Skip to content

docs(taxonomy): #498 AIF chantier PR-1 — Fallacious comparison cluster (proposition, gated) #699

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jsboige merged 4 commits into from
Jul 5, 2026
Merged

docs(taxonomy): #498 AIF chantier PR-1 — Fallacious comparison cluster (proposition, gated) #699

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c96051b
docs(taxonomy): #498 AIF chantier PR-1 — Fallacious comparison cluste…
jsboige Jul 5, 2026
94e55c8
fix(taxonomy): #498 PR-1 correct cluster boundary — pk 838 is Faulty …
jsboige Jul 5, 2026
2d3a0fe
docs(taxonomy): #498 PR-1 show both 833 Conflict nodes (anchor fideli…
jsboige Jul 5, 2026
eb96e65
docs(taxonomy): #498 PR-1 add §7 enriched-DoD mapping (I-node/RA-node…
jsboige Jul 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
280 changes: 280 additions & 0 deletions docs/taxonomy/498-aif-fallacious-comparison-cluster.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,280 @@
# 2026-07-05 — #498 AIF chantier, PR-1 : cluster « Fallacious comparison » (Misleading language)

**Scope**: first cluster of the #498 chantier — model the **Fallacious comparison** subtree (Misleading
language family) as **defeasible exceptions to legitimate argument schemes**, in the Walton/AIF
paradigm. **Proposition only — GATED, 0 write to prod CSV.** Triggered by ai-01 dispatch `yj7u3j`
(primary). This is a multi-session chantier; this PR advances one coherent cluster.

**Repo reference**: master `dc02e847`. Issue: #498 (reformulated, GO jsboige 2026-06-17 verified).
Linked: #133/#130 (existing OWL ontology, regenerated to reflect new structures — not created).

> **⚠ STATUS = PROPOSITION ONLY.** No prod CSV write, no DB write, no OWL regen. This doc proposes
> the AIF structure (scheme + exception/CQ) for the cluster's unmapped leaves, derived code=truth
> from the CSV. jsboige ratifies → worker applies the CSV edits in a follow-up PR (gated).

---

## TL;DR

The **Fallacious comparison** subfamily (`Subfamily=Fallacious comparison`, Misleading language
family) has **2 mapped nodes** — pk 833 (d2 subfamily root "Comparaison fallacieuse",
`BiasedClassification_Conflict` + `ExceptionSimilarityCase_Conflict` directRefs, `skos:broadMatch`)
and pk 839 (d3 "Fausse analogie", `Analogy_Inference` defeated by
`DifferencesUndermineSimilarity_Conflict`, `skos:closeMatch`) —
across 3 sub-sub clusters: **Faulty comparison** (834-838), **False analogy** (839-843), and
**Association fallacy** (844-845). **This PR scopes to the False analogy sub-sub**: anchor pk 839 +
**4 unmapped leaves** (pk 840, 841, 842, 843). The anchor establishes the cluster's pattern —
**exception to `Analogy_Inference` via the `DifferencesUndermineSimilarity_Conflict` critical
question** — and the 4 leaves are specialized variants of the same defeasible structure. This PR
proposes the AIF exception-modeling for the 4 leaves, reusing the anchor's scheme, and **failing
loud** where a leaf does not honestly fit the Analogy_Inference scheme (documenting it as such
rather than fabricating).

> **⚠ Cluster-boundary correction (code=truth re-verification).** An earlier draft of this
> proposition incorrectly placed pk 838 ("Distinction sans différence") in the False analogy
> sub-sub and labeled pk 834 as the subfamily root. Re-verification against the CSV `Soussousfamille`
> column shows: (a) the true subfamily root is **pk 833** (d2, MAPPED — missing from the earlier
> draft); (b) **pk 838 belongs to Faulty comparison**, not False analogy. The False analogy sub-sub
> is strictly pk 839-843 (anchor + 4 leaves). pk 838 (and its honest scheme-divergence modeling) is
> deferred to **PR-2** (Faulty comparison cluster). This correction strengthens PR-1's methodology:
> all 4 False-analogy leaves now honestly reuse `Analogy_Inference`, with no scheme-divergence
> needed in this cluster.

