Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor

Los Angeles’s automated welfare system collects, stores, shares, catalogs, classifies, and ranks numerically information about the unhoused. Designed to triage those with immediate need of housing, CES should be “a standardized intake process to reduce waste, redundancy, and double-dipping across agencies” (85), but in practice it often comes up short and is difficult to navigate. Moreover, CES poses a threat to the people it claims to be trying to help because it so deeply invades individuals’ privacy. The data collected during this entry system is stored in the Homeless Management Information System (HMIS) for at least seven years and used indiscriminately for law enforcement fishing traps.

The digital poorhouse is a metaphor that paints automated algorithms and systems used in welfare as a modern version of the historical poorhouse. Eubanks argues that the digital poorhouse’s “goal is what it has always been: to profile, police, and punish the poor” (38). Just as the physical poorhouses of the 19th and early 20th centuries demonized, isolated, and contained the poor, so too do today’s digital and dehumanizing datasets put “administrative power in the hands of a small elite” while creating “towering inequalities” (200) via welfare and social assistance programs.

Scientific charity is an outdated methodology of assistance that attempted to quell the backlash to poor relief through “more rigorous, data-driven methods” (22). Rather than providing blanket welfare for citizens like the countries of Europe, the US created moral distinctions between the “deserving poor from the undeserving” (22), making public assistance contingent on monitored behavior and thus eventually criminalizing poverty itself.

Passed by President Bill Clinton after he swore to “end welfare as we know it” (37), this 1996 law decimated US welfare systems: After its passage, “almost 8.5 million people were removed from the welfare rolls between 1996 and 2006” (37).

Eubanks uses this term to describe how automated decision-making systems and artificial intelligence entrench and proliferate bias rather than being the impartial tools they are supposed to be. Because algorithms are created from existing datasets and programmed by people who necessarily have blind spots, these systems end up perpetuating pre-existing biases.

In 1996, alongside PRWORA, President Bill Clinton replaced the children’s welfare program Aid to Families with Dependent Children (AFDC) with TANF, which carried much stiffer compliance regulations and harsher penalties for failing to adhere to the law’s requirements, with the result that many fewer poor children started receiving assistance: “In 1973, four of five poor children were receiving benefits from AFDC. Today, TANF serves fewer than one in five of them” (37).