Extraction Accuracy Stratum Divergence (EASD)

GMI-EGD-012Evidence-Governed DataIntegration-RequiredPattern O-2: Model Drift / Currency

Definition

The spread in extraction and normalization accuracy across corpus strata, measured against the accuracy on the material the models were accepted on.

The Measurement Gap

There is no metric that tells you whether extraction and normalization accuracy on the current corpus stratum has diverged from accuracy on the material the models were tuned and accepted against, which means degradation enters the method as evidence rather than as error.

Formula

EASD = max_k ( acc_baseline - acc_k ) , reported with the full per-stratum table

Proposed formula, pending validation pilots.

At a Glance

Unit
Difference (0 to 1)
Trend Direction
Lower
Category
Grading Integrity and Determinism

Decision Supported

Whether extraction output for a given stratum is still fit to be admitted as evidence.

← End-to-End Grade Reproducibility (EEGR)Grade Movement Attribution Rate (GMAR) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment