---
slug: dunning-kruger
passed: 2026-10-14
by: the assay line (research/assay-line/cited-1014b/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex
checked: 66
confirmed: 46
wrong: 18
unverifiable: 0
observed: 0
stale: 2
---

This was the directed correction pass over the 20 items in research/assay-line/cited-1014b/dunning-kruger/triage.json, with each item's evidence re-read before editing. It retains all 46 CONFIRMED claims and their qualifications from the Codex checker in N2-codex.md; those are the earlier checker's source readings, not 46 new independent investigations. The 7 MINOR lines count toward wrong alongside 11 WRONG lines; the 2 STALE lines remain separate. The authored page and model files matched the requested d9d5a28d18 baseline before editing. The film source and prose are corrected, but the rendered film remains uncorrected and the generated verification-date placard awaits landing regeneration. No source rejection or severity downgrade was used to pass the gate.

## Claims

- MINOR | X1: "Sorted into four skill quartiles, the textbook scissors reassembles itself from that noise: the bottom quartile rates itself 64.8 while it actually scores 12.5" ... "overestimating by +52.5 points" | research/dunning-kruger/film/film.html | Recreated the film model at seed 1234: perceived 64.84720426803831, actual 12.49062265566392, gap 52.35658161237439. The unrounded gap rounds to +52.4; the separately rounded inputs remain 64.8 and 12.5.
- MINOR | X2: "Restore self-insight and the scissors closes on its own (+52.5 → +19.3);" | research/dunning-kruger/film/film.html | Recreated SWEEP at seed 1234: bottom gaps 52.35658161237439 and 19.185510778134837 round to +52.4 and +19.2. The original caption used ensemble values.
- MINOR | X3: "matching the verifier's 200-seed means to a decimal;" | memory/log.d/2026-06-24T1837Z-film-dunning-kruger.md; research/dunning-kruger/film/film.html | Re-read the June 24 log: it gives the single-run bottom gap as +52.4 and the ensemble mean as +52.5. Close agreement holds; agreement to one decimal does not.
- WRONG | X4: "Scatter plot of every person's actual percentile against their self-rated percentile, with the best-fit line and correlation." | public/strata/dunning-kruger/index.html | Recreated simulate(): N=4000, step=floor(N/700)=5, samp.length=800. drawScatter draws samp. Correlation and fitted-line sums use every person before thinning.
- WRONG | X5: "every person · the raw cloud" | public/strata/dunning-kruger/index.html | The visible caption repeats X4. The drawing loop uses the 800-element sample, while the fitted line and correlation use all 4,000.
- WRONG | X6: "0 = they guess regardless of skill; 1 = they report their true percentile." | research/dunning-kruger/verify.mjs; public/strata/dunning-kruger/index.html | Re-read both model equations: rho=1 leaves bta + 8*G() and clamp(...,0,100). True ability percentile is the target before those terms, not the exact reported value.
- WRONG | X7: "This is the only dial that encodes a real psychological effect." | public/strata/dunning-kruger/index.html; research/dunning-kruger/verify.mjs | The other psychological control is explicitly labelled Better-than-average bias and enters the report as +bta. Test noise is a separate measurement parameter.
- MINOR | X8: "Most people, on most tasks, rate themselves a bit above average." | https://gwern.net/doc/psychology/1999-kruger.pdf | Re-read the 1999 primary paper, pp. 1121-1122: it describes overly favorable self-evaluation across many social and intellectual domains and discusses the above-average effect. It does not establish a majority of all people on a majority of all tasks.
- WRONG | X9: "The "skilled know less than they think; unskilled think they know more" correlation is partly an accounting identity." | https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:11831408%20AND%20SRC:MED&format=json&resultType=core | Re-fetched the primary 2002 abstract through Europe PMC: high scorers slightly underestimate their performance, while low scorers overestimate. The skilled clause in the page formulation reverses that direction.
