#PovertyMapping

ML wealth maps from space 🛰️ are great, but suffer from "shrinkage" bias, which waters down policy impact results (causal inference). We developed correction methods that fix this bias *without* new data.

arxiv.org/abs/2508.01341

#CausalInference #DataForGood #AI #PovertyMapping #EarthObservation

Debiasing Machine Learning Predictions for Causal Inference Without Additional Ground Truth Data: "One Map, Many Trials" in Satellite-Driven Poverty Analysis Machine learning models trained on Earth observation data, such as satellite imagery, have demonstrated significant promise in predicting household-level wealth indices, enabling the creation of high-...

Debiasing ML predictions for causal inference—no new labels needed. We propose Tweedie’s correction to fix shrinkage enabling “one map, many trials.”
arxiv.org/abs/2508.01341
#CausalInference #MachineLearning #EarthObservation #PovertyMapping