Skip to content

What this is

The core terms. Mental model: graded source records are resolved into a canonical registry of constructs, instruments, and effect sizes, then synthesized into Bayesian priors that downstream apps and analysts consume.

The measurement vocabulary (@measurement/core)

The canonical types every sibling shares: Construct, Instrument, Item, Measure, Model, EffectSize, Publication, StudyQualityGrade. Apps consume these via type imports — they do not re-implement the schema. Principia is the canonical registry (the actual graded rows); the core is the vocabulary.

Construct family

The unit of survey work — one construct family per document (e.g. engagement). A family survey covers: definition + history, the instruments that measure it (item-level schemas), reliability/validity evidence per instrument, a graded meta-analytic effect-size table, and cross-cultural/operational notes.

Source grade

Every record is graded for study quality (StudyQualityGrade) and citation-verified. Grading is what separates Principia from a scraped bibliography — claims carry an evidence quality, not just a citation.

Prior (the substrate)

Graded effect sizes synthesize into Bayesian prior distributions — the priors researchers and practitioners plug into their own analyses. The product arc layers this as:

  • Credible — artifact-corrected, bias-checked synthesis.
  • Living — a primary-evidence network + item benchmarks that confirm/refine priors with real deployments (the file-drawer escape).
  • Useful — Value-of-Information for decisions.

Source-side vs resolved-side

CanonicAI's pipelines emit ConstructCard, InstrumentCard, Citation as authoritative source-side records. Principia's resolved-side spine (CanonicalVariable, CanonicalSurveyItem) sits on top of — not in place of — those. Asset IDs reuse CanonicAI's {domain}:{type}:{source_id} scheme verbatim.

Scoring (θ̂)

Given item responses + a model, the Scoring Engine returns θ̂ (grid-MLE) with Fisher SE/CI, info_gain_next, and threshold_met — "how confident are we in competency X, and is the threshold met?"

See also

APIs & Data Contracts · Architecture · Trust & Grading