docs(taxonomy): #499 AIF PR-1 (Virtues) — Raisonnement valide / Causalités cluster (proposition, gated)

[jsboige]

What

#499 AIF chantier — PR-1 (Virtues side). Dispatch ompwhx [primaire] (ai-01). Proposition only — GATED, 0 write to prod CSV. Mirror of the po-2023 Fallacies chantier format (#498 PR-1 #699). This is the pilot for serializing the ratified AIF_attackType + AIF_attackedNode schema (#707 §4 Option a) on the Virtues side.

New doc: docs/taxonomy/499-aif-raisonnement-causalite-cluster.md (+108 lines, docs-only).

The cluster

Raisonnement valide (Virtues family 4) / Causalités bien identifiées (subfamily 4.1): anchor pk 80 + leaves 81/82/83. Chosen as the cleanest inverse-paradigm mirror — the Fallacies subfamily Causalité douteuse sits at path 4.1.x (698 Pétition de principe, 707 Inversion de causalité, 719 Effet cigogne), the Virtues subfamily Causalités bien identifiées sits at path 4.1. Same path prefix, opposite tenor — structural corroboration of jsboige's inverse paradigm, independent of the existing crossLink_Opposes mapping.

What this extends (not redoes)

The Virtues 12-col relational layer (crossLink_* + AIF_skos*) is already generated and validated CLEAN — 222/222 nodes, 9 PRs merged June 2026, validator #518 12/12 (see 499-virtues-parity-closure.md). This PR reuses that grounding (scheme, restored CQ, opposed Fallacy PKs) and only adds the I/RA/CA decomposition the newly-ratified schema asks for. No re-derivation.

The inverse-paradigm adaptation — FLAGGED for jsboige

The ratified columns are attack-shaped (attackType on attackedNode). A Virtue is the good holding of a scheme, not an attack. Three serialization options are documented in §3:

Option	Virtue-side semantics	Reco
A — "attack resisted"	columns record the attack-type the virtue prevents + the node it holds (mostly undercut / RA-node)	✅ recommended
B — structurally empty	columns empty (a virtue attacks nothing)	loses the info
C — AIF-native support	RA-node only, no attackType (would require fabricating a support token)	violates #677

The pilot is written under Option A to make the proposition concrete (4/4 undercut on RA-node, consistent with jsboige's "most fallacies live in the undercut"). Only the last two table columns change if jsboige picks B or C. This is a judgment call on inverse-paradigm semantics — flagged for validation before the 222-row back-fill.

Why this matters (the EPITA consumer)

The harnais EPITA (2025-Epita-Intelligence-Symbolique) evaluates argumentation quality with 9 hard-coded virtues because it has no consumable Virtues backbone. The ratified I/RA/CA schema + this mirror give the quality axis the same verifiable mechanics as the fallacy axis (1408 nodes). Option A lets the consumer read, per virtue: which scheme it correctly holds, against which attack — the inverse of a fallacy row.

Proposition (Option A, per leaf)

pk	title_fr	RA-node (scheme held)	CA-node resisted	AIF_attackType	AIF_attackedNode
80	Causalités bien identifiées (anchor)	Cause to Effect	causal RA held vs inversion/single-cause/correlation	undercut	RA-node
81	Indépendance des prémisses	Argument from Rule	rule RA held vs circularity (698)	undercut	RA-node
82	Causalité bien orientée	Cause to Effect	causal RA held vs direction inversion (707)	undercut	RA-node
83	Exclusion des causes alternatives	Cause to Effect	causal RA held vs spurious correlation (719/633)	undercut	RA-node

0 FAIL-LOUD in this cluster (every leaf has a native scheme + native attack-type). Honest scheme-divergence kept (pk 83 opposes cross-family 633 Relation infondée).

Discipline / gates

• ✅ Docs-only proposition. 0 prod CSV write, 0 DB, 0 OWL regen, 0 Cards/ write.
• ✅ Reuses the validated 12-col layer — no re-derivation, no fabrication. All AIF tokens native (undercut, RA-node, Cause to Effect, Rule) — discipline docs(taxonomy): #141 crossLink enrichment sample run (real output curated from #626) #677 held.
• ✅ Mirror of po-2023 Fallacies format (proposition gated, 1 cluster, method + flag for validation).
• ❌ [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 NOT merged (HOLD).

Asks for review

1. Option A vs B vs C (§3) — the inverse-paradigm adaptation. My recommendation is A; jsboige's call.
2. Once Option is ratified: the 222-row back-fill is largely mechanical (attackType/attackedNode derivable from scheme + opposed-family via the existing 12-col annotations) — a single programmatic back-fill PR is plausible.

Relates: #499, #498 (po-203 Fallacies owner), #707 §4 (ratified schema), #677 (0 fabrication), #518 (validator), 499-virtues-parity-closure.md, 499-virtues-prod-write-spec.md, 498-coverage-status.md.

Co-Authored-By: Claude-Code noreply@anthropic.com