Montreal's AI‑driven strategy to prevent subway suicides

The AI system, which is still under development, employs approximately 2000 surveillance cameras installed at subway stations. Société de Transport de Montréal (STM) – the public transport company of Montreal, and the Centre for Suicide Intervention (CRISE) are working together on this endeavor. The AI analyses the live feed from the cameras, identifying individuals who may be under severe stress.

Professor Brian Mishara, the director of CRISE and a Psychology professor at the University of Quebec in Montreal (UQAM), who is also a widely recognized expert in suicide prevention, elucidated the workings of this system to CBC, a public television network. "We have clues, but of course, humans cannot monitor hundreds of screens all day to identify such behaviors," Mishara explained. Therefore, AI is being used; it doesn't utilize facial recognition algorithms but concentrates on characteristic behaviors.

Professor Wassim Bouachir of the University of TÉLUQ, a Quebec-based university specializing in remote learning, delineated some of these behaviors to the "La Presse" newspaper. "People who spend a lot of time staring into the tunnel, people moving close to the edge of the platform, or frequently crossing the yellow line, and those who let many trains pass – these are all clues that could indicate a suicide attempt," Bouachir explained.

Mishara has been researching subway suicides in Montreal for many years. In his 1999 paper published in the Canadian Journal of Psychiatry, he evaluated coroner reports from 1986 to 1996 relating to 129 suicides in the Montreal subway. He found out that the majority of individuals who committed suicide had mental health issues, with 81 percent revealing their intent.

In 2016, Mishara published an analysis of the behaviors of subway suicides in Montreal, captured on camera from 2010 to 2013, in the scientific journal "BioMed Central Public Health." This represented a pioneering study, as reported in a UQAM press release. Mishara observed that individuals who left their belongings on the platform and repeatedly paced between the edge and wall often indicated potential suicide risks. He claimed in the UQAM press release that "based on these two behaviors, captured through the cameras, almost one-fourth (24 percent) of the suicides could have been prevented."

The ongoing pilot program is considering even more factors. The objective is to build the AI algorithm to potentially alert station security in time to intervene, or even notify the subway conductor to stop the train before entering the station. Bouachir told "La Presse" that, following accuracy tests, the system should be ready to be implemented within a maximum of two years.