Trust criteria
Five versioned indices — no sixth composite “truth” score. Version: mvp-2026-08
Freshness
Measures: Dates, versions and signals of dated content.
Auto / human: Auto when signals are clear; lower level if undated.
What it is not: Not a guarantee of absolute pedagogical currency.
Licence clarity
Measures: Consistency between declared licence and signals in the resource.
Auto / human: Auto if consistent; always human review on conflict.
What it is not: Not legal advice.
Metadata completeness
Measures: Coverage of required / optional LRMI fields.
Auto / human: Auto. Empty optional fields lower the level without blocking alone.
What it is not: Not a rating of pedagogical content quality.
Accessibility
Measures: Technical / readability signals (structure, alternatives, limits by type).
Auto / human: Partial auto; gaps → lower level or N/A.
What it is not: Does not claim full RGAA/WCAG certification.
Sources and references
Measures: Identifiable author/organisation and presence/consistency of references.
Auto / human: Auto only if trivially clear; otherwise AI assistance + human.
What it is not: Not a measure of “truth” or absolute credibility.
A–E / N/A scale
| Level | Meaning (MVP) |
|---|---|
| A | Strong / favourable signals |
| B | Rather favourable |
| C | Mixed / incomplete |
| D | Weak / concerning |
| E | Very weak / critical |
| N/A | Not applicable or insufficiently assessable |
Scores can evolve and be recalculated. See also the Manifesto.