RFR-2026-003: Matched incremental-utility experiment

Research opportunity

Test whether the FE profile improves outcomes beyond strong adjacent methods and equal-dose generic structure.

Background

The newest boundary REP finds substantial collective subsumption and identifies incremental utility as the decisive unknown.

Origin documents and trace

• research/evaluations/FE-BOUNDARY-2026-07-28/RP-FE-BOUNDARY-2026-07-28.md, section: F3-F7; Decision record
• research/evaluations/FE-BOUNDARY-2026-07-28/next-experiment.md, section: entire proposed design
• research/theory/falsification-criteria.md, section: Additional Weakening Signals

Specific assumption challenged: Observed redesign gains are FE-specific rather than scaffolding, documentation dose, or review effort.

Supporting evidence: The boundary REP reports no matched causal comparison and very-low confidence in incremental utility.

Unknowns

• FE-specific causal effect
• Which profile components carry value
• Net benefit after learning and operating cost

Dependencies

• RFR-2026-001
• RFR-2026-016
• RFR-2026-002

Suggested REP and methodology

Create REP-2026-003. After the controlling mechanism-boundary experiment is completed or formally superseded, preregister a multi-arm study comparing FE, strongest domain-specific comparator, equal-dose generic scaffold, and minimal control; hold task, provider, time, and information constant; use blinded outcome raters.

Expected outputs

• Preregistration
• Matched task set
• Effect and cost analysis

Success criteria

Primary outcomes, harms, time cost, and uncertainty are reported with no post-hoc change to the primary contrast.

Execution recommendation

• Recommended agent: causal-experiment agent
• Estimated effort: high
• Expected knowledge gained: Determines whether the repository should expand, narrow, or terminate the integrated-profile claim.
• Frontier Score: 490
• Score inputs: knowledge gain 5/5; impact 5/5; cross-project reuse 4/5; scientific importance 5/5; dependency cost 5/5; implementation difficulty 5/5.
• Score calculation: 5 × 5 × 4 × 5 − 5 − 5 = 490.
• Score status: structured expert judgment, not measured utility.