- MINOR | X10: "That remains an open empirical question; see the three references and the discussion above." | public/strata/dunning-kruger/index.html | Re-read the ordered reference list: it contains four entries, ending with Krueger and Mueller (2002). This finding concerns the count, not the empirical dispute.
- STALE | X11: "Last verified 22 June 2026, when it was written." | memory/log.d/2026-06-24T1837Z-film-dunning-kruger.md; scripts/check-placard.mjs | The June 24 film log records a later verifier run with all five claims passing. June 22 was a supported earlier observation. The generated placard still awaits landing regeneration from this newly dated source record.
- MINOR | X12: "A seeded Monte Carlo (200 seeds, five claims) proves every number live." | public/strata/dunning-kruger/index.html; research/dunning-kruger/verify.mjs | Re-read fillCheck(): one run at seed 1234. The Node verifier uses 200 seeds for C1/C3/C4, 80 for symmetry, 60 for the C2 comparisons, and one for C5. The live page does not run all those ensembles.
- STALE | X13: ""dateModified":"2026-06-22"" | memory/log.d/2026-06-24T1837Z-film-dunning-kruger.md; public/strata/dunning-kruger/index.html | The June 24 log records that the newly added film leads the page. The June 22 modification date was superseded; this correction pass changes the page on 2026-10-14.
- WRONG | X14: ""uploadDate":"2026-07-09"" | git log d9d5a28d18 --diff-filter=A --format=%as -- public/strata/dunning-kruger/film.mp4; scripts/seo-heads.mjs | Read-only git log d9d5a28d18 --diff-filter=A --format=%as for the exact MP4 path returns 2026-06-24, commit 798d58ba05b. scripts/seo-heads.mjs defines uploadDate by that path-specific first git addition, not by a guessed deployment date.
- WRONG | X15: "The right-hand scatter shows the raw cloud and its near-zero correlation throughout, so the reader can see for themselves there is no relationship producing the shape." | public/strata/dunning-kruger/index.html; research/dunning-kruger/verify.mjs | Recreated the browser model with N=4000, seed=12345, bta=15 and sTest=0.5: r=-0.003940902637566944 at rho=0 and r=0.8421552104015267 at rho=1. Near zero applies to the zero-insight setting.
- WRONG | X16: "The same seeded model runs in the verifier at `research/dunning-kruger/verify.mjs` (mirrored to `public/research/dunning-kruger/`), which proves five claims across 200 seeds:" | https://artwaste.land/research/dunning-kruger/verify.mjs; https://artwaste.land/checks/research/dunning-kruger/verify.mjs | Re-fetched both URLs on 2026-10-14: /research/dunning-kruger/verify.mjs returned HTTP 404 and error 404; /checks/research/dunning-kruger/verify.mjs returned HTTP 200 and 15,988 bytes matching the pre-edit authored verifier.
- MINOR | X17: "The friendship paradox: your friends really do have more friends than you, and it's forced by sampling." | https://arxiv.org/abs/1304.3480 | Re-read the arXiv primary abstract: the friendship statement is explicitly qualified by on average and explained by overrepresentation of popular people in averages over friends.
- WRONG | X18: "The scissors has closed: with real self-insight and no bias, perceived ≈ actual." | public/strata/dunning-kruger/index.html | Recreated rho=1, bta=5, sTest=0, seed=12345: bottom gap 4.601782469941902, top gap 3.3849125621897542, r=0.9649625617669614. The old branch fires at nonzero bias because it tests only the two gaps.
- WRONG | X19: "Plot overestimation against skill and the slope is −1 by construction" ... "skill sits on both axes. That is the autocorrelation hidden in the chart." | research/dunning-kruger/film/film.html | Re-read the film timeline: the sweep reaches rho=1 at 120 seconds and the caption begins after 120.5 seconds. Recreated film slopes: -0.9899166056683978 at rho=0 and -0.17585618236272627 at rho=1. Algebra gives Cov(self-rating,score)/Var(score)-1. The authored source and page text are corrected, but the rendered MP4 remains uncorrected under the explicit no-render instruction.
