Citations
Every footnote on every /about/science/* page resolves to an entry below. The registry is the single source of truth for methodological citations across all three audience tiers — public science pages, reviewer-grade infrastructure pages, and gated reports.
The list is alphabetical by author. Each entry includes the publication year, the venue, a primary-source URL, and a one-sentence "what this paper gave us" annotation that explains its load-bearing role in our work.
Anglim, J.; Knowles, E. R. V.; Dunlop, P. D.; Marty, A. (2017).
HEXACO Personality and Schwartz's Personal Values: A Facet-Level Analysis.
Journal of Research in Personality 68: 23–31; OSF deposit wkc5u.
N=1,244 Australian adults, HEXACO × Schwartz PVQ-RR. One of three Anglim deposits (with 2020 and 2024) that backbone the Tier-0 empirical layer of our Σ assembly. CC-BY 4.0.
Primary source ↗Anglim, J.; Sojo, V.; Ashford, L. J.; Newman, A.; Marty, A. (2020).
Predicting employee attitudes to workplace diversity from personality, values, and cognitive ability.
Journal of Research in Personality 84; OSF deposit uwdgs.
N=1,370 Australian adults, HEXACO × trait emotional intelligence (Schutte SEIS) × cognitive ability. Provides the HEXACO × EI Tier-0 cells for Σ. CC-BY 4.0.
Primary source ↗Anglim, J.; Marty, A. (2024).
HEXACO Personality, Values, and Cognitive Ability in a Large Sample of Job Applicants.
Social Psychological and Personality Science (in press); OSF deposit 2ah5z.
N=15,552 Australian job applicants, HEXACO × Schwartz PVQ-RR × cognitive ability. Largest Tier-0 source by sample size in our Σ assembly. CC-BY 4.0.
Primary source ↗Cheung, G. W.; Rensvold, R. B. (2002).
Evaluating goodness-of-fit indexes for testing measurement invariance.
Structural Equation Modeling 9(2): 233–255.
ΔCFI ≤ 0.01 threshold for measurement-invariance verdicts in hierarchical CFA. We implement Gate 10 (functions/synthesis/calibration/invariance.py) against this threshold, with the Meredith 1993 / Vandenberg & Lance 2000 / Millsap 2011 framework as the surrounding methodology.
Primary source ↗Condon, D. M.; Revelle, W. (2017).
The SAPA Personality Inventory: an empirically-derived, hierarchically-organized self-report personality inventory.
Journal of Open Psychology Data 5(1).
N≈4,000, SAPA Personality Inventory (SPI) cross-instrument correlations, distributed via the psychTools R package. GPL-2 CRAN tarball. One of five Tier-0 datasets in our Σ assembly.
Primary source ↗Demarta, S.; McNeil, A. J. (2005).
The t copula and related copulas.
International Statistical Review 73(1): 111–129.
Provides the closed-form upper-tail dependence coefficient λ_U(ρ, ν) = 2·t_{ν+1}(−√((ν+1)(1−ρ)/(1+ρ))) that we use to validate the block-t copula on the Dark-Triad triplet at ν=5.
Primary source ↗Embrechts, P.; McNeil, A. J.; Straumann, D. (2002).
Correlation and Dependence in Risk Management: Properties and Pitfalls.
Risk Management: Value at Risk and Beyond, Cambridge University Press.
The canonical demonstration that linear correlation is a brittle dependence summary outside elliptical distributions. Justifies our explicit block-t treatment of the Dark-Triad and our Henze-Zirkler MVN test on the latent layer.
Primary source ↗Embretson, S. E.; Reise, S. P. (2000).
Item Response Theory for Psychologists.
Lawrence Erlbaum Associates / Routledge.
Standard contemporary IRT reference. Chapter 12's discussion of measurement invariance under recalibration is the methodological warrant for our 'recalibrate against real-user data, not synthetic' Phase-3 commitment.
Primary source ↗Friesdorf, R.; Conway, P.; Gawronski, B. (2015).
Gender differences in responses to moral dilemmas: a process dissociation analysis.
Personality and Social Psychology Bulletin 41(5): 696–713.
Sex-d source for moral foundations (Care, Fairness, Loyalty, Authority, Sanctity, Liberty). Women score higher on Care; Schwartz Self-Transcendence shows similar patterns. Used in conditioning step alongside Schwartz & Rubel 2005.
Primary source ↗Higham, N. J. (2002).
Computing the nearest correlation matrix — a problem from finance.
IMA Journal of Numerical Analysis 22(3): 329–343.
Iterative projection algorithm we apply to the assembled four-tier Σ to guarantee positive definiteness while minimising Frobenius distance from the empirical+meta-analytic prior. We achieve relative Frobenius distance ≈ 0.05 on the locked manifest.
Primary source ↗Joe, H. (2014).
Dependence Modeling with Copulas.
Chapman & Hall/CRC Monographs on Statistics & Applied Probability.
The standard contemporary reference on copula families. We follow Joe's taxonomy when choosing Gaussian copula for the bulk of trait dependence and switching to t-copula on the Dark-Triad block.
Primary source ↗Jones, D. N.; Paulhus, D. L. (2014).
