Revisiting Everett M. Rogers: A systematic review and random forest analysis of technology adoption in modern agriculture

Agricultural innovation is pivotal to meeting global food, climate, and livelihood challenges. This study systematically reviews publications from 2021 to 2025 on agricultural technology adoption. We combine Everett Rogers’s diffusion of innovation theory with a supervised machine learning technique: Random Forest (RF), an ensemble tree-based method within the broader AI toolkit, to identify key factors that influence adoption outcomes across 571 cases from 531 publications. Our stepwise approach integrates systematic bibliometric searches, rigorous textual coding and numerical conversion, and RF modeling to synthesize diverse empirical evidence into actionable, data-driven guidance. The RF model, which demonstrates good predictive performance, highlights extension access, climate risk awareness, and perceived relative advantage (along with perceived simplicity and training participation) as the most influential predictors of adoption decisions. Education, or more broadly, innovation literacy, emerges as essential in specific local contexts but less influential across all cases, while peer networks exert moderate, context-dependent effects. These findings suggest that extension messages and programs should emphasize clear, observable benefits, manageable complexity, and climate-related risk information that directly address farmers’ needs and concerns. Overall, this integrated methodological approach provides robust and nuanced insights, offering practical guidance for agricultural development policy, extension strategies, and future research.