---

## 1. Cluster state (code=truth)

From `Cards/Fallacies/Argumentum Fallacies - Taxonomy.csv`, family **Misleading language**, subfamily
**Fallacious comparison**:

| pk | text_fr | Subsubfamily | DirectRef | ExceptionRef | MappingType | State |
|----|---------|--------------|-----------|--------------|-------------|-------|
| **833** | **Comparaison fallacieuse** (d2) | *(root)* | `BiasedClassification_Conflict`, `ExceptionSimilarityCase_Conflict` | — | `skos:broadMatch` | **✅ MAPPED (subfamily root)** |
| 834 | Comparaison abusive (d3) | Faulty comparison | — | — | — | unmapped (d3 anchor → PR-2) |
| 835 | Comparaison incomplète | Faulty comparison | — | — | — | unmapped (→ PR-2) |
| 836 | Classification non exclusive | Faulty comparison | — | — | — | unmapped (→ PR-2) |
| 837 | Comparaison incohérente | Faulty comparison | — | — | — | unmapped (→ PR-2) |
| 838 | Distinction sans différence | Faulty comparison | — | — | — | unmapped (→ PR-2) |
| **839** | **Fausse analogie** (d3) | False analogy | `DifferencesUndermineSimilarity_Conflict` | `Analogy_Inference` | `skos:closeMatch` | **✅ MAPPED (PR-1 anchor)** |
| 840 | Pétition de principe analogique | False analogy | — | — | — | unmapped (**PR-1**) |
| 841 | Analogie étendue | False analogy | — | — | — | unmapped (**PR-1**) |
| 842 | Argument de la similarité fallacieuse | False analogy | — | — | — | unmapped (**PR-1**) |
| 843 | Fausse équivalence | False analogy | — | — | — | unmapped (**PR-1**) |
| 844 | Sophisme d'association (d3) | Association fallacy | — | — | — | unmapped (d3 anchor → PR-3) |
| 845 | Amalgame | Association fallacy | — | — | — | unmapped (→ PR-3) |

The mapped anchor (839) is the **template**: a fallacious comparison = an `Analogy_Inference` scheme
**defeated** because the critical question "do the differences between the compared items undermine
the similarity?" (`DifferencesUndermineSimilarity_Conflict`) is violated. The 4 leaves are
specialized ways that analogy goes wrong.

**This PR scopes to the `False analogy` sub-subfamily (pk 839-843)** — anchor 839 + 4 leaves — the
tightest coherent cluster with an in-sub-sub mapped anchor. The `Faulty comparison` sub-subfamily
(834-838, no in-sub-sub anchor → anchored by the d2 root 833) is **PR-2**; `Association fallacy`
(844-845) is **PR-3**.

---

## 2. The parent model (839, the template)

**Legitimate scheme**: `Analogy_Inference` (AIF native) — "A is like B in property P; B has property
Q; therefore A has property Q." (Walton's argument from analogy.)

**Critical question violated** (the defeaser): `DifferencesUndermineSimilarity_Conflict` — "Are there
relevant differences between A and B that undermine the transfer of Q?" The fallacy ignores these.

**Mapping type**: `skos:closeMatch` — "Fausse analogie" is the informal-logic name for the violated-CQ
case of `Analogy_Inference`.

This is **exactly the paradigm #498 asks for**: a fallacy = a defeasible exception to a legitimate
scheme, named by the violated critical question, not by a Latin label alone.

---

## 3. Proposed AIF structure for the unmapped leaves

For each leaf, the proposal: **(a) legitimate scheme targeted, (b) exception/CQ that defeats it,
(c) `AIF_skosMappingType`**. Reuse the parent's scheme where the leaf is a specialized analogy-failure;
**fail loud** where a leaf targets a different scheme honestly.