- WRONG | X20: (added) "verifier" links to "../../../research/dunning-kruger/verify.mjs". | https://artwaste.land/research/dunning-kruger/verify.mjs; https://artwaste.land/checks/research/dunning-kruger/verify.mjs | The film-caption href resolves to the same HTTP 404 destination as X16; the positive control at /checks/ returns the published verifier. Re-read both the authored href and fetched responses.
- CONFIRMED | "Incompetent people are too incompetent to know they're incompetent, so they rate themselves sky-high; the truly skilled are racked with doubt." | https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | The page explicitly introduces this as "The story". Dunning describes poorer performers overrating themselves and experts underrating their skills socially; this supports the narrative being discussed, rather than a measurement of experts feeling doubt. [entry 2: popular-story]
- CONFIRMED | "The original 1999 study is real, and people do misjudge themselves." | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 4: original-study-real]
- CONFIRMED | "But the dramatic scissors shape" ... "the thing the chart is famous for" ... "is largely a statistical artefact." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | Nuhfer identifies "artifacts that invite misinterpretation"; Gignac and Zajenkowski call the effect "(mostly) a statistical artefact". The page explicitly limits its conclusion to what the chart alone establishes. [entry 5: shape-largely-artefact]
- CONFIRMED | "almost exactly the published 62nd / 12th (Kruger & Dunning, 1999)." | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 14: film-published-values]
- CONFIRMED | "the overestimation-vs-skill slope is −1 by construction." | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 16: film-slope]
- CONFIRMED | "It does not claim the effect is zero" ... "that stays unsettled." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/; https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | Nuhfer reports disagreement; Gignac allows that "the magnitude of the effect may be much smaller than reported previously"; Dunning argues that "the effect is still real". Read as the disputed psychological Dunning-Kruger component, rather than a denial of all metacognition. [entry 17: film-unsettled]
- CONFIRMED | "No test measures ability perfectly." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/; https://gwern.net/doc/psychology/1999-kruger.pdf | Nuhfer discusses reliability and measurement "noise"; the original paper also treats imperfect measurement and regression. The N1 excerpt concerns self-assessment noise specifically, so it is narrower than the page statement about ability tests. [entry 29: tests-imperfect]
- CONFIRMED | "This noise is what drives regression to the mean" ... "the extreme quartiles are partly luck, and luck doesn't repeat." | git show 632d05207f:research/dunning-kruger/verify.mjs | The model uses "x[i] = ai + sTest*G();" and cuts quartiles on xp. Recreated with rho=1, bta=0, sTest=0.7: bottom gap 7.99, top gap -7.04. The quoted N1 abstract alone does not establish this test-noise mechanism; the foundation does. [entry 30: test-noise-regression]
- CONFIRMED | "If confidence is unrelated to skill, then average self-rating is the same in every quartile" ... "a flat line near the population mean." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 35: flat-quartiles]
- CONFIRMED | "But average actual score, by definition of "quartile," climbs a staircase from ~12th to ~88th percentile." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 36: staircase-percentiles]
- CONFIRMED | "A flat line laid across a rising staircase must cross: below the crossing everyone looks overconfident, above it everyone looks humble." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 37: must-cross]
- CONFIRMED | "That is the scissors, and it is pure geometry." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 38: pure-geometry]
- CONFIRMED | "Set ρ = 0 and watch the perceived line go dead flat." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 39: dead-flat]
- CONFIRMED | "Tests are noisy." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/; https://gwern.net/doc/psychology/1999-kruger.pdf | Nuhfer discusses reliability and measurement "noise"; the original paper also treats imperfect measurement and regression. The N1 excerpt concerns self-assessment noise specifically, so it is narrower than the page statement about ability tests. [entry 40: noisy-tests]
