TrustStat AI to Begin Monitoring Spokane Sheriff Deputies’ Bodycam
Artificial intelligence could be the new “virtual sergeant” at the Spokane Sheriff’s Office. AI may soon be assessing every word spoken by patrol deputies on their bodycams. The U.S. Attorney announced that the sheriff’s office is receiving $945,520 to implement TrustStat AI, which will “analyze body-worn camera video and identify key behaviors and language related to de-escalation, use of force, and other critical areas of deputy performance.” As a defense lawyer in the court system, it will be interesting to see how this is implemented.
While high-tech surveillance in Spokane has traditionally been aimed at the public—through drones and license plate readers—AI is now being used to monitor police bodycam footage, with software from TrustStat and Truleo. It’s unclear whether the local union was (or will be) involved in the implementation of AI monitoring of its members. Some police unions have opposed AI surveillance, and departments in cities like Seattle and Vallejo, CA, have successfully shut down AI monitoring of their officers’ interactions.
Many people assume that bodycam footage is reviewed routinely, but it isn’t. There are simply too many hours of video, and police supervisors typically only monitor footage when there is a complaint. Prosecutors and criminal defense lawyers usually review only selected clips relevant to their cases. As defense lawyer I sometime use AI to analyze bodycam videos and transcribe them, but the software I use is rather rudimentary. Artificial intelligence software like TrustStat and Truleo, however, analyzes every word spoken by officers, their tone of voice, use of swear words, whether they interrupt others, and whether their language is clear. The AI system begins by collecting a voice sample from each officer, known as a “voice print,” to identify them. Keep in mind that an officer’s voice can be captured not only on their own bodycam but also on the bodycams of their colleagues. TrustStat also collects data on arrestees and members of the public, but it reportedly makes efforts to anonymize that data. Truleo, in its initial public stock offering, promised to create “baseball card stats” for each officer and emphasized that “body camera data needs to be analyzed so we can a) reward good officers, b) remove or retrain bad officers, and c) train new officers.”
The ACLU initially opposed AI analysis of police bodycams, but civil libertarians have expressed concerns even when police departments abandon AI monitoring. After pressure from the police union, the Seattle Police Department abandoned the use of Truleo. This decision followed an incident in which a police detective was overheard joking about the death of a pedestrian killed by a police officer speeding through town on an emergency call. A similar controversy arose when Vallejo, CA, cancelled its use of AI to monitor officer conduct. Truleo’s Twitter account reportedly stated: “[Police] chiefs around the country lack the courage to analyze 100% of their videos because they suspect a HUGE portion of the department is performing unprofessionally.” So, it will be interesting to see how TrustStat is received in Spokane. The idea of employees being monitored for professionalism shouldn’t be too surprising in modern American culture. For example, employees on the phone are often monitored for their language use. In many workplaces, emotion-detecting AI monitors biological signals such as tone of voice, facial expressions, and data from wearable devices to gauge how employees are feeling. It seems the motivation behind police agencies using AI is to avoid lawsuits, as the software could help them identify and address potential problems. The technology also appeals to police agencies seeking to exit federal consent decrees.
It will be interesting to see how this unfolds. In the past, police officers often had colorful vocabularies and could seem rather hard-nosed to outsiders. The stressful nature of their work traditionally leads to a certain level of gallows humor. The use of TrustStat and Truleo will undoubtedly offer benefits but could also result in police officers sounding more like customer service agents at a call center. Officers will likely learn to tailor their interactions to please the algorithm. As one article pointed out, “Truleo is clearly following a formula—officers who use more than 25 words to explain something get points for ‘professionalism.’” Some officers reportedly told researchers they were “playing the game” according to Truleo’s rules.
A major question is how much of the AI analysis will be made public. Would a criminal defense attorney defending a resisting arrest charge be entitled to access an officer’s stats? Typically, if a public employee is coached or given guidance, that information is not considered a public record. Personnel records only become public when an officer is disciplined, as established in Cowles Publishing Co. v. State Patrol (1988).
Who’s next in the criminal justice system for AI analysis? Would judges be open to having their proceedings assessed by AI? Given that most court hearings are already audio-recorded, it would seem relatively easy to implement. Judges exercise far more discretion than police officers, so the potential for AI to analyze judicial conduct could raise different ethical and legal questions.
What do you think? Add your comments below.