### pk 840 — Pétition de principe analogique (Analogical begging the question)
- **desc_fr**: "Vous utilisez une analogie qui suppose déjà la conclusion que vous cherchez à
démontrer, en assimilant de manière discutable deux situations."
- **Legitimate scheme**: `Analogy_Inference`
- **Exception/CQ**: the analogy **begs the question** — the disputed conclusion (A and B are
assimilable) is smuggled into the analogy's premise. This is **not** the same CQ as 839
(differences undermining similarity) — it's a circularity defeater. The closest AIF native conflict
node is a **question-begging / circularity** CQ.
- **Proposal**: `ExceptionRef=Analogy_Inference`, `DirectRef=` *(no clean AIF native circularity
Conflict node exists for analogy)* → **FAIL LOUD**: document that no AIF native CQ cleanly
captures "the analogy presupposes its conclusion"; propose `AIF_skosOther="Circularity:
analogy presupposes the disputed assimilation (no native AIF Conflict node)"`,
`MappingType=skos:closeMatch`. Do **not** fabricate a `*_Conflict` token.

### pk 841 — Analogie étendue (Extended/Transitive analogy)
- **desc_fr**: "À partir d'une chose qui ressemble à une deuxième et partage une propriété avec une
troisième, vous concluez que la deuxième et la troisième se ressemblent."
- **Legitimate scheme**: `Analogy_Inference` (transitive chain: A~B, A~C ⟹ B~C)
- **Exception/CQ**: transitivity of similarity is not guaranteed — similarity is not transitive.
This **is** a clean AIF conflict: the differences between B and C undermine the inferred similarity.
- **Proposal**: `ExceptionRef=Analogy_Inference`, `DirectRef=DifferencesUndermineSimilarity_Conflict`
(same CQ as 839, applied to the B-C link rather than A-B), `MappingType=skos:narrowMatch` (a
narrower, transitive-chain variant of 839).

### pk 842 — Argument de la similarité fallacieuse (Fallacious similarity)
- **desc_fr**: "Vous considérez que deux choses qui se ressemblent sont liées ou similaires."
- **Legitimate scheme**: `Analogy_Inference` (or more loosely `Sign_Inference` — resemblance as a
sign of relatedness)
- **Exception/CQ**: superficial resemblance does not establish a real link/causal relation. The CQ:
"is the resemblance merely superficial, without an underlying connection?" Closest AIF:
`DifferencesUndermineSimilarity_Conflict` (the resemblance is too thin to transfer any property).
- **Proposal**: `ExceptionRef=Analogy_Inference`, `DirectRef=DifferencesUndermineSimilarity_Conflict`,
`MappingType=skos:broadMatch` (the fallacy is broader — covers mere-resemblance-without-property-
transfer, a looser case than 839's property-transfer analogy).

### pk 843 — Fausse équivalence (False equivalence)
- **desc_fr**: "Vous faites un parallèle impropre entre deux arguments distincts."
- **Legitimate scheme**: `Analogy_Inference` (two arguments treated as equivalent)
- **Exception/CQ**: the two arguments differ in a **relevant** respect (strength, structure,
evidence) that defeats their equivalence. This is again `DifferencesUndermineSimilarity_Conflict`,
but applied to **arguments** rather than objects — a well-known specialized case.
- **Proposal**: `ExceptionRef=Analogy_Inference`, `DirectRef=DifferencesUndermineSimilarity_Conflict`,
`MappingType=skos:closeMatch` (a direct, argument-level variant of 839).

> **Note on pk 838** ("Distinction sans différence"): this leaf sits in the **Faulty comparison**
> sub-sub (`Soussousfamille=Faulty comparison`), **not** False analogy — see §1 boundary correction.
> It is the mirror fallacy (claiming a distinction where none exists) and honestly targets a
> **classification** scheme (`ArbitraryVerbalClassification_Inference` +
> `PropertyNotExistant_Conflict`) rather than `Analogy_Inference`. Its full modeling is deferred to
> **PR-2** (Faulty comparison cluster), to keep this PR scoped to a single coherent sub-sub.

