v7.33 public methodology

Lighthouse Research Object methodology

How one governed issue becomes an inspectable research object without editorial conclusions, hidden inference, or synthetic evidence.

1. Selection

All 1,139 v7.26 issue-state records are scored from governed counts only: published relationships, bounded evidence objects, source breadth, explicit geographies, existing Studio fixtures, Decision Lens fixtures, related research, and exact locators. Scores sort descending, then by canonical issue ID. The procedure selected issue:latinidad-worker-power. It does not rank importance, urgency, harm, truth, or viewpoint.

Selection report and weights

2. Governed inputs

The object references existing issue, evidence, source, stakeholder, relationship, Observatory, Studio, Decision Lens, and Workspace registries. It does not duplicate or rewrite the underlying governed corpus.

3. Source-supported values

Only bounded propositions attached to governed evidence objects are labeled source-supported. Evidence state, exact locator, source type, verification scope, and limitations travel with the proposition. A verified evidence proposition never becomes issue-level verification.

4. Calculated values

Counts, set intersections, and scenario deltas are deterministic calculations over stable canonical IDs. Each displayed metric carries a definition, method, denominator, scope, and limitation. Calculations are visually distinct from source-supported claims.

5. Stakeholder relationships

Only published governed v7.28 edges appear. Relationship type, direction, mechanism, evidence, geography, limitation, uncertainty, and reciprocal route remain accessible. An edge does not imply endorsement, causality, responsibility, representation, consensus, or influence.

6. Intersectionality bounds

The Studio accepts K=2–7 canonical issues. The Lighthouse uses an existing deterministic featured K=2 fixture. Shared published stakeholders are a corpus intersection, not an exhaustive account of identity, experience, representation, or causal effects.

7. Decision Lens interpretation

The Lens compares exact record sets, evidence coverage, category coverage, and gaps. It produces no aggregate score, preferred scenario, recommendation, or outcome forecast.

8. Missing and contradiction states

Missing means unknown, unavailable, incomplete, or unreviewed according to the labeled state. It never becomes “none” or “false.” Contradictions remain unknown where no governed evaluation exists.

9. Workspace privacy

Saving, collections, and private notes use the v7.32 browser-local IndexedDB model. No account, server synchronization, personal profile, telemetry expansion, or public personal state is introduced. Public exports contain the public object and provenance; private notes are excluded.

10. Reproduction

Run python3 tools/build_v733_lighthouse.py from the admitted v7.32 source. Generation uses stable input files, fixed release identity, canonical identifier sorting, deterministic JSON serialization, and no network dependency. The object content hash excludes no governed field and can be recalculated from the serialized representation with its hash field removed.