It’s not always simple to spot disadvantaged people, but in the event of a pandemic, it might mean the difference between life and death. This was especially important in Togo, where 50% of the population lives on less than $1.90 per day. Togo used machine learning to analyze satellite imagery and cell phone data without a social register. The government was able to identify citizens in need of financial assistance, and cash was provided via cell phones.
Togo’s Minister of Digital Economy and Digital Transformation, Cina Lawson, was the lady in charge. She discusses the country’s unique approach to social support as well as the country’s intentions to create a national digital identity system. According to Lawson, the aid program was first implemented in metropolitan areas and around Togo’s capital, Lomé, in April of last year. In the native Ewe language, it’s called ‘NOVISSI,’ which means ‘solidarity.’
Citizens with cell phones were urged to text a special number and complete a survey. According to Lawson, the system would evaluate their eligibility based on their occupation and location, such as if they resided in a curfew-enforced region. Their identities were checked against a voter database that encompasses almost 93% of the population.
Later, the team turned its attention to rural regions, but it required to be able to pinpoint the poorest cantons and most vulnerable populations. “It was vital to make sure that the minimal cash available went to individuals who truly needed it,” Lawson adds. She says that while occupation-centered targeting worked well in urban areas, it did not work well in rural areas where most people are farmers or informal workers.
To find susceptible people, the researchers used mobile phone data and satellite photos. Experts from the University of California, Berkeley, the University of Mannheim, Northwestern University, and the non-profit Innovations for the Poverty Action contributed to the project. She explains that the work was done in two stages.
The researchers began by analyzing satellite pictures with AI. To determine the region’s wealth, they looked at things like roof material, road condition, agricultural quantity and quality, and settlement density. It calculated poverty levels for every 2km2 tile in Togo and prioritized aid to the 200 poorest cantons.
The team then used mobile phone data to decide which persons in these areas should receive assistance. Wealthier people, according to Lawson, make more calls and utilize more mobile network data. The team was able to anticipate the consumption of the mobile phone users in the originally identified locations using this data. NOVISSI aid was available to those who consumed less than $1.25 per day, she adds.
Togo partnered with the GiveDirectly, a non-profit that provides funds to residents of the country’s 200 poorest cantons. According to Lawson, it provided $8 million to over 139,000 individuals between November of last year and July of this year.