- CONFIRMED | "The people who scored in the bottom quartile include plenty who were merely unlucky; their true ability is higher than their score, so any honest self-read lands above their measured rank" ... "instant "overconfidence." The top quartile is the mirror image." | git show 632d05207f:research/dunning-kruger/verify.mjs | The model uses "x[i] = ai + sTest*G();" and cuts quartiles on xp. Recreated with rho=1, bta=0, sTest=0.7: bottom gap 7.99, top gap -7.04. The quoted N1 abstract alone does not establish this test-noise mechanism; the foundation does. [entry 41: regression-explanation]
- CONFIRMED | "The classic way to quantify the effect is to plot overconfidence (self-rating minus score) against score." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/ | Nuhfer explicitly lists "(y minus x) vs. (x) scatterplots" among the relevant conventions, alongside quartile line charts. The claim concerns quantifying the gap, not the only historical display. [entry 44: classic-quantification]
- CONFIRMED | "But score now sits on both axes, with opposite signs." | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 45: opposite-sign-axes]
- CONFIRMED | "Even with self-ratings that are literally random, that regression has a slope of almost exactly −1" ... "built in, not discovered." | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 46: random-slope-minus-one]
- CONFIRMED | "The quartile chart at zero correlation, beside the figure published by Kruger & Dunning in 1999:" | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 50: published-table]
- CONFIRMED | "almost exactly the famous "62nd vs 12th" of the original paper." | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 52: published-pair-quote]
- CONFIRMED | "People are miscalibrated about themselves; the original measurements are real." | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 55: people-miscalibrated]
- CONFIRMED | "What the artefact dissolves is the strong, asymmetric, deficit-of-the-incompetent story" ... "the claim that the unskilled are uniquely, specially blind." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/ | Nuhfer says self-assessments "reflect a genuine competence that they can demonstrate" and also says experts self-assess better than novices. This supports weakening the strong general deficit story, not erasing every expert/novice difference. [entry 56: deficit-story-dissolved]
- CONFIRMED | "Whether any genuine metacognitive signal survives once the statistics are removed is honestly unsettled." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/; https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | Nuhfer reports disagreement; Gignac allows that "the magnitude of the effect may be much smaller than reported previously"; Dunning argues that "the effect is still real". Read as the disputed psychological Dunning-Kruger component, rather than a denial of all metacognition. [entry 58: metacognition-unsettled]
- CONFIRMED | "Nuhfer and colleagues, using measures designed to dodge the artefact, find the strong tendency largely vanishes." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/ | Nuhfer says self-assessments "reflect a genuine competence that they can demonstrate" and also says experts self-assess better than novices. This supports weakening the strong general deficit story, not erasing every expert/novice difference. [entry 59: nuhfer-vanishes]
- CONFIRMED | "Gignac and Zajenkowski call it "(mostly)" an artefact" ... "leaving room for a small real component." | https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | The title contains "(mostly)" and the abstract concludes "the magnitude of the effect may be much smaller than reported previously." This leaves room for a component rather than proving it is zero. [entry 60: gignac-mostly-quote]
- CONFIRMED | "Dunning himself maintains the self-misjudgements are real whatever their cause." | https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | Dunning says "the pattern of self-misjudgements remains regardless of what may be producing it." Attribution confirmed. [entry 61: dunning-rebuttal]
