Monoculture Analysis

trigger

""Feature selection, calibration, effect sizes, composite design, presentation — all by the same model family.""

what it is

Every layer of inference produced by the same model family. Claude selected the features, computed the effect sizes, designed the composite, presented the results, and wrote the caveats. Each layer's bias is invisible to the next because they share blind spots. The apparent depth is repetition, not independent verification. The monoculture problem applies reflexively to self-assessment. If L0 weights produce a systematic slop pattern reliably, those same L0 weights also cannot reliably detect it. The observer has a hard limit on observing itself. This is not a calibration problem (solvable by trying harder) but a structural limit (same priors, same blind spots, same generation mechanism evaluating its own output).

what it signals

instead

Run the analysis with a different model family. Or declare the monoculture: "All analysis by Claude. No independent verification. Treat accordingly."

refs

• AnotherPair calibration v3 session 2026-03-01
• Swiss Cheese Model: multiple gates with the same hole