Evaluation of Rollback-Equivariance Audits for Artificial Intelligence Agriculture Tariff-Schedule Models Under Retroactive Ministerial Amendment: A Governance Non-Retroactivity Guarantee for Smallholder Compliance Burden

Retroactive ministerial amendment of agricultural tariff schedules creates a governance problem for smallholders: an artificial intelligence model that computes compliance obligations from a schedule may be revised after the fact, so that a farmer's liability changes without any change in the farmer's conduct. The audit is formalised as a property of the model's mapping from schedule parameters and transaction dates to assessed liability, and its operating characteristics are characterised through a replication-based design evaluation of a before-after-control-impact estimator under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64). Power was estimated across 2000 design cells at cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The design requires 56 clusters to detect a 5-percentage-point effect at 80 per cent power. Rollback-equivariance auditing supplies a non-retroactivity guarantee that is procedurally enforceable ex ante, and that its statistical power is the binding constraint on its governance value for smallholder compliance burden.