Fighting Poverty with AI

An algorithm that uses artificial intelligence to identify people most in need of financial assistance was recently launched in Togo, where it has been estimated that 55 percent of the population lives in poverty. In rural areas, that number is more than 80 percent.

As COVID-19 shutdowns affect people throughout the world, charities like GiveDirectly are more focused than ever on getting cash directly into people’s hands. The new algorithm helps identify those individuals more quickly and accurately, and it’s totally contactless, an important aspect as the pandemic continues to spread.

Instead of relying solely on government data or poverty indexes, the GiveDirectly algorithm analyzes high-resolution satellite imagery to look at features that distinguish poorer areas from wealthier ones, like building size, building density, roofing materials and paved vs. unpaved roads. A second stage of analysis looks at mobile phone data for other clues such as data usage and length of calls. Selected individuals are prompted by text message to enroll and then receive payment via a cash transfer system that the Togo government set up several months ago. Cash can be withdrawn at local shops.

For a period of five months, recipients receive $15 per month for women and $13 per month for men, which is the amount that has been determined as necessary to minimally meet their survival needs. So far, the program has reached 30,000 people, with a goal of reaching 58,000, or the poorest 10 percent. To donate, visit The Audacious Project.