Counterfactual-Fairness Evaluation for Computer Agriculture Credit-Scoring Models Under Missing Land-Title Records: An Artificial Intelligence Causal-Effect Audit

Credit-scoring models deployed in computer agriculture increasingly allocate loan decisions on the basis of features that are themselves shaped by historical land-tenure inequality, and the fairness of those models is routinely assessed with counterfactual criteria that assume complete covariate records. The problem is framed as a causal-effect audit: the quantity of interest is the counterfactual disparity in approval that would obtain if a protected attribute were intervened upon while the missingness mechanism is held fixed. A formal framework is developed in which missing land-title status is treated as a coarsening of the protected attribute rather than as a nuisance to be imputed away, and three propositions are derived that characterise when the observed counterfactual disparity is an unbiased, attenuated or sign-reversed estimator of the target disparity. The framework is then applied to the operating characteristics of a before-after-control-impact design used to evaluate the audit procedure, with a declared variance structure of between-cluster SD 6.0, residual SD 4.5 and intraclass correlation 0.64, and power estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The analysis indicates that 54 clusters are required to detect a 5-percentage-point effect at 80 per cent power. The contribution is a specification of the conditions under which counterfactual-fairness audits of agricultural credit models remain interpretable when tenure records are incomplete, and a demonstration that the audit design, not only the model, must be evaluated for its own operating characteristics.