Information is not Implementation: Early evidence from using machine learning for climate policies when state capacity is low

Climate change adaptation policies are increasingly using machine learning to make use of more information and provide more accurate predictions of extreme weather events. Equally, Payment for Ecosystem Services to monitor and affect land use are increasingly relying on remote sensing data. This talk will draw upon a set of recent (ongoing) field experiments to highlight the importance of designing ML-based climate policies that account for last mile implementation constraints and low citizen trust in the state.