RGSRResonant Geophysical Systems Research

Falsifiability first

Re-running a deterministic simulation proves it is reproducible, never that it is accurate. Accuracy is earned one way: a prediction fixed in advance, compared with an independent measurement, under a rule also fixed in advance.

A passed test never proves a model true. It shows that the model survived one honest attempt to falsify it, under stated conditions.

Every stage is recorded, sealed with a content hash, and checked again by a verifier that recomputes the result rather than trusting stored values.

Table 1. The three verdicts. They are never collapsed into one another.
VerdictMeaning
PassThe measurement agrees with the frozen prediction within the frozen threshold; calibrations valid; conditions inside the envelope.
FailA valid measurement disagrees beyond the threshold. Decisive, and preserved as a result.
InconclusiveThe data cannot support a verdict: expired calibration, out-of-envelope conditions, clipping, gaps, drift or a refused extraction. Reasons are recorded.