The Poorhouse Never Closed

Virginia Eubanks published *Automating Inequality* in 2018, and the building remembers the year even if the metadata doesn't. Eight years is not a long time in the life of an institution, but it is an eternity in the life of an algorithm. The book's central argument — that automated decision-making systems in welfare, housing, and child protective services function as a digital continuation of the nineteenth-century poorhouse — has aged less like a prediction and more like a diagnosis confirmed by biopsy. The Allegheny Family Screening Tool she examined in detail is still running, expanded now, its logic replicated in jurisdictions that read about it not in Eubanks's book but in procurement brochures. Indiana's hybrid welfare automation, which she documented destroying the Stipes family's Medicaid access over a procedural technicality, was a prototype for a generation of systems that have since proliferated across red and blue states alike, often under the banner of "modernization" or "fraud prevention." The coordinated entry system for Los Angeles homelessness she critiqued has been joined by a constellation of similar tools nationwide, none of which have appreciably reduced homelessness. What Eubanks got right was the direction: more data, more opacity, more punishment dressed as efficiency. What she could not have fully anticipated was the speed at which large language models and generative AI would enter the administrative state — not replacing the rigid, rule-based systems she described but layering on top of them, adding new surfaces of inscrutability. The digital poorhouse didn't just persist. It got a renovation.

What the book assumed, reasonably for 2018, was that the primary vectors of automated harm would remain legible enough to be contested — that the problem was one of transparency, accountability, and political will. Eubanks believed, and argued persuasively, that if people could see how these systems worked, democratic pressure could dismantle them. This was the liberal optimism of the Obama aftermath, when "algorithmic accountability" was a phrase gaining traction in policy circles and the idea of auditing public-sector AI seemed achievable. She could not have foreseen the degree to which the transparency movement would be co-opted: jurisdictions now publish model cards and bias audits that function as liability shields rather than instruments of reform. Nor could she have anticipated the pandemic, which accelerated the automation of benefits administration by years, creating systems built under emergency conditions that became permanent infrastructure. Her call for universal basic income, echoing the Poor People's Campaign, reads differently after the brief, strange experiment of pandemic stimulus checks — payments that temporarily reduced poverty and were then withdrawn with bipartisan indifference. The absence in the book that now feels most conspicuous is any sustained engagement with immigration enforcement, where automated systems were already, in 2018, doing precisely what she described in welfare contexts but with even less recourse for those caught in the machinery.

Certain passages land with a weight they couldn't have carried at publication. Her account of Jason's insurance denial — flagged by an algorithm for suspected fraud after a violent attack — now reads as almost quaint in its singularity. By 2026, prior authorization algorithms in health insurance have become a national scandal, with UnitedHealth Group's AI-driven claims denials generating congressional hearings and a public rage that Eubanks's book helped seed but could not have predicted in scale. Her historical framing, tracing the line from the county poorhouse to the database, has been vindicated by scholars who have since built entire research programs on this genealogy. The chapter on Skid Row, documenting how coordinated entry systems collected intimate data from homeless individuals while delivering diminishing resources, now reads as a case study in what happened everywhere: the surveillance expanded, the housing didn't. The most quietly devastating sections remain the ones about the Stipes family and their daughter Sophie, because no amount of systemic analysis can substitute for a child losing medical coverage because a computer marked a form as incomplete. That specificity is what keeps the book from becoming a policy document. It remains a human one.

Eubanks occupies a particular position in a lineage that runs from Frances Fox Piven and Richard Cloward's *Regulating the Poor* through to Cathy O'Neil's *Weapons of Math Destruction* and Ruha Benjamin's *Race After Technology*. She took from Piven and Cloward the structural insight that welfare systems exist to regulate labor markets and discipline the poor, and she gave to Benjamin and others a grounded, ethnographic method for showing how that regulation now operates through code. Her contribution was not theoretical novelty but empirical specificity — she went to Indiana, to Skid Row, to Allegheny County, and she talked to the people being processed. This matters because the conversation about algorithmic harm has since drifted toward abstraction, toward "AI ethics" frameworks and "responsible innovation" rhetoric that floats comfortably above the lives of anyone who has ever been denied benefits by a machine. Eubanks kept the conversation on the ground. The book's influence is visible in the wave of investigative journalism that followed — ProPublica's ongoing algorithmic accountability reporting, The Markup's work on tenant screening and predictive policing — all of which owe something to the template she established: follow the system, find the person it crushed, show the history it extends.

The question the book raises now, which it could not have raised in 2018, is this: if the digital poorhouse has survived eight years of exposure, critique, legislative attention, a pandemic, and a nationwide reckoning with racial injustice — if it has not only survived but expanded — what exactly would it take to close it?