
Predictive AI tool helps Edmonton organizations decide how to respond to safety challenges
An architect of Unison, a City of Edmonton tool that uses artificial intelligence to map data from safety incident reports to predict where responses will be needed and inform what those responses should be, said that after several years in use, the tool is now coming into its own.
"Unison sets itself apart in three really important ways," Kris Andreychuk, the manager of data science and research at the City of Edmonton, told Taproot. "The first one is that we are building one common operating picture by aggregating all of this data. Thing two is privacy by design. And thing three is that these solutions are just wildly expensive if purchased from third-party vendors, but our corporate leadership team has given us permission to facilitate access to our partners who are in the business of public safety."
Launched in 2022, Unison draws data from the City of Edmonton, Edmonton Police Service, REACH Edmonton's 24/7 Crisis Diversion program, the University of Alberta, Southgate Centre, and Edmonton City Centre (the owner of which entered receivership on July 7).
Andreychuk's team uses AI to analyze the data and determine what kind of safety incidents happen where and when with the above partners — as well as with Allied Universal, ATB Place, the Edmonton Downtown Business Association, Edmonton Public Library, GardaWorld, NAIT, Paladin Security and Paladin Technologies, and Securitas
But part of the "privacy by design" Andreychuk mentioned is limiting what data Unison uses. Unison, for example, only draws on the type of incident, like an assault, where it happened, and when it happened. Andreychuk said his team does not use data from incidents that specify race or gender, or types of incidents that include, for example, transit fare evasion or creating graffiti. Unison only includes demand data, which means when a member of the public has a direct interaction with one of the partners.
"This is an area and a discipline that I think is rightfully subject to a lot of scrutiny and criticism around data ethics and data privacy," Andreychuk said. "From the perspective of data minimization, what's the least amount of data that we require to get our partners the insight they need?"
This point is particularly important when it comes to policing, Andreychuk said. "We don't want to be predicting against actual crime data, because it can create a feedback loop," he said. "It can put us in a situation where we could be over-policing."
The first version of Unison launched three years ago but Andreychuk said it has come into its own over the past year. He demoed the tool during the EDBA and BOMA Edmonton and North's Downtown Safety & Vibrancy Summit on June 17. Later, he shared on LinkedIn that the demo generated at least 16 requests for access to Unison.