Gallery

Explore our visual archive of maps and figures combining satellite imagery, machine learning, and causal inference to study development worldwide. Each item links to the underlying paper, data, and code—making it easy to understand, reproduce, and build on the work. Explore this gallery to see how AI, remote sensing and causal inference come together to illuminate the causes and consequences of human development!

A One Map, Many Trials Paradigm A One Map, Many Trials Paradigm (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 1 Chinese vs World Bank Development Projects visual abstract panel 1 (dark mode)
Multiscale satellite representation Multiscale satellite representation (dark mode)
Research visualization 1 Research visualization 1 (dark mode)
Table comparing properties of tabular, text, network, and earth observation data streams, including origin, high dimensionality, interpretability, and availability Table comparing properties of tabular, text, network, and earth observation data streams, including origin, high dimensionality, interpretability, and availability (dark mode)
Composite illustration summarizing eight aspects of integrating earth observation data into causal inference, including research questions, data measurement, spatial and temporal resolution, information leakage, computer vision models, and causal inference approaches Composite illustration summarizing eight aspects of integrating earth observation data into causal inference, including research questions, data measurement, spatial and temporal resolution, information leakage, computer vision models, and causal inference approaches (dark mode)
Research visualization 1 Research visualization 1 (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 2 Chinese vs World Bank Development Projects visual abstract panel 2 (dark mode)
Oracle vs single-/multi-scale diagram Oracle vs single-/multi-scale diagram (dark mode)
Research visualization 2 Research visualization 2 (dark mode)
Pair of directed acyclic graphs demonstrating image-based confounding at the scene level and confounding due to objects not in the image Pair of directed acyclic graphs demonstrating image-based confounding at the scene level and confounding due to objects not in the image (dark mode)
Earth-observation workflow diagram
Research visualization 2 Research visualization 2 (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 3 Chinese vs World Bank Development Projects visual abstract panel 3 (dark mode)
Uganda heatmap Uganda heatmap (dark mode)
Research visualization 3 Research visualization 3 (dark mode)
Maps showing Nigeria highlighted in Africa and Nigeria with treatment and control sites indicated by black and gray points Maps showing Nigeria highlighted in Africa and Nigeria with treatment and control sites indicated by black and gray points (dark mode)
Research visualization 3 Research visualization 3 (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 4 Chinese vs World Bank Development Projects visual abstract panel 4 (dark mode)
RATE ratio differences table RATE ratio differences table (dark mode)
Research visualization 4 Research visualization 4 (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 5 Chinese vs World Bank Development Projects visual abstract panel 5 (dark mode)
Chinese vs World Bank Development Projects visual abstract panel 6 Chinese vs World Bank Development Projects visual abstract panel 6 (dark mode)