Citation Integrity Spot-Check — Phase 6 Report
Date: 2026-04-25
Audit phase: Phase 6 (per AUDIT_PLAN.md v1.3, M4 sample-size + S2 audit-produced inclusion + R3 caveat)
Sample design
- Sample size: 20 citations (within the 15-25 range; M4 anchors ~3-5% of estimated 400-600 corpus citations).
- Bias: weighted toward (a) citations to authors not in
raw/(highest-risk per M4) and (b) ~30% audit-produced pages from Phases 1-2-4 (per S2 — actual: 6 of 20 = 30%, exceeding the ~20% S2 floor). - Faul fabrication check (mandatory): performed prior to sampling.
Faul fabrication check
grep -rn -iE "Painter.s Courage|JBSP 55[: ]2|Journal of the British Society for Phenomenology.*2024" wiki/
grep -rn -E "\bFaul\b" wiki/
grep -rn "Ontologically Interactive Painting" wiki/
All hits are confined to audit-meta files discussing the absence of the fabrication: wiki/log.md, wiki/overview.md, wiki/.audit/paper-a-gaps-2026-04-25.md. Zero hits in wiki/concepts/, wiki/sources/, wiki/entities/, or wiki/questions/. The wiki content layer is uncontaminated by the Paper A drafting incident; the meta layer correctly logs the audit checks.
Update from Phase 1: the new raw/ files now include Heinbokel and Kee — their works were uploaded between Phase 1 (2026-04-25) and Phase 6 (2026-04-25, same day). Neither has been ingested yet, but this changes the citation-verifiability landscape: future citations to Heinbokel and Kee can now be verified against the raw/ source if the ingest is performed.
Sample (20 citations)
Tier 1 — citations to authors not in raw/ (highest risk, 10 samples)
| # | Citation | Wiki location | Verification |
|---|---|---|---|
| 1 | Steinbock — static/genetic/generative tripartite | passivity §"Beith"; generative-passivity; don-beith | PASS — citation-of-Beith's-citation-of-Steinbock; Beith 2018 IS in raw/ and the extraction note records the Steinbock attribution. The wiki claim is about Beith's appropriation, not direct Steinbock. |
| 2 | Schopenhauer 1874 Nachlass 34[21] | granite-of-fate | PASS — Chouraqui 2014 raw line 2576 records the Schopenhauer middle-term attribution explicitly; the Chouraqui-2014-motifs extraction note logs this. |
| 3 | Cassirer PSF III, 1929, p. 380 (Schwungrad) | volant, ernst-cassirer, passivity §"Dough, Blisters, Flywheel" | PASS — Saint Aubert E&C II Ch I § 1b is the source; Saint Aubert IS in raw/. The Cassirer page reference is internal to Saint Aubert's reconstruction. |
| 4 | Wahl 1951 (HoXX p. 353-354/237) | ambiguity-vs-ambivalence; questions/is-ambiguite-the-sortie-from-ambivalence | PASS — citation is to Saint Aubert's transcription of MP's reference to Wahl, anchored in HoXX p. 353-354/237. Saint Aubert IS in raw/. |
| 5 | Cousteau-Malle 1955 Le Monde du silence + 1956-7 obstetric ultrasound | pregnancy-pragnanz, silence | PASS — Saint Aubert E&C II Ch V is the source; Saint Aubert IS in raw/. Film and ultrasound are real historical artifacts (Palme d'Or 1956 verifiable). |
| 6 | Husserl Origin of Geometry / Urstiftung-Nachstiftung-Endstiftung | institution §"Husserl Stiftung Lineage" | PASS — MP's In Praise of Philosophy Course 11 pp. 188-194 is the citation source; IPoP IS in raw/. |
| 7 | Heraclitus B217 (ever-living fire) | flesh-as-element §"Motif Cluster: fire/ignition/spark"; motifs.md | PASS — Kaushik 2019 ch. 1 + ch. 3 is the source; Kaushik IS in raw/ and the extraction note logs the Heraclitus B217 attribution. |
| 8 | Stendhal De l'Amour 1822 (salt-mine analogy) | cristallisation, texture-imaginaire-du-reel | PASS — Saint Aubert E&C II Ch II § 4 + Ch IV § 3; Saint Aubert IS in raw/. Stendhal is a real published work. |
| 9 | Breton L'Amour fou 1937 | cristallisation, texture-imaginaire-du-reel | PASS — same as #8. |
| 10 | Lacan 1949 Revue Française de Psychanalyse "Le stade du miroir comme formateur du fonction du je" | specular-image | PASS — citation is to MP's own Primacy of Perception fn. 18 at p. 156; PoP IS in raw/. Citation traces via MP, not direct. |
Tier 2 — audit-produced page citations (per S2, 6 samples = 30%)
| # | Citation | Audit phase | Wiki location | Verification |
|---|---|---|---|---|
| 11 | E&M raw line 35 ("secret science / fundamental of painting") | Phase 1 | science-secrete | PASS — line 35 anchored directly in raw E&M; Phase 1 gap report records the location. |
| 12 | PhP haecceity attestations (pp. lxxv, 15, 40, 45, 128, 221, 428, 440, 477) | Phase 2 | haecceity | PASS — 9 specific page references to PhP (in raw/); Phase 2 silent-key scan report records all 9. |
| 13 | V&I Ch 4 raw line 535/1235 ("opening of a dimension") | Phase 2 | initiation | PASS — line 535/1235 anchored directly in raw V&I (the duplicate line numbering is a translation feature noted in the page). |
| 14 | PhP Preface lines 1150 + 1261 ("phenomenology in a nascent state") | Phase 2 | nascent-state | PASS — lines anchored directly in raw PhP; Phase 2 report records the 14+ attestations across the book. |
