Appearance
What this is
Principia's value is that its claims are graded and sourced, not scraped. This states how trust is established and where the limits are.
Source grading
Every record carries a study-quality grade (StudyQualityGrade). Effect sizes are graded before they enter a synthesis, so a prior reflects evidence quality — not just the existence of a citation.
Citation verification
Records are citation-verified with source-graded BibTeX committed per construct family. A construct survey is not "done" until reliability/validity evidence is cited per instrument and the meta-analytic effect-size table is populated and graded.
Novelty verification
A novelty pass guards against re-registering an instrument the registry already covers under another name — keeping the canon clean as siblings file new discoveries.
Prior provenance (Credible → Living → Useful)
- Credible — priors are artifact-corrected and bias-checked, not raw pooled means.
- Living — priors are confirmed/refined against real client deployments and item benchmarks (the file-drawer escape), so they don't ossify.
- Useful — Value-of-Information framing ties a prior to the decision it informs.
Provenance chain
Source-side cards (from CanonicAI) → resolved-side canonical spine → graded synthesis → prior. Each layer is preserved, so a prior traces back to graded primary evidence — defensible, not asserted.
Honest scope
Coverage is uneven and growing (construct-family by family); a prior is only as strong as the graded evidence behind it, which PIPELINE_STATUS.md reports honestly. Principia grades and synthesizes evidence; it does not replace a researcher's judgment for a specific application.