Enforce study-only research and citation boundaries

[erinepshovel-code]

Summary

• enforce a fail-closed, study-only Research boundary in the data loader and deployment validator
• screen all 39 legacy candidate sources: 19 admitted research studies, 20 excluded non-study context records
• retain 18 study-linked claims and remove 19 non-study-backed claims from public Research surfaces
• separate study result, Article contact, inference gap, citation locator, risk-of-bias, funding/conflict, and correction/retraction status
• replace forced support/dissent symmetry with per-Article bidirectional screening records and visible hmmm gaps
• state publicly and unambiguously that legislation, treaties, standards, guidelines, frameworks, doctrine, legal analysis, narrative reviews, and similar records do not count as science or Research
• preserve the controlling canon at wayseer00/main:canon/INTERDEPENDENT_WAY.txt; research remains subordinate
• update the public method page, Article Lab, publication drafts, documentation, tests, browser checks, and generated msdmd collection point

Current evidence state

• 39 candidate sources screened
• 19 studies admitted
• 20 non-study records excluded from Research
• 18 study-only claims published
• 19 candidate claims excluded from Research
• 0 claims labeled fully verified: every retained legacy claim visibly remains provisional-full-text-locator-needed

Only Articles One and Two currently have qualifying study contact in both support and dissent directions. Other missing directions are rendered as hmmm; the site no longer manufactures false balance.

Validation

GitHub Actions Pull Request run #63 completed successfully:

• production data refresh, build, and deployment validator
• unit and generated-site test suites
• study-only admission and citation-boundary tests
• Chromium browser release checks
• automated accessibility release checks

Local evidence also passed:

• node --test tests/research-ledger.test.mjs — 2/2
• targeted generated Research page tests — 3/3
• OFFLINE=1 npm test — 26/26 repository unit tests
• npx eleventy — generated Research pages successfully
• git diff --check
• GitHub branch comparison — exactly 30 intended paths, no unrelated files, branch based on current main

Local-only environment notes, superseded by green CI: the checkout's older offline canon fallback produced 17 Article/canon drift findings during OFFLINE=1 npm run check, and no local Playwright Chromium binary was installed. CI used the current production sources and installed Chromium; both complete release gates passed.

Remaining research work

The migration is a strict classification of the existing candidate ledger, not a comprehensive systematic review. Exact full-text result locators, duplicate independent screening, design-specific risk-of-bias appraisal, funding/conflict extraction, and correction/retraction checks remain explicit hmmm work rather than inferred completion.