Active-Learning Evaluation for Computer Agriculture Pest-Surveillance Classifiers Under Budget-Constrained Scouting: An Artificial Intelligence Query-Selection Bound for Detection Efficiency

Active learning promises to reduce the annotation cost of pest-surveillance classifiers by selecting only the most informative field queries, yet the evaluation of query-selection strategies under realistic scouting budgets remains informal, and reported gains are rarely tied to a formal bound on detection efficiency. No observational data are used or claimed: the analysis characterises the operating characteristics of the design under a declared variance structure with between-cluster SD = 6.0, residual SD = 4.5 and intraclass correlation = 0.64, 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 bound decomposes detection efficiency into an informativeness term and a coverage term, and it predicts that the marginal value of an additional query falls as the intraclass correlation rises. The design study indicates that 54 clusters are required to detect a 5-percentage-point effect at 80% power, and that power curves, minimum detectable effect curves and residual diagnostics behave as the bound anticipates. The contribution is a formal, falsifiable criterion for evaluating active-learning query selection in agricultural pest surveillance, together with a design-level statement of the cluster counts at which the criterion becomes informative.