Decision preparation, bounded by evidence
The Decision Lens structures exact record sets from the published governed corpus. It does not forecast outcomes or recommend a decision.
Scenario definition
A scenario means “analyze this defined subset of the published governed corpus.” It is not a future state, intervention, forecast, or predicted outcome. Each configuration admits one through seven canonical issues, exact registry-backed filters, the v7.28 governed source release, and declared method versions.
The scenario ID is the first 20 hexadecimal characters of SHA-256 over canonical analytical fields. Arrays use canonical ordering. The optional local label is excluded from the ID, analysis, content hash, share URL, and analytics.
Question and comparison semantics
Seven bounded question templates constrain the permissible output. Two or three scenarios may be compared. The engine computes exact additions, removals, and shared sets for issues, relationships, stakeholders, bridge stakeholders, formal-authority relationships, evidence, categories, tensions, and gaps.
No aggregate scenario score, winner, recommendation, expected value, or policy rank is produced. “Higher” means only a larger labeled corpus count.
Coverage, authority, evidence, and tensions
Missing published coverage
The engine subtracts stakeholder categories represented by included governed stakeholders from the canonical category registry. The result describes corpus coverage; it does not establish applicability, exclusion, injustice, or a participation requirement.
Formal authority
Only relationships admitted through FORMAL_MANDATE_OR_JURISDICTION enter this subset. Authority does not become endorsement, support, opposition, preference, or exclusive jurisdiction.
Evidence asymmetry
Distinct evidence records and edge–evidence associations remain separate units. Missing dates and unknown geography remain unknown rather than zero. Counts do not measure truth or real-world research quality.
Tension gate
A tension appears only when an exact source-qualified canonical tension ID is included. Unequal counts, stakeholder sets, lanes, categories, or geographies do not independently establish tension.
Decision considerations and research questions
Fixed templates respond only to declared triggers such as uncovered selected issues, absent formal-authority coverage, missing dates, unknown geography, zero bridge coverage, or missing published categories. Every record carries its trigger, supporting IDs, method, research significance, and prohibited interpretation.
URL, privacy, and security contract
- Canonical issue and filter IDs are validated against immutable registries.
- Duplicate IDs are eliminated; one through seven issues and two through three comparison scenarios are enforced.
- User-visible labels are bounded to 64 characters, sanitized, local-only, and never executed.
- All rendered values are HTML-escaped; downloads use safe filenames and CSV formula neutralization.
- Invalid URL state fails safely and does not silently truncate a valid selection.
- No account, remote scenario storage, private-note transmission, custom analytics event, hosted AI, or third-party runtime request is added.
Accessibility and performance
The six states are keyboard-operable and deep-linkable. Selectors use programmatic labels, live selection announcements, clear removal controls, visible focus, semantic tables, row and column headings, focus-returning dialogs, linear mobile layouts, reduced-motion handling, 200% zoom support, and print reading order.
The browser loads the compact issue selector and bounded registries first, then lazily loads the existing v7.30 record index only after a valid scenario exists. No scenario combinations are precomputed.
Reproduction contract
Rebuild with python3 tools/build_v731_decision_lens.py, qualify the engine and source, package the flat static root, and fresh-extract the candidate. JSON, CSV, Markdown, and print output reconcile to the same analysis fields and canonical IDs.
Application release: v7.31 candidate. Governed source corpus: v7.28 production. Production deployment is not authorized.