Principia↗ peopleprincipia.com

Customer

Builders, researchers, and operators working on organizational measurement who need a single source-graded place to look up constructs, instruments, reliability/validity evidence, and effect sizes — alongside the People Analytics Platform consuming canonical priors over MCP.

Problem · external

The same construct is measured five different ways across five different studies; effect-size tables are scattered through textbook chapters; high-quality instruments get reinvented in low-quality form; and the existing measurement-handbook tradition synthesizes at publication date rather than continuously.

Problem · internal

You can't tell whether the engagement number on the dashboard means what the field thinks engagement means, because the literature you'd need to check is fragmented across decades and you don't have a week to assemble it.

Problem · philosophical

Load-bearing measurement decisions should be defensibly grounded in the source-graded primary literature, not in the closest survey vendor's marketing copy.

Guide

Principia is the continuously-updated, source-graded, citation-verified, Bayesian-prior-bearing registry of organizational science — a survey-not-original-research curation layer that sits on top of meta-factory's extraction pipelines and feeds canonical priors to the rest of the portfolio.

Plan

(1) Consume `@measurement/core` as the canonical schema spine, (2) ingest construct-family-by-construct-family with source grading and effect-size tables, (3) expose the registry over a versioned REST + MCP contract so consumers vendor typed reads and don't re-implement.

Success

The customer looks up a construct once and sees the instrument inventory, item-level schemas, graded reliability/validity evidence, meta-analytic effect-size table, and the Bayesian prior synthesized from it — with provenance preserved end-to-end.

Failure avoided

Measurement decisions made on tertiary summaries, instruments reinvented in degraded form, effect sizes pooled across incomparable studies, and a portfolio of products that each carries its own slightly-drifted definition of the same construct.

The same construct is measured five different ways across five different studies; effect-size tables are scattered through textbook chapters; high-quality instruments get reinvented in low-quality form; and the existing measurement-handbook tradition synthesizes at publication date rather than continuously. You can't tell whether the engagement number on the dashboard means what the field thinks engagement means, because the literature you'd need to check is fragmented across decades and you don't have a week to assemble it. Load-bearing measurement decisions should be defensibly grounded in the source-graded primary literature, not in the closest survey vendor's marketing copy.

Hub-and-spoke around `@people-analyst/measurement-core` — the canonical vocabulary (Construct, Instrument, Item, Measure, Model, EffectSize, Publication, StudyQualityGrade, CanonicalPrior) lives in the shared package; Principia is the canonical registry of rows, not the canonical vocabulary. Tightly coupled with meta-factory upstream — asset IDs, variable codes, and classification flow forward into the registry; the registry is downstream of extraction, not parallel to it. Public reader UI live at `peopleprincipia.com/registry/*` (PRN-038a–d: scaffold, cross-entity linking, distribution plots, JSON-LD + OG + partitioned sitemap + Vercel Analytics) with the CanonicalTheoreticalModel layer (PRN-037, schema + REST + 8 seed theories), instrument-equivalence networks (PRN-042, 5 engagement-family seeds), validity-decay + MethodologyCritique + cultural sub-priors (PRN-043/044/045), and a construct → "valid & reliable measures" surface (PRN-040). Curator admin console at `/admin/*` (PRN-029): citations, queues, watchlists, plus a promote-effect-size CLI (PRN-030, D4-compliant write path) and novelty-verification queue (PRN-032). VoI as a REST endpoint (PRN-046) plus an intervention library (PRN-047). Book-build orchestrator (PRN-041a) with XeLaTeX + HTML5 + APA-7 CSL templates, against the "Principia of Organization Measurement" outline (PRN-041). Engagement-family ingest pipeline first; additional construct families follow as the substrate matures.

• 01Continuous citation-verified curation, not point-in-time handbook synthesis — the registry refreshes as the literature does, with provenance preserved end-to-end.
• 02Source grading as a first-class field on every claim — StudyQualityGrade lives in the schema, so downstream consumers can weight evidence rather than treat all citations equally.
• 03Bayesian-prior bearing — canonical priors are a first-class output of the registry, not a downstream interpretation; the rest of the portfolio vendors typed priors instead of re-fitting from scratch. First live CanonicalPrior is engagement → task performance, synthesized from Corbeanu 2023 + Neuber 2021 (k=3); Mazzetti 2023 promoted engagement → satisfaction (r=.60) + engagement → commitment (r=.63); Christian, Garza & Slaughter 2011 resynthesizes engagement → performance.
• 04CanonicalTheoreticalModel layer as Pillar-1 — schema + handlers + REST + 8 seed theories (PRN-037), with instrument-equivalence networks (PRN-042) linking measures across nominally-distinct instruments. The vocabulary recognizes that the same construct gets measured by different instruments that should be calibrated against each other, not treated as separate variables.
• 05Validity-decay + MethodologyCritique + cultural sub-priors (PRN-043/044/045) — explicit modeling of how evidence ages, how methodology shapes finding strength, and how priors differ by cultural context. The registry doesn't pretend universality where the literature doesn't support it.
• 06Survey-not-original-research positioning — Principia does not run new studies; it curates the existing literature into a legible, queryable, source-graded registry.
• 07MCP-native distribution — the registry is consumable by AI agents over Model Context Protocol from day one, not via a UI someone else's product sits on top of. VoI exposed as a single-tuple Bayesian-prior shortcut at `/api/v1/voi` for callers that just need the next-best-measurement decision.
• 08Book-build orchestrator (PRN-041a) — the registry contents compose into "The Principia of Organization Measurement" via directive expansion + appendix auto-builders + XeLaTeX/HTML5/APA-7 templates. The registry is its own primary source.

