RFR-2026-015: Automated confidence-decay and frontier refresh

Research opportunity

Make conclusions time-aware and regenerate frontier candidates when evidence, dependencies, or status change.

Background

The repository has dashboards and registries but no consistent machine-readable confidence history or decay trigger.

Origin documents and trace

• research/operating-system/confidence-update-guide.md, section: confidence update process
• research/operating-system/evidence-dashboard/evidence-dashboard.json, section: current state
• research/evaluations/FE-BOUNDARY-2026-07-28/ros-fe-profile-v1-audit.md, section: PA-008, PA-016

Specific assumption challenged: Static confidence labels remain current as evidence and technology change.

Supporting evidence: Generated indexes can be stale and omit confidence/lineage query fields; current narratives already conflict.

Unknowns

• Appropriate decay functions by claim type
• Trigger sensitivity
• How to avoid automating epistemic authority

Dependencies

• RFR-2026-008
• RFR-2026-009

Suggested REP and methodology

Create REP-2026-015. Define claim-type-specific review intervals and event triggers; retain human approval; build check-only stale-state detection and an append-only confidence history.

Expected outputs

• Confidence-history schema
• Decay policy experiment
• Frontier refresh checker

Success criteria

Stale or contradicted conclusions are flagged reproducibly without automatically changing accepted claims.

Execution recommendation

• Recommended agent: research-operations tooling agent
• Estimated effort: medium
• Expected knowledge gained: Lets the repository detect when its own frontier analysis is obsolete.
• Frontier Score: 102
• Score inputs: knowledge gain 3/5; impact 3/5; cross-project reuse 4/5; scientific importance 3/5; dependency cost 3/5; implementation difficulty 3/5.
• Score calculation: 3 × 3 × 4 × 3 − 3 − 3 = 102.
• Score status: structured expert judgment, not measured utility.