Police Bodycams: If You Film It …
by Michael Dean Thompson
One hundred petabytes is a difficult quantity to comprehend. In plain English, that is about 113 quadrillion or 113 followed by 15 zeroes. According to ProPublica, that is the rough data equivalent of 25 million copies of the movie Barbie. One hundred petabytes is also approximately the volume of bodycam video held by Axon’s cloud storage system. In that sense, then, even the comparison to Barbie does not quite capture the magnitude of the data, as the bodycam data is not as prolific a bit generator as a high-definition movie. For that reason, it is far more than the 5,000 years of high-definition video the Barbie comparison implies. New York City alone generates millions of hours of bodycam video per year. The numbers continue to grow. And the majority of it remains unwatched.
Police bodycams came about as an intended solution to a problem. It was hoped that the tools would help build back public trust after several high-profile police killings. It is certainly true that transparent use of video footage can reveal what actually happened during a disputed encounter. In 2020, Louisiana State Police arrested Antonio Harris, during which the troopers kneed, slapped, and punched the man after he surrendered. Despite the troopers’ “wholly untrue” reports about the incident, the bodycams showed at least part of the truth. The first trooper to approach Harris—who had already surrendered—kneed and slapped Harris before thinking to turn off his bodycam. The troopers later laughed and bragged about the incident via text messages.
The promise of bodycam truth telling extends beyond examining incidences after the fact, which is too late. Such violent police encounters are rarely isolated incidents. Cameras that are always on enable systematic reviews of officer behaviors so that problematic behaviors can be captured, flagged, and addressed before they escalate. Furthermore, the videos can then be used to train future cops to identify both effective and destructive behaviors. The problem is one of scale. Cops can either hire hordes of bodycam video watchers or find some automated mechanism of flagging suspicious or problematic police activity.
Polis Solutions and Truleo are among an increasing number of companies attempting to use AI-based solutions to do just that. Paterson, New Jersey, hired Truleo to review their footage after police killed a community activist who in the midst of a mental health crisis called 911 for help. Truleo’s software allows police supervisors to identify which behaviors to flag. Those behaviors can range from interrupting civilians and using profanity to muting the camera or using force. These are behaviors that would have flagged the Louisiana troopers who initially failed to reveal any footage existed to investigators. Truleo has found that “There are officers who don’t introduce themselves, they interrupt people, and they don’t give explanations. They just do a lot of command, command, command, command, command,” Anthony Tassone, cofounder of Truleo, told ProPublica. “That officer’s heading down the wrong path.”
Polis Solutions of Dallas, Texas, has its own software called TrustStat that grew out of a Defense Advanced Research Projects Agency project that sought to understand how soldiers in what may be hostile environments could prevent conflicts from escalating. The AI tools within TrustStat analyze speech, facial expressions, and even body movements to try and flag both positive and negative encounters.
Similarly, Washington State University’s Complex Social Interactions Lab uses a combination of 50 reviewers drawn from the university’s students and AI to review videos from Pullman, Washington, to identify features and outcomes of police behavior.
Unfortunately, one of the key promises of bodycam video—police transparency—has not been fulfilled. As ProPublica notes, departments using Truleo have not been willing to make their findings public. Meanwhile, police departments in Seattle and Alameda, California, canceled their contracts after backlash from police unions. In Philadelphia, department policy prevents officers caught violating procedures during bodycam spot checks from being disciplined. And, across the country, police departments themselves are the gate keepers when it comes to public access to the video.
The truth is that catching the trends early and intervening can save lives and prevent unnecessary brutality, but that seems to have little effect on policy. As the case with the Louisiana troopers who attacked Harris shows, we cannot expect errant cops to be prosecuted. All three of Harris’ attackers are having their charges quietly dismissed.
Sources: ProPublica.org, Associated Press