Introducing the Short Dark Triad (SD3): a brief measure of dark personality traits.
Assessment 21(1): 28–41.
Validation paper for the SD3 instrument used in the OSP raw data. Authors' choice of three correlated-but-distinct dark traits (Machiavellianism, Narcissism, Psychopathy) is what makes the block-t copula the right tool: the joint upper tail co-occurrence is what a pure Gaussian copula systematically under-samples.
Primary source ↗Mickelson, K. D.; Kessler, R. C.; Shaver, P. R. (1997).
Adult attachment in a nationally representative sample.
Journal of Personality and Social Psychology 73(5): 1092–1106.
Base-rate proportions for the four-style adult attachment taxonomy (Secure, Avoidant, Anxious, Fearful). Our GMM-BIC validation on (ECR_Anx, ECR_Avo) targets these proportions, though Stage D Mickelson-calibrated thresholds are an open Phase-2 hardening item.
Primary source ↗Open-Source Psychometrics Project; Jones, D. N.; Paulhus, D. L. (2014).
Short Dark Triad (SD3): a brief measure of dark personality traits.
Assessment 21(1): 28–41 (instrument); openpsychometrics.org raw data.
N=18,192 web volunteers, Short Dark Triad (Machiavellianism, Narcissism, Psychopathy). Provides the DT × DT inner-block correlations and the Tier-0 empirical anchor for the block-t copula. Public-domain raw data; instrument cited per Jones & Paulhus 2014.
Primary source ↗Roberts, B. W.; Walton, K. E.; Viechtbauer, W. (2006).
Patterns of mean-level change in personality traits across the life course: a meta-analysis of longitudinal studies.
Psychological Bulletin 132(1): 1–25.
Per-decade z-slope source for our age conditioning. Conscientiousness, Agreeableness, and Emotional Stability rise with age; Openness and Extraversion are roughly stable. Drives the additive age shift on the latent layer before the probability integral transform.
Primary source ↗Samejima, F. (1969).
Estimation of latent ability using a response pattern of graded scores.
Psychometrika Monograph Supplement 17.
The graded-response model itself. We fit a hierarchical Bayesian variant via NumPyro NUTS, with one item per PSL rule, five categories per construct, and per-family pooling on log-discrimination.
Primary source ↗Schmitt, D. P.; Realo, A.; Voracek, M.; Allik, J. (2008).
Why can't a man be more like a woman? Sex differences in Big Five personality traits across 55 cultures.
Journal of Personality and Social Psychology 94(1): 168–182.
Cross-cultural Cohen's d source for sex differences on the Big Five. Largest effects on Agreeableness (~0.40) and Emotional Stability (~−0.50, women report lower). We use these as the seed values for the additive sex shift in the persona sampler.
Primary source ↗Sklar, A. (1959).
Fonctions de répartition à n dimensions et leurs marges.
Publications de l'Institut de Statistique de l'Université de Paris 8: 229–231.
The foundational result that any joint distribution can be decomposed into a copula plus its marginals. Underwrites the entire copula sampler architecture: we get to model trait dependence (Σ) and per-trait shape independently.
Primary source ↗Vovk, V.; Gammerman, A.; Shafer, G. (2005).
Algorithmic Learning in a Random World.
Springer.
Foundational text for conformal prediction. Our conformal recalibration step (functions/synthesis/calibration/conformal_fitter.py) follows the split-conformal recipe; the calibration band is guaranteed to cover the truth at the rate we promise without asymptotic hand-waving.
Primary source ↗
Methodology of the registry
The registry lives at docs/citations.json (the <CitationsList /> component above renders it). Every entry has:
authors— the author list as it appears on the original publication.year— publication year.title— the paper or book title.venue— the journal, conference, or publisher.url— a primary-source URL (DOI link, OSF deposit, publisher page, or open-access PDF).gloss— a one-sentence annotation that names what we use the work for.
The registry is the only place citation strings are authored in this site. Pages reference entries by short key (e.g., sklar1959), and the <Footnote> component looks up the entry, renders the popover, and surfaces the gloss. Adding a new citation is a one-line addition to docs/citations.json; it then becomes available to every page that imports the typed registry.
What is not here
The citations page is the bibliography for methodological claims. It does not include:
- The proprietary PSL rule library — the rule weighting was informed by the work above, but the specific rules are not published in any of these papers.
- The exact construct → trait weight matrix — the methodology is documented (statistical fits against published instrument correlations), but the per-cell magnitudes are proprietary and gated to the reports tier.
- The exact thinking-budget values used in offline corpus generation — these are specific cost-engineering choices, not methodological claims, and live in the gated R4 FinOps report.
- Internal engineering decisions like the Higham 2002 nearest-PD projection's tolerance threshold or the conformal coverage target tolerance — these are documented in the gated reports, not the citations.
A methodological claim is a published-literature claim. An engineering choice is a build-versus-buy decision. The citations registry is for the former; the reports tier is for the latter.
If you find a claim on this site that should resolve to a citation but does not, the omission is a defect — please flag it. The registry is meant to be exhaustive for methodological claims; gaps are bugs, not redactions.