Algorithmic rationing requires sacrificing one human right for another
In November 2017, Philip Alston, the U.N. Special Rapporteur on extreme poverty and human rights, undertook a historic two-week mission across the United States. Noting that there exists "pretty extreme levels of poverty in the United States given the wealth of the country," Alston set out to document the human rights implications of systemic poverty in places including Charleston, West Virginia, Selma, Alabama, and Los Angeles' own Skid Row.
In a damning report released last month, Alston summarized his findings. Many of its central observations, while refreshingly candid compared to U.S. politicians' rhetoric on the same topic, were predictable: compared to other high-income countries, the U.S. provides a woefully inadequate social safety net, and policymakers place an "an illusory emphasis on employment" to justify cutting it further. Poverty itself is increasingly criminalized as a means of "conceal[ing] the problem," while "racism is a constant dimension." Those living in poverty are "demonized," while fraud is "used as a smokescreen" to make dwindling benefits even less accessible.
Yet Alston's report also examined emerging issues that have yet to receive much mainstream attention—including the use of algorithms and automated decision-making in rationing access to sparse resources.
Skid Row provides a prime example. The Coordinated Entry System (or CES) is a centerpiece of Mayor Garcetti's "Comprehensive Homelessness Strategy" and has been used since 2014 to rank the needs of Los Angeles County's nearly 58,000 homeless residents. To match the unhoused with available apartments, CES relies on a survey called the Vulnerability Index-Service Prioritization Decision Assistance Tool, or VI-SPDAT, administered by outreach workers and service providers. The VI-SPDAT not only collects extensive personal information, such as Social Security Numbers and immigration status, but also asks respondents about a wide range of intimate topics including their mental health, substance abuse, domestic violence history and experience selling drugs or exchanging sex for money.
The data from the VI-SPDAT then feeds into the LA Homeless Management Information System, where an algorithm ranks respondents from 1 to 17 in order of their perceived need. Those with the lowest scores do not qualify for a housing intervention, while those scoring between 4 and 7 are assessed for "rapid re-housing," which aims to provide individuals who have only recently become homeless with a short-term rental subsidy and limited case management services. Those with the highest scores, generally deemed to be at the greatest risk of a health emergency or even death, are prioritized for permanent supportive housing.
In a compelling new book, political scientist Virginia Eubanks explores the origins and impacts of CES through the experiences of people who have participated in the VI-SPDAT as well as those who designed and are implementing the system. While she's careful to point out that it has worked for some, at its core, CES is yet another system designed to manage poverty rather than solve it. According to Eubanks, more than 13,000 units of low-income housing in Skid Row have been eliminated since 1950. Given the diminishing accessible housing stock, the goal of CES is to make sure the remaining limited resources have the maximum impact. But in the face of such a fundamental flaw—the shortage of affordable housing in Los Angeles—CES ultimately serves not to fix homelessness, but to recast a massive political problem as a technological challenge.
For participants in the system, the low odds of actually getting placed in housing mean the CES can produce frustrating and seemingly arbitrary outcomes. Monique, a woman Eubanks interviewed—who said completing the VI-SPDAT was "like talking to my therapist,"—described getting off the waitlist for a supportive housing unit downtown as "the best Christmas gift I ever got." Yet she was troubled that many of the other women she knew from the Downtown Women's Center "went through the same shit that I did, and three years later they're not housed." Outreach workers who administer the VI-SPDAT are advised not to tell people they will get a score, keeping the process by which it operates opaque. Monique estimated she was likely a 10.
Another interviewee, Gary, illustrated the perils of being ranked somewhere in the middle. In the early 2000s, Gary was laid off from his job with a mortgage lender—one later implicated in fueling the Great Recession—due to outsourcing. A few months later, Gary's car was impounded for being parked illegally, and he didn't have the cash to get it back. Having lost his only asset and with his unemployment running out, Gary could no longer afford to pay his rent. Newly homeless, he relocated to Santa Ana to access Orange County's social services. After accumulating 25 tickets for offenses like jaywalking and unlawfully remaining in a public park, a judge told Gary he could either go to jail or leave the county permanently. He chose the latter, ending up on Skid Row.
