����ɫ��

A Schulich School of Engineering graduate believes there might be a role for technology in helping address chronic homelessness.

Ahead of his convocation in June, Caleb John put the finishing touches on his master’s thesis, which focuses on how data collected by homeless shelters could be used to identify clients at risk for chronic homelessness.

“Supporting vulnerable populations is very important to me,” he says. “This project was a way for me to take my engineering education and apply it in an area where it can directly help those in need.”

John says the results of his work show that modern homeless shelters could benefit greatly with the use of machine learning algorithms.

Housing first

Over the last few years, shelters have been shifting toward a “housing first” strategy, where clients are placed in stable housing without pre-conditions or barriers.

John says his work is in lockstep with that strategy by identifying individuals who will need additional supports to get out of homelessness.

Using a recommender system will help shelter operators identify individuals in need who might fly under the radar.

John warns the proposal is only for a tool that operators can use to help them identify clients and that an automated tool shouldn’t be the only factor in deciding how housing is allocated.

Inspired work

Dr. Geoffrey Messier, PhD, is a professor in the Department of Electrical and Software Engineering at the Schulich School of Engineering. He’s been working on and was the adviser for John’s thesis.

“Caleb’s research provides a way to create tests that can identify individuals in need of housing supports who may have otherwise fallen through the cracks,” Messier says. “He has taken the practical approach of creating threshold-style tests that are both accurate and possible to implement on the IT infrastructure found in most emergency shelters.”

He says the metric will be built into the DI’s system for use in the coming year.

Expanding the scope

John is looking forward to getting the system into the hands of shelter operators.

“The next step will be to re-evaluate the definitions of chronic homelessness and build a system more aligned to the lived-experience of shelter users,” he says. “Aside from that, there are some interesting technical directions that this work can take, and I know some of those will be explored by my teammates and supervisor.”

One of those possibilities is to expand the use of machine learning to include other segments of the homeless population.

“Another important group consists of people who experience episodic homelessness,” Messier says. “These individuals access shelter services sporadically over long periods of time and often fly under the radar. Caleb’s work could be extended to identify and help those people as well.”