WSU lab part of growing AI field helping law enforcement

The Spokane County Sheriff’s Office is one of a number of agencies statewide and across the country turning to artificial intelligence and researchers to review to what Sheriff John Nowels referred ‘as law enforcement’s single largest data set: body camera footage.

In the Evergreen State, much of that work is done at Washington State University’s Complex Social Interactions Lab, which has reviewed more than 40,000 hours of footage for a number of law enforcement agencies since its inception in 2016.

Lab Director David Makin said the trend is an effort to turn those large data sets into usable and actionable information.

“It’s an untapped data source that you’re spending a considerable amount of money to collect and store, but unfortunately, without these tools or integrated practices, it just sits there,” Makin said. “It’s a data source that largely is untapped, when you could use it to improve outcomes.”

The lab partners with a number of law enforcement agencies across the state, and can tailor analysis to what an agency might want to learn from its trove of footage, Makin said. He added that any Washington agency could participate in the lab’s studies free of charge.

For example, an agency may ask the lab to explore possible discrimination within the agency’s force, the effectiveness of an agency’s de-escalation tactics or a review of how often resources are provided following a domestic violence call.

The lab also partners with Axon, the largest provider of police cameras and footage storage in the country. The company’s database of body camera footage sat at around 4 terabytes in 2016, as reported by ProPublica.

As of August, that figure has grown to around 400 petabytes, according to a company letter to shareholders. That’s the storage capacity equivalent to 3,125,000 iPhone 16s, if opting for the cheapest model.

Mankin said his field has grown exponentially over the years, due in part to a somewhat competitive landscape and the piggybacking successes of involved parties. Makin said there are two main camps on the analysis side: researchers like himself who are developing open-source technologies that can be shared widely, and private companies developing proprietary programs and techniques. The company the county has partnered with, Polis Solutions and its software TrustStat, is an example of the latter.

Where TrustStat relies predominantly on artificial intelligence to carry the load, Mankin’s lab uses around 50 student researchers to assist in analysis. The students process the footage, take notes on the information they can gather, then enter it into machine learning algorithms to look for identifiable patterns or connections between what was observed and the eventual outcome of the police encounter.

Mankin said that the human element helps with cases where a person’s judgment might be needed.

“We need some element of subjective kind of assessment, and so that’s what we need our students for, as well,” Mankin said. “They help us label data, but they also give us that vantage of, ‘Well, how well do you think the officer did in this interaction?’”

While implementing an analysis tool, or a program like the Sheriff’s Office’s, is one piece of the puzzle, Mankin said incorporating the technology into an institution plays an equally large role. Integrating the tech into existing practices and daily operations is often where agencies fall short, he said.

“Integrating it means I got to align policy and practice so that if the technology identifies a set of body-worn material videos that I should go look at for various reasons, do I have a policy that tells the person how they have to do it?” Mankin said. “What has to happen next?”

Mankin said they evaluate the success of integrating technology on five main principles: it should lead to an improvement in efficiencies, effectiveness, resource allocation and the ability to deploy resources, while also having an accountability aspect.

“You can have a successful integration if you can do all five of those,” Mankin said. “What’s uncommon is actually going through it when you do your integration and checking, because you could actually implement a technology and actually it produces operational inefficiencies, introduces organizational ineffectiveness.”

“Then we’re kind of back at square one, and most people end up scrapping the technology and saying it didn’t work,” Mankin said.

Nowels said he was not aware of the lab’s ongoing work, and that the agency could have partnered with the lab free of charge. Mankin takes the blame for that, saying the lab and researchers as a whole could do more to build awareness of their efforts.

“We just don’t do well to market ourselves,” Mankin said. “I’m a researcher and I’m not a salesperson.”

Mankin said he looks forward to seeing the results of the Spokane County program, and he hopes the aforementioned board overseeing the project has a research plan on what they hope to learn, how those lessons will be implemented, and so on.

“I think that’s the missing part of all of this, right?” Mankin said. “It’s: ‘We just got a grant to buy this technology.’ Well, does it come with a plan to evaluate its effectiveness or its efficiency?”