Evaluation of Coalition-Proof Scoring Rules for Artificial Intelligence Agriculture Extension Rankings Under District Patronage Reassignment: A Governance Stability Bound for Smallholder Access Equity

Artificial intelligence systems that rank agricultural extension agents increasingly allocate smallholder access to advisory services, yet the scoring rules underlying those rankings remain vulnerable to strategic manipulation by district-level coalitions that can reassign patronage across administrative boundaries. The framework treats district patronage reassignment as a cooperative game in which coalitions of district officials may coordinate to alter agent rankings, and it derives a stability bound that specifies the conditions under which no coalition can profitably deviate from the truthful ranking. The analysis characterises the bound as a function of the number of districts, the granularity of the scoring rule, and the reassignment cost structure. Three scoring-rule families are compared: additive threshold rules, ordinal bucket rules, and marginal-contribution rules. The framework proposes that marginal-contribution rules with district-level caps satisfy the stability bound under weaker conditions than the alternatives, but at the cost of reduced sensitivity to within-district variation.