LA Experiments With AI To Predict and Prevent Homelessness

By Maggie LaBarge
LOS ANGELES, CA– In 2019, UCLA researchers began experimenting with artificial
intelligence and California databases to predict trends in homelessness with hopes of preventing

They use data from hospitals, prisons, and financial assistance programs to track who is more
likely to experience homelessness. Once they determine this, they plan to help financially with
bills or living situations.

Artificial intelligence uses 580 different factors to rank a list of 90,000 California residents who
have recently used county Health Services or Mental Health departments based on the risk of
experiencing homelessness in the future.

Those deemed highest risk use services like CalFresh more frequently and are seen at a higher
risk in emergency rooms and jails.

In the beginning, researchers trained the algorithm by using data from those who have recently
begun experiencing homelessness. The algorithm then would “predict” trends in those who use
services and their chances of losing their housing.

Although the technology is still in the early stages of development, of the 90,000 on the list
studied, 7% lost their shelter in 18 months. There were 10,000 of those who were deemed at the
highest risk of experiencing homelessness, and 24% became homeless.
Researchers reported that the data becomes even more accurate when smaller groups of 1,000 to
10,000 are studied. However, working in smaller groups becomes challenging due to
communication issues.

Skeptics have brought up concerns about the reliability of artificial intelligence over a social
worker or a human with a background in the subject; however, researchers have determined that
the algorithm is 3.5 times more accurate than a human and eliminates all biases a person may

This program is also unique in removing the need for people to seek out help actively. Many
factors like pride, not knowing where to go, or not knowing they need assistance can make
receiving that assistance very difficult; this way, those at risk of losing housing get the help they
need without feeling the weight of asking for it.

The program is being funded by 26 million dollars from the COVID-19 era funds from the
federal American Rescue Plan. It will continue developing until 2026 when the data will be
released for other researchers to build off.
This technology can be revolutionary in the progression of limiting the amount of people who
encounter homelessness. It will assist counties and cities in spending their funds efficiently and
maximize the effect of programs in the entire United States.
To learn more about the future of this program visit


Scroll to Top