Over a decade before becoming an assistant professor of criminology and criminal history at the University of South Carolina, Dr. Ian Adams worked as a police officer in Utah, and he had the rare chance to help test body-worn cameras for the department. The opportunity excited Adams, who always had been drawn to the intersection of technology and law enforcement.
This article appeared in the January/February issue of OFFICER Magazine. Click Here to subscribe to OFFICER Magazine.
The project was a mixed blessing, though. Adams wasn’t only using the device to record video, but he was also responsible for reviewing the footage and getting the videos properly distributed.
“I’m a tech guy and found it very interesting, but it pretty quickly grew overwhelming,” Adams tells OFFICER Magazine. “I was spending more and more of my week sending footage to defense attorneys and prosecutors and all the other interested parties you might need to provide this technology or this video to. So eventually I went to my sergeant and said, ‘Yo, I can’t do this. I’m a patrol cop. I can’t be spending all my week just distributing and editing or redacting these videos.’”
Years later, that challenge provided a springboard of sorts for Adams’ doctorate dissertation that looked at the practical considerations police departments face when it comes to reviewing and managing hundreds of hours of video footage collected by dashboard and body-worn cameras. Instead of requiring officers to spend more time in front of a computer screen and not on patrol, Adams and other law enforcement officials are turning to artificial intelligence to help close cases faster and keep officers safe on the streets.
Artificial intelligence and law enforcement
Artificial intelligence had a coming out party in 2023. After lingering for years in the shadows in everything from video games to search engines, the technology became the year’s media darling, appearing in headlines on the news, financial and entertainment pages. Terms like deepfake, ChatGPT, deep neural network and Midjourney inundated the public, which at times expressed ethical and legal concerns over AI’s application.
When it comes to law enforcement, artificial intelligence has trended in similar patterns with the public. For years, the technology has been instrumental in everyday tools such as automated license plate readers, that are deployed by agencies during traffic stops and criminal investigations. As the ubiquity of body-worn cameras and security cameras increases, agencies have employed AI-powered facial recognition software in criminal investigations.
Companies, such as i-PRO, incorporate AI software into their cameras and other devices with a goal of eliminating the tedious work of maintaining body camera footage, so that officers can spend more time in their communities, says David O’Connor, i-PRO’s director of public safety
“If (an officer) is never out there on the street or with the folks, then it seems like something that’s strange, foreign, removed, isolated and not it’s not as relatable,” he says. “So, I think by and large, most police departments treasure the time that they are able to keep their officers in contact with members of the public and make progress on those relationships and improve the situation. We want to automate and improve anything that’s a commonplace, regular task.”
According to a report by Criminal Justice Testing and Evaluation Consortium (CJTEC), AI-powered software has been developed sophisticated enough to allow police officers to dictate reports instead of typing them out. Along with saving time, the software would maintain proper grammar and eliminate biased language. Portability of the software also lets officers dictate reports at their desk or in their cruiser.
“We’re in probably what will eventually be recognized as the most extreme technological change that the profession has seen,” says Adams. “I think we’re in the middle of that, from not just body cameras, but the increasing reliance on technology for crime control, for operations, for budgeting, intelligence gathering, I think across the gamut of police activity, you’re going to see technological innovation. And that’s not that unusual, right? As a sector, policing is a lot like any other public or private sector, like technology is advancing those in those areas as well.”
Artificial intelligence vs. machine learning
“Artificial intelligence” might have been last year’s buzzword, but it’s not a term Adams bandies about. He blames the “scary” connotations conjured up in popular culture, thanks to fictional AIs like the autonomous, human hunting Skynet of the Terminator movies, or even the real-world script and likeness generators that were points of contention during this summer’s Hollywood strikes.
“We tend to throw around artificial intelligence any time something seems exciting, and at its heart, what we’re dealing with is just some pretty complex algorithms that don’t come anywhere near true artificial intelligence,” says Adams. “But that’s sort of just my little professorial soapbox.”
Instead, he prefers the term “machine learning” as a better way to describe the high computational algorithms and calculations being performed. That paints a more accurate digital picture, so to speak, of what’s actually being accomplished.
“If you think about artificial intelligence as being on the other side of the digital divide, that’s sort of responsive to us, I don’t think we’re there yet,” says Adams. “I’m a believer that we’ll get there, but we’re not there.”
‘It’s not a problem humans can solve’
Some law enforcement agencies are using artificial intelligence’s strengths to create a better organizational culture. Chicago-based Truleo has contracted with several police departments and sheriff’s offices to review in-car and body-worn camera footage. This approach, however, isn’t just about creating efficiencies for departments. It’s also about reviewing body camera videos in order to evaluate officers, spotting good and bad practices and using the footage as a teaching tool.
“The body camera videos we look at like an athlete. We’ll look at it like game time footage,” says Anthony Tassone, CEO and co-founder of Truleo. “That’s like the exact same approach in mindset we take to policing the body camera videos. They contain all of the officer’s interactions across the department. With AI and technology, you can do things that you couldn’t have done five years ago, for example.”
The Aurora Police Department in Colorado is one of the departments that has contracted Tassone and Truleo. The company also has worked with departments in California—Alameda, Atwater and Anaheim—Castle Shannon, Pennsylvania, and Paterson, New Jersey.
“We can process like one year of Aurora Police Department data in a single weekend,” says Tassone. “Now every interaction across the entire department, across an entire year can now be cataloged and indexed in a weekend. So on Monday, you could just search and say, ‘Show me every stop and search. Show me every stop and search where my officers were highly professional. Show me every stop and search where my officers are highly professional and there was non-compliance by the community member, ‘OK, great.’ Those are excellent reels that I want to look at. I’ve got I’ve got community members that are not following commands. My officers giving explanation, introducing themselves. They don’t use any threats, insults or profanity.”
As Truleo and Art Acevedo, then-interim chief for the Aurora Police Department, began their project, Adams put together his research team of and Dr. Jeff Alpert, also of the University of South Carolina, and Dr. Kyle McLean of Clemson. Adams said the trio is conducting randomized controlled trials in Aurora and an agency in South Carolina, but it’s still too early to draw any conclusions.
Although Adams might still be waiting on his research findings, Tassone and others already see the benefits by performing functions—in this case, reviewing thousands of hours of body camera footage—without dismissing the human element.
“It would be impossible for humans to review all of this data. It’s impossible,” Tassone says about evaluating the footage. “What we did was we had humans train and teach our AI so we hired a bunch of former cops, and they taught and trained our AI for what they’re looking for…
This is what a pursuit is. This is what non-compliance is, what a frisk is. Even going into more granular detail, like this is a frisk vs. this is a non-consensual search vs. this is a consensual search. So it’s all been trained by cops, and it’s further refined by our departments using the product.
“It’s just making the AI get smarter and smarter and smarter. We’re basically crowdsourcing what phenomenal policing looks like. That’s what’s happening. The AI model is just learning. What do departments want to happen vs. what do they not want to happen.”