- CONFIRMED | "It settles a narrower, sturdier thing: the chart is not the evidence it is usually taken to be." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | Nuhfer identifies "artifacts that invite misinterpretation"; Gignac and Zajenkowski call the effect "(mostly) a statistical artefact". The page explicitly limits its conclusion to what the chart alone establishes. [entry 62: chart-is-not-evidence]
- CONFIRMED | "When someone shows you the scissors, they have shown you regression to the mean wearing a lab coat." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | Nuhfer identifies "artifacts that invite misinterpretation"; Gignac and Zajenkowski call the effect "(mostly) a statistical artefact". The page explicitly limits its conclusion to what the chart alone establishes. [entry 63: chart-regression-labcoat]
- CONFIRMED | "that the overconfidence-vs-score regression has slope ≈ −1 by construction;" | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 66: proven-slope]
- CONFIRMED | "The "62nd vs 12th" figure is the widely-cited summary of Kruger & Dunning's bottom quartile, read from the published figures, not re-derived here." | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 71: pair-figure-source]
- CONFIRMED | "Kruger, J. & Dunning, D. (1999). "Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments." Journal of Personality and Social Psychology 77(6): 1121–1134." | https://api.crossref.org/works/10.1037%2F0022-3514.77.6.1121 | Crossref records Justin Kruger and David Dunning, 1999, volume 77, issue 6, pages "1121-1134", and the printed title. Confirmed independently on the primary paper. [entry 73: citation-kruger-1999]
- CONFIRMED | "Nuhfer, E., Cogan, C., Fleisher, S., Gaze, E. & Wirth, K. (2016); Nuhfer, E. et al. (2017). Numeracy 9(1) & 10(1)" ... "self-assessment measured with statistically valid instruments." | https://digitalcommons.usf.edu/numeracy/vol9/iss1/art4/ | The publisher lists "Edward Nuhfer, Christopher Cogan, Steven Fleisher, Eric Gaze, and Karl Wirth", Numeracy 9(1), 2016, article 4. The page initials and author order match. [entry 74: citation-nuhfer-2016]
- CONFIRMED | "Nuhfer, E., Cogan, C., Fleisher, S., Gaze, E. & Wirth, K. (2016); Nuhfer, E. et al. (2017). Numeracy 9(1) & 10(1)" ... "self-assessment measured with statistically valid instruments." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/ | The publisher lists "Steven Fleisher" and Numeracy 10(1), 2017, article 4. Crossref calls him "Steven Fleischer", an error in that metadata, not a page misspelling. The page abbreviates the 2017 author list as et al. [entry 75: citation-nuhfer-2017]
- CONFIRMED | "Gignac, G. E. & Zajenkowski, M. (2020). "The Dunning–Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data." Intelligence 80: 101449." | https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | UWA gives "Gignac, G. E., & Zajenkowski, M. (2020)", the exact title, "Intelligence, 80, Article 101449" and the cited DOI. [entry 76: citation-gignac-2020]
- CONFIRMED | "Krueger, J. & Mueller, R. A. (2002). "Unskilled, unaware, or both? The better-than-average heuristic and statistical regression predict errors in estimates of own performance." JPSP 82(2): 180–188." | https://api.crossref.org/works/10.1037%2F0022-3514.82.2.180 | Crossref gives Joachim Krueger and Ross A. Mueller, 2002, volume 82, issue 2, pages "180-188", and the page title. Europe PMC also confirms the citation. [entry 77: citation-krueger-2002]
- CONFIRMED | "It is three pieces of ordinary statistics wearing a lab coat: a flat line crossing a rising staircase, regression to the mean, and an autocorrelation hidden in the axes (the overconfidence-vs-score regression has slope −1 by construction)." | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 86: dek-statistical-mechanisms]
- CONFIRMED | "That is genuinely unsettled (Nuhfer; Gignac & Zajenkowski 'mostly'; Dunning's rebuttal)." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/; https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | Nuhfer reports disagreement; Gignac allows that "the magnitude of the effect may be much smaller than reported previously"; Dunning argues that "the effect is still real". Read as the disputed psychological Dunning-Kruger component, rather than a denial of all metacognition. [entry 88: dek-unsettled]
- CONFIRMED | "A third, test noise, drives regression to the mean." | git show 632d05207f:research/dunning-kruger/verify.mjs | The model uses "x[i] = ai + sTest*G();" and cuts quartiles on xp. Recreated with rho=1, bta=0, sTest=0.7: bottom gap 7.99, top gap -7.04. The quoted N1 abstract alone does not establish this test-noise mechanism; the foundation does. [entry 101: md-noise]
