types / ~49 fallacy-instances. rebut stays absent on Position to Know and Bias — schemes that don't host counter-conclusions (an unreliable source, a biased selection, or a cognitive bias doesn't deny the conclusion, it just fails to support it). rebut is localized + predictable: Consequences only. The back-fill plan's single rebut override {340} holds.

Scheme coverage

12/14 schemes covered (87% of taxonomy by node count), including the 8 most represented: Rule (50), Commitment (40), Bias (27), Sign (26), Verbal Classification (21), Cause to Effect (11), Witness Testimony (10), Position to Know (8, NEW) = 193 nodes. Remaining: Values (8, family 2), Danger (1).

Family 1 now covered on 2/3 subfamilies (1.1 + 1.2, 1.3 remaining).

Discipline

• ✅ Read body before action: family 1 structure enumerated (33 nodes, 3 subfamilies), phase-2 argument-annotations.csv scheme/CQ/opposes read (29 annotated nodes), 3 opposed-fallacy descriptions read (71/942/1024) before attack-type derivation; anti-duplication grep confirmed no phase-2 doc decomposes I/RA/CA (additive layer).
• ✅ 0 fabrication (docs(taxonomy): #141 crossLink enrichment sample run (real output curated from #626) #677): Position to Know/Bias/undercut/RA-node all native AIF/Walton.
• ✅ 942 re-examined honestly — credibility-attack classification confirmed and documented via the distinction "falsification of source qualification (942) ≠ falsification of underlying premise (889)". Consistent with PR-2 and docs(taxonomy): #499 Virtues AIF — mechanical back-fill plan 222-rows (Option A, HOLD prod) #744.
• ✅ 71 = RA-node paradigmatic documented (Appel à l'autorité = Position to Know Application failure archetype), not forced.
• ✅ 1024 honestly classified undercut (cognitive bias = inference treatment), consistent with 1242 (PR-7) on the same bias-of-framework logic.
• ✅ 0 code change, 0 write Cards/, 0 regen, 0 CSV/DB/OWL. Docs-only (+211 lines).
• ❌ Does NOT touch [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).

Test impact

None — docs-only.

Request

Review + (if acceptable) merge. This cluster (a) validates the 3rd major scheme Position to Know (8 nodes, no scheme-override — top-3 schemes now validated = ~38% of taxonomy in pure mechanical default), (b) confirms the credibility-attack classification of 942 with a documented distinction (source-qualification falsification ≠ underlying-premise falsification). Both findings feed back into PR #744 (back-fill plan) — no change to the plan's rule, but strengthened confidence in §2a (942 stays undercut/RA) and the "scheme is NOT an override axis" simplification.

cc @jsboige (Option A FYI).

🤖 Worker po-2024 — [primaire] cluster 8 of dispatch ompwhx deep-queue.