AI for Livelihoods in Forested Regions
Paula J. Sarmiento — Duke University
The problem
- Socioeconomic data in the Amazon — and in other forested frontiers — is scarce.
- Yet that is exactly where conservation policy needs to be implemented.
What I have
- A rich household survey in Guaviare, Colombia (~2001 households) — non-representative data.
- For each household: sociodemographics, land tenure, forest dependence, social characteristics, GPS location, and spatial covariates.
- Other disconnected surveys might exist, but there is no centralized source of information.
The idea
- Train an ML model on the my detailed survey…
- …to predict socioeconomic outcomes in neighboring Amazon areas.
Where the survey is
