In Hollywood, nestled between a strip mall and a recording studio where bands like the Rolling Stones have recorded, the residents of a small homeless encampment greet passers by with a friendly “Hi, hello, how are you doing?”
Some people respond in kind; others seem nervous and terse. But according to one of the most outgoing people here, Cedric — who didn’t want to give his last name — they simply hope that if their neighbors see them as friendly and nonthreatening, they won’t call the cops and have their tents removed. L.A. police and the Bureau of Sanitation have become increasingly strict about the “cleanup” of homeless encampments, even though most residents here have nowhere to move to.
Los Angeles has the second largest homeless population in the U.S. after New York, with an estimated 52,765 homeless individuals in 2018. The numbers are compiled by the Los Angeles Homeless Services Authority (LAHSA), a city agency that helps get people off the streets — and LAHSA says the number of people experiencing homelessness for the first time is increasing.
In an initiative started in January 2018, LAHSA is now sharing data from the Homeless Management Information System (HMIS) with researchers at the Center for Artificial Intelligence in Society (CAIS) at the University of Southern California. The researchers are using the data to build a system that can identify behaviors and outcomes, and allocate the type of housing with the greatest statistical chance of long-term success, while also reducing racial discrimination in the system. The project — Housing Allocation for Homeless Persons: Fairness, Transparency, and Efficiency in Algorithmic Design — brings together researchers from both the engineering and social work schools.
The project is informed by a 2018 study by two CAIS leads — engineering professor Phebe Vayanos and Eric Rice, a professor at the USC Suzanne Dworak-Peck School of Social Work — which examined the efficiency and fairness of housing allocation programs. They analyzed national data on homeless youth, aggregating information on rapid rehousing (schemes that provide money and moving assistance to get people off the streets quickly) versus permanent housing (long-term accommodation, often with support and subsidies), and looking at whether those young people became homeless again. They then devised a theoretical model to allocate housing more fairly and efficiently across the country, estimating that this system would allow 16% more youth to successfully exit homelessness within a year, and reduce the race gap — the percentage of white compared to minority youth who permanently leave the streets — by 72%.
“I can’t sleep around a lot of people — I get paranoid. And I’m actually you know, slow — special ed. So it’s hard for me to do the steps or follow-up on my own. But I’m getting better at it.”
In order to make the system transparent and easy to understand for the administrators who allocate housing, the team devised a policy that would assign a score to youth based on characteristics including age, the reason for homelessness, and where they were sleeping. In the L.A. model, Vayanos says the algorithm would assess candidates based on all of these factors, as well as on the statistical likelihood that the allocation will lead to a permanent exit from homelessness.
The team hopes this theoretical model can be expanded into a live, working system for L.A. That could help change the life trajectory for someone like Lulu, who also lives in the small cluster of tents near the recording studio in Hollywood.
Lulu says he’s been homeless since he fell out with his father when he was 15, some two decades ago. He has tried to get housing from the city throughout his youth and adulthood, but found it difficult to go through a process which can include moving between shelters and temporary housing, as well as a lot of paperwork. “I don’t get along with a lot of people,” he says. “I can’t sleep around a lot of people — I get paranoid. And I’m actually you know, slow — special ed. So it’s hard for me to do the steps or follow-up on my own. But I’m getting better at it.”
According to LAHSA data, out of 5,034 youth who received assistance from LAHSA’s Coordinated Entry System in 2018, only 1,344 ended up in some kind of permanent housing, while closer to 2,000 spent time in interim housing. Race is also a factor: A LAHSA study revealed that although housing allocated through the L.A. city system is fairly assigned among different ethnic groups, African Americans were more likely to become homeless again. The study showed 14.2% of black people would be homeless again with a year, compared to 8% of Latinx and 7.2% of white people.
Peter Lynn, the executive director of LAHSA, says that this is down to “American institutional structural racism” which manifests in employment discrimination, housing discrimination, and disproportionate law enforcement towards black people. While he’s supportive of the CAIS project, he cautions that it would need to go through an intensive “community engagement process” before being deployed so that it had the trust and support of the community.
There are other concerns, too, reflected in recent scandals about poorly engineered automated services based on biased data. From insurance and job applications to school admissions, algorithms are increasingly being used to make decisions that affect people’s lives. So what safeguards can CAIS put in place to make sure that the vulnerable people being assessed by its system are represented fairly?
HMIS, which provides information technology to agencies like LAHSA, has strict requirements regarding the privacy and consent of its clients. Homeless people are required to give informed consent for LAHSA to use their data, and LAHSA operates under an agreement to protect clients’ privacy or have its access to HMIS suspended. Getting any data on homelessness presents a challenge in itself. Since volunteers have to individually compile data on homeless people one by one, it’s very possible that these counts miss people — possibly by as much as half, according to some experts.
“This algorithm isn’t going to solve homelessness in Los Angeles. Is it going to put a dent? For sure.”
Vayanos says the team is trying its best to create a system that is prepared for the possibility of inaccuracy. “We are actively working on an algorithm that does account for inaccuracies in the predictions and in uncertainty in the arrival times of housing resources and homeless persons in the system,” she says. “We want the system to perform well, even if things turn out to be different than expected — for example, if particular individuals turn out to be less successful than expected when placed in a particular type of housing resource.”
Rice understands that the process to develop the system and win the understanding of the community will take time. A.I.-driven work in the real world must happen at real-world speed, “not at the pace of computer science, which is a much faster moving animal,” he says. Rice has worked on a variety of tools to help the homeless, including an A.I. tool that helps select peer facilitators to prevent HIV in homeless youth, and tools that assess the vulnerability of homeless youth that are now used nationwide. He knows from experience the long effort could be well worth it.
“This algorithm isn’t going to solve homelessness in Los Angeles,” he says. “Is it going to put a dent? For sure. Because if you think about, if we can improve the efficiency by even 5% or 6% , that translates into hundreds, if not thousands, of people getting housing that’s better for them.”