---

## 4. Cluster summary (proposed)

| pk | text_fr | Proposed ExceptionRef | Proposed DirectRef | MappingType | Honest? |
|----|---------|----------------------|-------------------|-------------|---------|
| 839 | Fausse analogie | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | `skos:closeMatch` | ✅ (existing anchor) |
| 840 | Pétition de principe analogique | `Analogy_Inference` | *(Other: circularity — no native CQ)* | `skos:closeMatch` | ⚠ FAIL LOUD (no native node) |
| 841 | Analogie étendue | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | `skos:narrowMatch` | ✅ |
| 842 | Argument de la similarité fallacieuse | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | `skos:broadMatch` | ✅ |
| 843 | Fausse équivalence | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | `skos:closeMatch` | ✅ |

**4 leaves proposed (840-843), 1 FAIL-LOUD (840)** — documents that AIF has no native circularity-CQ
for analogy rather than fabricating one. All 4 leaves honestly reuse the anchor's
`Analogy_Inference` scheme; the only variation is the violated-CQ specialization (840) and the
`MappingType` granularity. The honest scheme-divergence case (pk 838, classification fallacy)
moves to **PR-2** (Faulty comparison cluster), keeping this PR a single coherent sub-sub.

---

## 5. Method notes (for the chantier's后续 PRs)

- **Cluster = a sub-sub-family with an in-sub-sub mapped anchor + its unmapped sibling leaves.**
This PR's cluster (False analogy, anchor 839 + 4 leaves) is the unit of work per PR. Where a
sub-sub has no in-sub-sub anchor (e.g. Faulty comparison 834-838), it is anchored by the parent
d2/d3 node — that is PR-2's shape.
- **Reuse the anchor's scheme + CQ** where the leaf is a specialized violation of the same scheme;
vary only the `MappingType` (`narrowMatch` = more specific, `broadMatch` = looser, `closeMatch` =
direct variant).
- **Fail loud** when no native AIF Conflict node captures the leaf's defeater (840) — use
`AIF_skosOther` to document, never fabricate a `*_Conflict` token.
- **Honest scheme-divergence** — when a leaf sits in a subfamily but actually targets a different
scheme (pk 838: a comparison-subfamily leaf that is genuinely a *classification* fallacy), say so
and use the right scheme; don't force the anchor's scheme. (838 lives in the Faulty comparison
sub-sub → its modeling is in PR-2, not here.)
- **Coverage accounting** (#498 DoD): this PR adds **4 mapped leaves** to the 70 existing → 74/1408.
At cluster-level, the False-analogy sub-sub goes from 1/5 mapped to 5/5. Effective cluster
coverage of the comparison subfamily improves; leaf-level "100%" is explicitly **not** the goal.

---

## 6. DoD status

| DoD item (#498) | Status |
|-----------------|--------|
| Effective ~31% cluster-level coverage documented (not "100% leaves") | ✅ §5 (70→74 mapped, cluster-level) |
| Per unmapped leaf: legitimate scheme + exception/CQ (not Latin alone) | ✅ §3 (4 leaves, scheme + CQ each) |
| `AIF_skosMappingType` coherent (broadMatch when narrower) | ✅ §4 (varied per leaf) |
| Fail-loud when no honest scheme fits | ✅ §3 pk 840 (no native circularity CQ — documented, not fabricated) |
| OWL regen reflects new structures | ⏳ Deferred — jsboige ratifies this proposition first, then CSV edits applied, then OWL regen (#133) — not in this PR |

---

## 7. Mapping to jsboige's enriched DoD (I-node / RA-node / CA-node decomposition)

The last jsboige comment on #498 enriches the DoD: modeling a fallacy "AIF-style" decomposes the
argument into **I-nodes** (premises + conclusion), an **RA-node** (the inference principle = the
Walton scheme, a first-class attackable object), and a **CA-node** (the Conflict Application, typed
by attack: **undermine** a premise / **undercut** the RA-node / **rebut** the conclusion). This
section maps the 4 leaves of this cluster to that decomposition. (PR-2 #701 carries the parallel
section for the Faulty comparison cluster.)

### pk 841 — Analogie étendue → **undercut**
- **RA-node**: `Analogy_Inference` applied transitively (A~B, A~C ⟹ B~C).
- **I-nodes**: premise P1 "A~B" (similarity); premise P2 "A~C" (similarity); conclusion C "B~C".
- **CA-node**: `DifferencesUndermineSimilarity_Conflict`, applied as an **undercut** on the RA-node
— the inference (transitivity of similarity) is defective because similarity is not transitive.
The conclusion C may still hold; it is the *inference* that is defective. Classic undercut.
- **CQ that fails**: "is similarity transitive?" — no.