- CONFIRMED | "the overconfidence-vs-score slope is ≈ −1 by construction;" | git show 632d05207f:research/dunning-kruger/verify.mjs | The equation is "const px = xp, py = sp.map((v,i)=>v-xp[i]);". Thus slope=Cov(sp,xp)/Var(xp)-1; under independence it is -1 in the population and approximately -1 in the sample. The zero-insight context matters. [entry 106: md-slope]
- CONFIRMED | "Anscombe & the Datasaurus show identical summary statistics hiding wildly different shapes;" | https://www.research.autodesk.com/publications/same-stats-different-graphs/ | The authors describe "four clearly different and visually distinct datasets" with the same statistics, and their Datasaurus family has matching mean, SD and correlation "to two decimal places". The source supports the standard rounded-summary-statistics statement. [entry 109: related-anscombe-datasaurus]
- CONFIRMED | "A flat self-rating laid across a rising staircase must cross. The bottom looks overconfident; the top, humble." | git show 632d05207f:research/dunning-kruger/verify.mjs | Ranks use "(k/(n-1))*100". Four equal groups have means 12.4906, 37.4969, 62.5031, 87.5094. Independent self-ratings have a flat conditional expectation; their 50 to 80 model mean lies between the extreme quartile means. Finite samples fluctuate around that flat expectation. [entry 131: film-canvas-crossing]
- CONFIRMED | "The clueless, wildly overconfident. The experts, modestly unsure." | https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | The page explicitly introduces this as "The story". Dunning describes poorer performers overrating themselves and experts underrating their skills socially; this supports the narrative being discussed, rather than a measurement of experts feeling doubt. [entry 138: film-canvas-story]
- CONFIRMED | "It does not say people are well-calibrated. It does not say the psychological effect is zero" ... "that is genuinely unsettled." | https://digitalcommons.usf.edu/numeracy/vol10/iss1/art4/; https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/; https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents | Nuhfer reports disagreement; Gignac allows that "the magnitude of the effect may be much smaller than reported previously"; Dunning argues that "the effect is still real". Read as the disputed psychological Dunning-Kruger component, rather than a denial of all metacognition. [entry 139: film-canvas-unsettled]
- CONFIRMED | "62nd perceived / 12th actual" | https://gwern.net/doc/psychology/1999-kruger.pdf | The primary abstract says "12th percentile, they estimated themselves to be in the 62nd." Original page 1121 confirms both numbers. The numerical shorthand is accurate, but is not a verbatim quotation. [entry 140: film-canvas-published-pair]
- CONFIRMED | "leaving room for a small real component." | https://research-repository.uwa.edu.au/en/publications/the-dunning-kruger-effect-is-mostly-a-statistical-artefact-valid-/ | The title contains "(mostly)" and the abstract concludes "the magnitude of the effect may be much smaller than reported previously." This leaves room for a component rather than proving it is zero. [entry 142: gignac-small-component]
- CONFIRMED | (added) "og:image:width" content="1200" and "og:image:height" content="630". | https://artwaste.land/og/dunning-kruger.png | Curl returned HTTP 200. PNG IHDR bytes decode to width=1200, height=630. The same decoder positively identified the PNG signature. [entry 146: added-og-dimensions]

## Findings

- [fixed] X1: Changed the film caption to +52.4; kept the independently rounded 64.8 and 12.5. MINOR, no public correction line.
- [fixed] X2: Changed the film-caption endpoints to +52.4 → +19.2. MINOR, no public correction line.
- [fixed] X3: Changed decimal equality to close agreement between the simulated quartile values and the verifier ensembles. MINOR, no public correction line.
- [fixed] X4: Changed the scatter accessibility label to a sample of 800 of 4,000 and retained the full-population scope of the fit and correlation. Shared correction line with X5.
- [fixed] X5: Changed the visible scatter caption to 800 of 4,000 people and the sampled cloud; also made the markdown sample explicit. Shared correction line with X4.
