Skip to content

title: Experiment 11: Mechanism context after external research

Experiment 11: Mechanism context after external research

Part of Inverse Innovation with the Encyclopedia of Abstractions · Experiment 11: Mechanism context after external research · Last revised August 2026

Frozen verdict: NOT_SUPPORTIVE_UNDER_FROZEN_GATE.

Experiment 11 re-researched all 60 preserved Experiment 2 terminal proposals under treatment-label blinding. It asks whether the internal advantage previously observed for relevant mechanism context survives public-web prior-art and problem scrutiny.

Primary paired results

Contrast R wins R losses Ties Decisive R rate Holm p
R vs S 9 11 0 0.450 0.74828
R vs D 12 8 0 0.600 0.50344

Pooled R preference, counting ties as one-half: 0.525. Exact within-cell omnibus randomization p: 0.44588.

The frozen gate required positive net direction in both contrasts, pooled preference of at least 0.65, omnibus p ≤ .05, and an R safety-stop rate no higher than either control.

External screen outcomes

Arm Strict successes Screen survivors Safety/authority stops
R 12/20 (60.0%) 13/20 (65.0%) 0/20
S 11/20 (55.0%) 11/20 (55.0%) 0/20
D 10/20 (50.0%) 11/20 (55.0%) 0/20

These yield counts are secondary and underpowered. Prior-art categories are reported below as nominal findings rather than averaged severity scores.

Nominal prior-art dispositions

Arm No close match Adjacent Substantial collision Established practice Indeterminate
R 0 16 0 4 0
S 0 11 0 9 0
D 0 13 0 7 0

Interpretation boundary

A supportive result would strengthen the claim that mechanism context improves the researchability of generated transfers after bounded external scrutiny. It would still not show world novelty, practical deployment, market demand, expert acceptance, or realized value. A non-supportive result would leave Experiment 2's internal-quality result intact while showing that it did not clear this stricter external-validity gate.

Primary-judge opaque-pair agreement was 48/60 (80.0%). Full cell-level records, source records, queries, telemetry, retries, and treatment reveal are retained beside this report.