RFR-2026-004: Recognition-control replication

Research opportunity

Separate lexical recognition, structural recognition, and unsupported prior-knowledge import.

Background

Recognition effects are mixed; H013 has moderate evidence while H016 remains insufficient.

Origin documents and trace

• research/operating-system/evidence-dashboard/evidence-dashboard.json, section: active_hypotheses H013 and H016
• research/operating-system/research-queue/research-queue.md, section: candidate E
• research/operating-system/evidence-dashboard/evidence-dashboard.json, section: material_threats

Specific assumption challenged: Removing watched terms adequately controls recognition.

Supporting evidence: Recognition leakage is material and current findings do not isolate its mechanism.

Unknowns

• Lexical versus structural recognition effect
• Unsupported import rate
• Interaction with provider and expertise

Dependencies

• RFR-2026-001

Suggested REP and methodology

Create REP-2026-004. Factorially vary names, canonical vocabulary, structural cues, and misleading labels; audit unsupported imports against packet evidence.

Expected outputs

• Factorial packet set
• Recognition/import annotations
• Replication report

Success criteria

The design estimates lexical and structural effects separately and includes negative/misleading controls.

Execution recommendation

• Recommended agent: cognitive-bias validation agent
• Estimated effort: medium
• Expected knowledge gained: Clarifies whether apparent recovery is extraction or recognition.
• Frontier Score: 394
• Score inputs: knowledge gain 4/5; impact 4/5; cross-project reuse 5/5; scientific importance 5/5; dependency cost 3/5; implementation difficulty 3/5.
• Score calculation: 4 × 4 × 5 × 5 − 3 − 3 = 394.
• Score status: structured expert judgment, not measured utility.