Improving the lives of millions of Latin Americans through better welfare targeting algorithms

More than 50 million people in Latin America are impacted by the decision of very simple linear algorithms which determine how much welfare they receive from social programs. Simple changes to the algorithm lead to hundreds of thousands of people being added or removed to major welfare programs which often determine over 50% of their income.

Despite this incredible impact, very few machine learning engineers and researchers work to improve these algorithms. Very few funders target this cause area. In the face of that neglect, dozens of countries continue to use antiquated algorithms and targeting systems that misallocate billions of dollars of capital.

There is a tremendous opportunity here to improve the accuracy and fairness of social welfare algorithms. My colleagues and I have implemented these kinds of algorithms in Colombia, Costa Rica, Panama and Honduras. Effective altruists looking to understand this opportunity more deeply can get in touch with me at anc@prosperia.ai and can read about our work at https://​​www.prosperia.ai/​​. Our academic paper goes into technical detail at https://​​dl.acm.org/​​doi/​​pdf/​​10.1145/​​3351095.3375784.

This is the highest point of leverage for social good that I believe exists. Please join us.

By: Alejandro Noriega-Campero
Thank you to Jeremy Nixon and Ivan Vendrov for feedback and editing.

No comments.