Benchmarking Evaluation of Speech-Based Pest-Reporting Assistants for Computer Agriculture Extension Under Low-Literacy Dialect Variation: An Artificial Intelligence Recognition-Robustness Audit for Advisory Equity

Speech-based pest-reporting assistants promise to widen agricultural extension access for low-literacy farmers, but their advisory value depends on recognition robustness across dialect variation that current benchmarks rarely audit. The framework formalises dialect-conditioned word error rate as a function of acoustic and lexical distance from the training distribution, and shows that aggregate word error rate masks subgroup disparities that determine whether pest reports reach advisory thresholds. A replication-based evaluation of the before-after-control-impact estimator was conducted under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.640), with power 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 analysis indicates that 52 clusters are required to detect a 5-percentage-point effect at 80% power under this variance structure, a requirement that exceeds the cluster counts typically available in extension-pilot evaluations. The contribution is a recognition-robustness audit protocol that specifies the cluster counts, effect sizes and reporting quantities needed for advisory-equity claims to be defensible.