Answers are grounded in this paper and cite their sources.
Generative artificial intelligence (AI) reduces the time and effort required to prepare competitive research proposals, potentially reshaping the allocation of agricultural research funding. We study these implications in the context of the U.S. Department of Agriculture (USDA)'s Agriculture and Food Research Initiative (AFRI), the primary federal mechanism supporting U.S. agricultural research and innovation. Using a formal economic model of the AFRI grant market developed from a no‐AI benchmark, we analyse two AI scenarios: independent adoption , in which AI use is uncorrelated with researcher quality, and correlated adoption . AI‐driven reductions in preparation costs can increase submissions from marginal projects, weaken peer‐review signal informativeness and amplify institutional funding advantages even when underlying research quality is comparable. Allowing AI adoption to correlate with researcher quality shows that compensatory adoption—a negative adoption–quality correlation deepens signal degradation, whereas the correlation must exceed approximately 0.10 before fixed‐pool informativeness improves relative to the no‐AI benchmark. In the calibration, uniform high adoption increases equilibrium submissions by 18.0%. Under a heterogeneous scenario in which 70% of potential investigators receive high‐intensity AI assistance and 30% receive none, submissions increase by 6.2%, while mean funded quality falls by 3.5%. These allocative effects coexist with proposal preparation resource savings. We evaluate second‐best responses implementable within USDA's existing authority, including mandatory pre‐proposals, meritorious‐pool lotteries, AI‐use disclosure and targeted AI‐access programmes for historically Black land‐grant and other minority‐serving institutions. Generative AI can therefore raise applicant productivity while weakening the information and selection mechanisms underlying efficient and equitable research funding.
No takes yet. Share an insight, caveat, or question.
Ng’ombe et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: