Automating Inequality

How Poverty Became a Data Problem

In 2015, Virginia Eubanks learned how quickly an automated system can throw a family into crisis. After her partner, Jason, survived a violent assault and needed major facial reconstruction, their health insurance suddenly disappeared inside a computer system. Even though their coverage was valid, the insurer claimed there was no policy start date and refused to pay more than $62,000 in medical bills. While the company called it a glitch, the pattern suggested that the family may have been flagged by an automated fraud screen because the claims were large, recent, and included pain medication.

That experience showed how many major decisions are now handed to software. Systems sort people for healthcare, loans, policing, housing, and public benefits, often with little explanation. They are sold as neutral tools that increase efficiency, but they can create life-changing barriers that are hard to see and even harder to fight. People with money, time, and connections may be able to push back, but poor and working-class families often cannot.

This burden is not spread evenly. In Maine, state officials used benefit card data to track where low-income people withdrew cash, then highlighted a tiny number of transactions to build a public story about waste and fraud. The point was not to improve assistance, but to cast suspicion on people who needed help. Data became a political weapon that made poverty look like personal failure.

Across the country, public agencies have built similar systems into daily life. Indiana automated welfare eligibility. Los Angeles created a shared database to rank unhoused people for services. Allegheny County in Pennsylvania began using predictive models to estimate which children might face abuse or neglect. These systems collect intimate personal information, limit access to help, and turn human need into a technical sorting problem.

This is the digital poorhouse: not a building, but a network of databases, algorithms, and screening tools that monitor and manage poor people from a distance. Older systems used home visits, case files, and institutions. Newer ones use risk scores, document portals, and hidden decision rules. The tools look modern, but they often serve an old purpose: making aid harder to get, making suffering easier to ignore, and treating the poor as people to be managed rather than citizens with rights.