Certified-Fairness Evaluation for Computer Agriculture Credit-Scoring Models Under Cooperative-Lending Disparities: An Artificial Intelligence Equalized-Odds Audit for Smallholder Credit Access

Credit-scoring models deployed by agricultural cooperatives can allocate smallholder credit in ways that satisfy overall accuracy while systematically misclassifying borrowers from disadvantaged groups. The framework formalises equalized odds as a pair of group-conditional error-rate constraints, derives the feasible audit region under a declared variance structure, and specifies a certification rule that a cooperative or regulator can apply to a fitted model without access to the training pipeline. A replication-based evaluation of the before-after-control-impact estimator was conducted to characterise the operating characteristics of the audit design under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.640). Power was estimated across 2000 design cells spanning 8, 16, 24 and 32 clusters and effects of 3, 5 and 8 percentage points. The analysis indicates that 54 clusters are required to detect a 5-percentage-point disparity at 80% power, and that the intraclass correlation of 0.64 dominates the power profile at small cluster counts. The contribution is a certification procedure that converts an abstract fairness criterion into an auditable, power-aware test that cooperatives can run on deployed models.