Source
Devezer et al. (2019), LIT-988, read in NOTE-792: the main finding (claims C1, strong, and C2, moderate) that reproducibility and time at the truth come apart because research strategies decide which comparisons are run.
What is asserted, and by whom
The thesis is that in real scientific communities, as in the model, reproducibility of results can be high while the standing view is false, so a high replication rate does not certify a field’s consensus. Devezer et al. draw this lesson for science. The narrowed entry keeps only what the model proves.
What the record holds toward it
- THEORY-295 (Active): in the memoryless 14-model framework, replication cannot change where the consensus goes, and a Bo-dominant population holds untrue models with near-perfect reproducibility.
- No study of a real field is held.
What would confirm or reject it
A model with memory and learning that keeps the decoupling would strengthen it; a meta-scientific study of a field that replicates well while independent evidence shows its standing model wrong, traced to which alternatives were tested, would confirm it. A field where high replication tracked truth whenever the model space was searched would count against it. QUESTION-151 asks the model half as a question; this entry is the answer Devezer et al. propose.