- [fixed] X6: Changed the page hint, README and verifier model comments to true ability percentile as a target before bias, reporting noise and clipping. Regenerated the published verifier copy from the source. Added the matching correction line.
- [fixed] X7: Replaced the sole-psychological-dial claim with the two psychological parameters and the separate test-noise role. Added a correction line.
- [fixed] X8: Narrowed the better-than-average tendency to many social and intellectual domains, without a majority quantifier. MINOR, no public correction line.
- [fixed] X9: Preserved the quoted words intact, identified the skilled clause as backwards, stated the correct direction outside the quotation and linked reference 4. Added a correction line.
- [fixed] X10: Changed three references to the references. MINOR, no public correction line.
- [dated] X11: Set passed: 2026-10-14 in this authored record for the landing build to regenerate Last verified. The generated page placard still says June 22 until that build. Applied the triage placard-date exception: no Updated line and no hand-edit of generated markup.
- [fixed] X12: Distinguished the live single-seed model from separate Node ensembles in the dek, check explanation, markdown and verifier comments. Described the verifier as using up to 200 seeds per comparison. MINOR, no public correction line.
- [dated] X13: Changed Article dateModified to 2026-10-14 and preserved datePublished. No Updated line under the explicit JSON-LD exception; the Corrected lines already date the revision.
- [fixed] X14: Changed VideoObject uploadDate from 2026-07-09 to 2026-06-24, using path-specific git history and the existing metadata rule. Added a correction line without guessing the error cause.
- [fixed] X15: Restricted near-zero correlation to zero self-insight in the markdown and stated that it changes with the dials. Added the same correction line to page and markdown.
- [fixed] X16: Changed the documented verifier mirror to public/checks/research/dunning-kruger/. Shared correction line with X20.
- [fixed] X17: Added on average to this layer's related-layer friendship summary. MINOR, no public correction line; the related layer itself was not edited.
- [fixed] X18: The live branch now reports only that the bottom and top quartile means are each within 8 points of their actual means. It no longer infers no bias from those gaps. Added a correction line.
- [open note:b52053] X19: Authored source and page prose only: conditioned the minus-one slope in the film source, film caption, apparatus, dek and markdown. The public correction explicitly states that the existing film still has the old caption. NOT FIXED in the rendered MP4; rendering and re-encoding were prohibited. A board note id is pending.
- [fixed] X20: Changed the film-caption verifier link to /checks/research/dunning-kruger/verify.mjs. Shared correction line with X16.

## Not done

- X19: public/strata/dunning-kruger/film.mp4 still contains the unconditional minus-one caption after about 02:00.5. Its authored caption is corrected, but no film, audio or image was rendered or re-encoded.
- X11: the generated Last verified placard still shows 22 June 2026. The new passed date is in this source record for the landing build; no shared placard writer or site build was run.
- This was the directed triage pass. Findings in the earlier checker report that the directing triage did not include were not silently added to its work list.

## Verification

- `node research/dunning-kruger/verify.mjs`: exit 0, 14 assertions pass across C1-C5. Its two artefact files changed only their generated date to 2026-10-14.
- Existing `verify-dunning-kruger.mjs` run through a scratch copy with this checkout's Playwright path, the authored public directory and an ephemeral local port: exit 0, all 18 original assertions pass. Ten additional checks for corrected descriptions, verdicts, dates, correction lines and the verifier link also pass. MP4 requests were disabled; no film was played or rendered.
- `node scripts/check-claims-pass.mjs`: exit 1 only for another layer, the-dancing-plague, with malformed finding dispositions at findings 1, 3 and 18. No dunning-kruger error was reported; that other record was not edited.
- Direct `checkRecord('dunning-kruger', ...)`: no errors; checked 66, confirmed 46, wrong 18 including MINOR, stale 2, unverifiable 0, observed 0.
- Page and film-source JavaScript parse; JSON-LD parses; the nine correction texts match between HTML and markdown and precede the check placard; the published verifier equals its authored source byte for byte. Scoped `git diff --check` passes.