Patrol officers and detectives have long relied on hunches and history to best predict where a crime may happen. But now the Peoria Police Department is the first in Arizona to employ a new technology that assigns patrol missions in areas of high crime probability.
Listening to dispatch and keeping their eyes open on patrol, police officers have long been reactive to instances of crime. Although Peoria police Chief Roy Minter said this has allowed for predictive patterns to emerge.
“You’re reacting to crimes that have already occurred. A lot of the deployment strategy had to do with where we were seeing crime patterns develop as opposed to what we’re doing now with predictive policing where we’re trying to get ahead of the game and we’re trying to take a look at where crime may occur as opposed to the crime that has already occurred," said Minter.
That predictive policing is coming in the form of HunchLab, a new technology based on statistics that calculates probability of when, where and what type of crime will occur.
Although Minter points out the force is still new to HunchLab, he sees potential.
“We’re very encouraged by what we’re seeing," Minter said. "So far we’ve seen an incident where officers were in a HunchLab mission area where there was a burglary in progress call that occurred and the officers were already there. And we’ve also seen officers respond to several suspicious occurrence calls where the officers were already available to respond.”
“It gives us a better direction of where to spend our patrolling time, rather than just driving around aimlessly hoping to find somebody," Peoria Officer Amanda Stucky said.
After the department inputted years-worth of raw data into the software, Stucky now can scroll through different missions HunchLab creates.
“We catch bad guys; we arrest people; we save the day," Stucky said. "That’s what we like to see, so it’s nice that HunchLab has the ability to tone it down for us so it gives us the ability to scroll down and say ‘OK, here’s what this formula just put together.'”
In the HunchLab user interface, color coding indicates the different crime likelihoods. The mission areas are small, covering 250-square-meter boxes, or nearly three football fields, on a map of Peoria.
“When you’re talking about spreadsheets and getting all that information, that’s how we used to get it," said Stucky. "And it’s really hard for visual people to see numbers on a spreadsheet and say, ‘Why is it that I have to go there again?' A lot of cops are visual people and they need to see the nuts and bolts of everything rather than just spreadsheets and numbers and formulas.”
After reviewing crime-pattern literature, Jeremy Heffner and a team who developed the HunchLab software identified different variables and how they may affect people’s behavior.
"Everything from the fact that people change their pattern of behavior from weekdays to weekends, daytime to nighttime," Heffner said. "People congregate in a certain geographic locations like bars. The weather changes whether people go outside or stay in. So we wanted to use all of those concepts to represent crime risk in an accurate way.”
Using algorithms to determine and calculate crime patterns, HunchLab learns how to display data based on individual crimes and how they may be related, essentially eliminating bias.
“That basically allows us to not impose our own thought processes on what should matter in terms of predicting crime in a given place or for a given type of crime," said Heffner. "And so what’s interesting is after the system builds these models we can peer into the models and see things like the nature of a community as opposed to the recent event that happened.”
As Officer Stucky demonstrated HunchLab on a laptop in a patrol vehicle along with other gear, one of those events comes in on the radio. "Break for an armed robbery occurred at the Safeway. 83 and Cactus at the bank," said the radio dispatcher.
When asked if that area is covered in HunchLab, Stucky laughed.
“And that’s kind of the joke amongst us is, ‘Well, was HunchLab gonna know that’s going to be there?" Stucky said. "What’s sad is probably, it did. It probably had it as a theft or something suspicious in that area. It just doesn’t give us the exact, ‘Hey, there could be a bank robbery.’ It’s not that specific but it gives us the, ‘Hey, there is probably going to be a suspicious circumstance or a crime of some kind in that general area.’ But if we went and looked, 83rd (Avenue) and Cactus would have something right now. So I’m going to go try to catch a bad guy."
It will take time to measure the success of the program. Since HunchLab was first employed in September, the department is waiting at least a year to start comparing numbers and see if there is a reduction in crime occurrences.