1. 2026-04-30Initial scaffold v1: research surface, AGENTS, methodology, roadmap (17190d5)
2. 2026-05-06Mike-voice reference pointer (PA-017) (fbedaa6)
3. 2026-05-11PRN-001..008 assignment queue + first-wave research entries (d7992ed)
4. 2026-05-13Portfolio SOTA snapshot landed in docs/SNAPSHOTS/2026-05-13-snapshot.md (2b3ffd1, PR #1 merged)
5. 2026-05-18Canonical engineering + portfolio spec v0.1 landed (docs/specification/SPEC.md) with PRN-010..015 engineering queue (3ad2b25)
6. 2026-05-19ASN-1013 closed: consume @measurement/core as the schema spine (e839d30); tight-coupling-with-meta-factory codified in AGENTS.md (da4c7e8)
7. 2026-05-19PRN-010 scaffolded packages/core + packages/registry; PRN-011 consumed Deltas 4-5 from @measurement/core v0.2 (2d9f144, 61e10bf)
8. 2026-05-19PRN-013 JSON-file registry store + CanonicalVariable resolver (6b66e84)
9. 2026-05-19PRN-012 (partial) engagement-family ingest adapters: 147 source constructs → 68 canonical variables, 38 instruments, 72 citations (aa98246)
10. 2026-05-19Bayesian-prior reframe: Deltas 6-9 + continuous-loop + 12 new assignments (d01d3ab)
11. 2026-05-19PRN-014 MCP tools + apps/web Next.js scaffold + /api/health + /api/mcp (37d6d90)
12. 2026-05-19PRN-014c follow-ons: namespace rename to @people-analyst/measurement-core@^0.5.1, .npmrc for Vercel build root (8d3e2de, 7803b89)

• ·PRN-014c — apps/web Vercel deploy so /api/health returns 200 from a public URL (open PR refactor/measurement-core-namespace-rename); unblocks toolbox PAT-114 coming-soon → live flip
• ·PRN-012 full — extend ingest beyond engagement to Tier 1 families (job satisfaction, organizational commitment, burnout, psychological safety) per docs/research/PROGRAM.md order
• ·PRN-014b + PRN-015 — full REST surface (GET /constructs, /instruments, /effects) + per-consumer auth keys + audit-table upgrade beyond JSON-line
• ·PRN-019 + PRN-020 — continuous-loop infrastructure: LiteratureMonitor entity + watchlist scheduler + EnrichmentJob queue + worker loop (Deltas 8-9)
• ·PRN-021 — Bayesian-prior synthesis engine: EffectSize[] → CanonicalPrior random-effects meta-analysis, quality-weighted, provenance-preserving
• ·PRN-001..004 — audience-tier writeups (general / peer-review / engineering / product) for peopleanalyst.com/research/principia/
• ·PRN-005 — engagement-family construct survey (Tier 1 proof-of-method), gated on schema landing (now unblocked by ASN-1013 close)

Principia is structurally a curation layer that sits on top of meta-factory's extraction pipelines and below the People Analytics Platform's analytical surfaces. Three structural commitments define the shape. First, hub-and-spoke against `@measurement/core` — the canonical vocabulary (`Construct`, `Instrument`, `Item`, `Measure`, `Model`, `EffectSize`, `Publication`, `StudyQualityGrade`, `CanonicalPrior`) lives in the shared package, and Principia is the canonical registry of rows, not the canonical vocabulary. Second, tight coupling with meta-factory — asset IDs, variable codes, and classification dimensions are consumed verbatim from meta-factory's metadata coding; Principia does not invent parallel schemes and does not run parallel extraction. Third, two-shaped output — the same underlying rows render as a book manuscript and as a queryable database, exposed over a versioned REST + MCP contract so consumers vendor typed reads instead of re-implementing. The repo itself is a TypeScript monorepo with `packages/core`, `packages/registry`, and `apps/web` (Next.js 16); the markdown research surface (`docs/research/`) and the engineering scaffold ship in the same git history because they describe two surfaces of one project, separated only by audience.

Live at peopleprincipia.com. Public reader UI built out 2026-05-13 → 2026-05-21 (PRN-038a–d: scaffold + cross-entity linking + distribution plots + SEO/JSON-LD/sitemap/Analytics). First live CanonicalPrior shipped (engagement → task performance, k=3, from Corbeanu 2023 + Neuber 2021); engagement-survey v1.1 promoted 8 FILL markers to live priors. CanonicalTheoreticalModel layer + 8 seed theories (PRN-037), instrument-equivalence networks + 5 seeds (PRN-042), validity-decay / MethodologyCritique / cultural sub-priors (PRN-043/044/045), valid-&-reliable measures surface (PRN-040), VoI REST + intervention library (PRN-046/047), curator admin console at `/admin/*` (PRN-029), and the book-build orchestrator with XeLaTeX/HTML5/APA-7 templates (PRN-041/041a) all live. Hub-and-spoke against `@people-analyst/measurement-core` (renamed from `@measurement/core` 2026-05-19) is the structural commitment that keeps the vocabulary canonical across the portfolio; `@people-analyst/measurement-core` is now consumed by the People Analytics Toolbox over the canonical-priors REST contract.