When information is not enough: AI-enabled advisory and the fragile pathways of low-carbon transitions under system constraints

Abstract

CONTEXT Low-carbon rice farming is widely recognized as a key strategy for agricultural decarbonization, yet adoption remains limited among smallholder farmers despite expanding AI-enabled advisory and increasing awareness of climate risks. Existing adoption-based models provide limited explanation for why improved information and awareness often fail to translate into low-carbon transition. OBJECTIVE This study explains how farmers translate information into transition intention under real-world constraints by conceptualizing low-carbon agriculture as a constrained system transition. It develops a process-based cognitive decision pathway linking risk perception, perceived feasibility, and decision confidence, while positioning AI-enabled advisory as cognitive infrastructure rather than a direct transition intention. METHODS Survey data were collected from 398 smallholder rice farmers in Vietnam's Mekong Delta. Partial least squares structural equation modelling (PLS-SEM) was employed to estimate a framework integrating sequential mediation and moderated mediation, with government support and market access modelled as system-level enabling conditions. RESULTS AND CONCLUSIONS AI-enabled advisory significantly enhances risk perception, perceived feasibility, and decision confidence but has no direct effect on transition intention. Perceived feasibility emerges as the critical implementation bottleneck, while decision confidence functions as the final cognitive mechanism translating favourable feasibility assessments into transition intention. In contrast, risk perception increases awareness but does not significantly improve feasibility, revealing a fundamental tension between recognizing climate risks and judging transition to be practically achievable. Government support and market access further strengthen this decision pathway by reinforcing the translation of cognitive readiness into transition intention. SIGNIFICANCE This study makes three principal contributions. First, it reconceptualizes low-carbon agriculture as a constrained system transition rather than a discrete adoption decision. Second, it advances a process-based cognitive decision pathway that explains how farmers progress from implementation feasibility to confidence to act under uncertainty. Third, it repositions AI-enabled advisory as cognitive infrastructure, demonstrating that its effectiveness depends on enabling institutional and market conditions rather than information provision alone. These findings provide a more integrated theoretical framework for understanding low-carbon agricultural transition and offer practical guidance for designing AI-enabled and system-oriented transition policies.