| 15 | Carbone 2015 ch 5 (khôra) + Carbone 2015 ch 4 ("the level") + Kaushik 2019 ch 5 (Erinnerung) | Phase 2 + motifs.md | initiation | PASS — Carbone 2015 and Kaushik 2019 both IN raw/ with extraction notes; Phase 2 silent-key scan report records the cross-source matrix-cluster observation. |
| 16 | Chouraqui 2014 Conclusion raw line 2324 ("this very circle itself is sedimentation") | Phase 1 era (motif-propagation 2026-04-21) — sampled here as audit-produced because of Phase 0a citation-rule retroactive coverage | sedimentation, circulus-vitiosus-deus, self-falsification | PASS — line 2324 anchored directly in raw Chouraqui 2014. Cardinal three-term identity (circle = sedimentation = self-falsification) recorded in the 2026-04-21 motif re-ingest log entry. |
Tier 3 — specific journal/page references on legacy pages (4 samples)
| # | Citation | Wiki location | Verification |
|---|---|---|---|
| 17 | Saint Aubert E&C II Ch I § 1b — volant ~20 occurrences 1951-1961 | volant | PASS — Saint Aubert in raw/; the survey claim is internal to Saint Aubert. |
| 18 | Gardner 2018 fn. 15 + fn. 16 (Kant Beweisgrund 2:82 quoted) | unvordenklich, dass-was-distinction, immanuel-kant | PASS — Gardner 2018 (in raw/ as Thought_s_Indebtedness_to_Being_Gardner.md); the Kant Beweisgrund citation is via Gardner's quotation, anchored to the paper's footnotes 15 + 16. |
| 19 | Beith 2018 Steinbock-tripartite (static/genetic/generative) | generative-passivity, passivity | PASS — Beith in raw/; same as #1. |
| 20 | MP Possibility of Philosophy Course 3 line 1872 ("history is, in this sense, the very flesh of humanity") + Marx 1844 Manuscripts | flesh-as-element §"The Marxist Source of the Metaphor" | PASS — PoP in raw/; line 1872 anchored directly. The 2026-04-11 PoP re-ingest log entry records this as one of three depth-preserving insights. The Marx 1844 Manuscripts reference is internal to MP's gloss in PoP, not a direct Marx citation. |
Summary
- Sample size: 20
- Pass: 20
- Flagged: 0
- Confirmed fabrications: 0
The Phase 6 spot-check passes cleanly. Every sampled citation traces to an extraction note or a verified raw/ source — typically both. No fabricated citations surfaced. The Faul fabrication-pattern hard-search returned zero hits in the wiki content layer.
R3 sampling caveat
Per AUDIT_PLAN R3: a clean biased sample is evidence that fabrications aren't concentrated where we'd expect, but is NOT certification that the corpus is fabrication-free. The estimated 400-600 citations across 234 pages were sampled at ~3-5%; the unsampled ~95% remain unverified by this audit.
The bias was strong: 50% of the sample (Tier 1) targets the highest-risk category (citations to authors not in raw/, since these cannot be verified against source within this audit). 30% (Tier 2) targets audit-produced pages — the post-Phase-0a-rule cohort that should obey the citation-traceability rule by construction. 20% (Tier 3) targets specific journal/page references on legacy pages.
Within these biased zones, the wiki is clean. This is a strong negative result for the high-risk zones, but the medium-risk zones (~95% of the corpus) remain unsampled. A future audit round should:
- Extend Tier 1 to include sampling of citations on pages updated between 2026-04-10 (wiki reset) and 2026-04-25 (Phase 0a commit) — the citations that pre-date the traceability rule.
- Add a Tier 4 (random sample) to test whether the bias is itself protective or whether it just confirms that fabrications cluster differently than expected.
What the spot-check did not test
- Quotation accuracy — citations were verified for existence of the cited passage, not for accuracy of the quoted text against the source. A spot-check that reads the cited passage in
raw/and verifies the quote against the wiki page would catch quotation drift; this audit did not perform that check. - Page-number accuracy — the wiki cites raw line numbers (machine-verifiable) and book page numbers (require verification against the printed edition). Line-number citations are pragmatically more verifiable; page-number citations to printed editions are taken on trust where the line-number anchor agrees.
- Citations to works in
raw/that are translations — translation drift between the cited French original and the English translation inraw/is not tested.
Phase 6 conclusion
Per AUDIT_PLAN Step 6.3 ("If zero problems in a sample of 15-25: spot-check passes, record negative result"), Phase 6 passes. No citation hygiene corrective action is required. The Faul-fabrication incident has not propagated into the wiki content layer. The Phase 0a citation-traceability rule (committed 2026-04-25) appears to be functioning as intended: audit-produced pages from Phases 1, 2, and 4 (Tier 2 of this sample) all trace cleanly to extraction notes or raw/ sources.