### pk 842 — Argument de la similarité fallacieuse → **undercut**
- **RA-node**: `Analogy_Inference` (resemblance ⟹ relatedness/causal link).
- **I-nodes**: premise P "A resembles B"; conclusion C "A and B are linked/related".
- **CA-node**: `DifferencesUndermineSimilarity_Conflict`, applied as an **undercut** — the
resemblance is too superficial to license the inferred link. Again the inference, not the
conclusion, is the target.
- **CQ that fails**: "is the resemblance more than superficial?" — no.

### pk 843 — Fausse équivalence → **undermine**
- **RA-node**: `Analogy_Inference` (two arguments treated as equivalent).
- **I-nodes**: premise P "argument X ≡ argument Y (in strength/structure/evidence)"; conclusion C
"what holds for X holds for Y".
- **CA-node**: `DifferencesUndermineSimilarity_Conflict`, applied as an **undermine** on P — the
premise "X ≡ Y" is false because X and Y differ in a *relevant* respect. Unlike 841/842, the
attack lands on a *premise* (I-node), not the inference itself.
- **CQ that fails**: "do X and Y differ in a relevant respect?" — yes.

### pk 840 — Pétition de principe analogique → FAIL-LOUD at the **CA-node level**
- **RA-node**: `Analogy_Inference` (exists — the analogy is the inference).
- **CA-node**: ⚠ **absent (FAIL-LOUD)** — circularity in analogy would be an **undercut** (the
analogy presupposes the disputed assimilation it is meant to establish), but **no native AIF
Conflict node captures circularity/question-begging** for analogy. The CA-node cannot be typed
without a native conflict token; fabricating one is forbidden (FAIL LOUD, §3 pk 840).
- This is a different FAIL-LOUD shape from PR-2's pure-comparison leaves (834/835/837): there the
**RA-node** was missing (no native comparison scheme); here the **RA-node exists** but the
**CA-node** is missing. Both are honest gaps, at different layers of the decomposition.

### Cluster attack-type distribution
| pk | RA-node | CA-node | Attack type |
|----|---------|---------|-------------|
| 840 | `Analogy_Inference` | *(absent — FAIL LOUD)* | would be undercut (circularity) — untyped |
| 841 | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | undercut |
| 842 | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | undercut |
| 843 | `Analogy_Inference` | `DifferencesUndermineSimilarity_Conflict` | undermine |

**3 of 4 leaves are undercuts** (the inference is defective, the conclusion may stand) — consistent
with jsboige's note that "most fallacies live in the undercut." 843 is the lone undermine (the
premise is false). 840 is the lone FAIL-LOUD (CA missing, not RA).

### Representation note
As in PR-2, the I-node/RA-node/CA-node decomposition is **recorded here but not serialized** in the
CSV — it would require new columns (`AIF_attackType`, `AIF_attackedNode`) and is the terminal target
of jsboige's comment. Recording it in the proposition lets jsboige ratify the attack-typing *before*
any CSV schema evolution. The two PRs together (PR-1 False analogy + PR-2 Faulty comparison)
demonstrate the decomposition across both cluster shapes (in-sub-sub anchor + borrowed anchor) and
both FAIL-LOUD layers (CA-missing vs RA-missing).

---

## Gate boundaries (HARD — proposition only)

- ❌ No prod CSV write, no DB write, no OWL regen, no `aif:` assertion emitted.
- ❌ No #674/#596 merge, no régén launch.
- ✅ Proposition derived code=truth from the taxonomy CSV (cluster + AIF columns) + AIF scheme
semantics (Walton/`Analogy_Inference` + `DifferencesUndermineSimilarity_Conflict` are AIF native).
- ✅ Fail-loud on 840 (no fabricated Conflict node); pk 838's honest scheme-divergence deferred to
PR-2 (cluster-boundary correction §1 — 838 is in Faulty comparison, not this cluster).

Relates: dispatch `yj7u3j` (primary), #498 (chantier), #133/#130 (existing OWL), #499 (inverse:
virtue = good tenor of a scheme), #192 (terminology), #458.
Loading