Despite completing the VI-SPDAT three times, Gary has never been matched with housing, which he attributes to scoring too low for perceived need. According to Eubanks, while those who receive low rankings may be eligible for rapid re-housing, the chronically homeless, absent a serious medical condition or mental health emergency, often fall into a no-man's-land of mid-range scores.
Further, there is the risk that the sensitive data collected through a process that may well yield no payoff will be used against the individuals who provide it. As Eubanks reports, the survey data is already shared with 168 organizations and agencies, including the LAPD, "when required by law or for law enforcement purposes...to prevent a serious threat to health or safety."
The requirement to share deeply personal information in exchange for public benefits or services is nothing new, nor is the deep entanglement of law enforcement with social services. Yet with big data and automated decision-making playing an ever growing role in the administration of public assistance, the potential for surveillance has never been greater. As Alston observed, "despite the good intentions of officials in Los Angeles, there is an Orwellian side to CES." Its requirements that unhoused individuals divulge "the most intimate details of their lives" in exchange for a slim shot at permanent housing leaves many feeling like they are sacrificing one human right for another.
The past few years have yielded some important steps to reduce homelessness and increase affordable housing in Los Angeles. Proposition HHH, which L.A. voters overwhelmingly approved in 2016, aims to finance the construction of 10,000 units of permanent supportive housing over the next ten years. Meanwhile, the L.A. Tenants Union has mobilized to end L.A.'s restrictions on rent control and strengthen protections against evictions. There have also been important legal victories, such as the 2012 Ninth Circuit ruling in Lavan v. City of Los Angeles, which established that seizing and destroying the belongings of homeless individuals violated the Fourth Amendment.
But the failures and setbacks perhaps tell us more than the victories about the nature of the challenges ahead. For example, skirting the limits of Lavan, an amendment to the municipal code (LAMC 56.11) limits the amount of property unhoused people can possess and criminalizes the storage of personal property on sidewalks—yet another effort to make homelessness less visible without addressing its fundamental causes, while increasing the potential for harassment. As documented by the Los Angeles Community Action Network (LA-CAN) and DSA-LA's Street Watch, police sweeps of homeless encampments remain common, which often lead to citations, arrests, the confiscation of property and, occasionally, violent encounters. Between 2011 and 2016, arrests of unhoused Angelenos increased 37%.
Despite voters' support for Proposition HHH, the city has made little headway toward realizing its objectives, while some of L.A.'s wealthier communities have actively resisted concrete measures to address the homelessness crisis. In Venice, for example, residents rallied in the fall of 2016 against converting a vacant senior center into a storage facility for the belongings of unhoused people living in the area, with the president of the Venice Stakeholders Association bemoaning that it would "attract hundreds of transients to the site." As Eubanks observes, this example underscores that the problem is not for lack of data: It is the result of "explicit political resistance from organized elites," for whom a data-based system like CES provides the veneer of a solution without disrupting the status quo, and assuages their guilt in the process.
California has long prided itself on being the most progressive state in the nation—perhaps no more so than in the Trump era. Following the Special Rapporteur's visit, much of the media coverage focused on his findings from Alabama. Readers were rightfully shocked by his descriptions of raw sewage on the outskirts of Montgomery. But neglect also manifests in mundane processes of categorization, and human rights abuses extend beyond those that can be vividly photographed. Addressing homelessness and poverty in Los Angeles will require moving beyond progressive politics as usual to a collective reorientation of priorities—and a collective demand of housing as a human right. As Eubanks concludes,
"If homelessness is inevitable—like a disease or a natural disaster—then it is perfectly reasonable to use triage-oriented solutions that prioritize unhoused people for a chance at limited housing resources. But if homelessness is a human tragedy created by policy decisions and professional middle-class apathy, coordinated entry allows us to distance ourselves from the human impacts of our choice to not act decisively."
Virginia Eubanks' new book is Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. A portion of proceeds will benefit the Los Angeles Community Action Network (LA CAN).