Artificial Intelligence and the Union: Protecting Your Members’ Rights

By: RLS Partner, Robert Wexler
From: PORAC LDF

Picture this: A police officer prepares for her shift. She loads her gear into her AI-enabled patrol vehicle and sets off to conduct patrol. Only, instead of deciding where to drive, her patrol car is self-driving, and her route has been predetermined for her by AI based on an algorithm that analyzed historical crime data to predict where criminal activity will occur. When the officer detects something afoul, she elects to deploy her autonomous robotic canine to investigate, which, using its integrated facial recognition software, alerts the officer that an individual previously identified as an armed and wanted suspect is located. If the suspect flees, the officer will simply notify dispatch to track the suspect using a sophisticated network of AI-enabled CCTV cameras to avoid engaging in a dangerous high-speed pursuit. And if the suspect engages the officer rather than fleeing, the officer may disable him by deploying a conducted energy device (e.g., Taser) from afar, using a drone mounted to her patrol vehicle. Upon returning to the station, the officer logs on to her computer to check the accuracy of the police report of the incident, generated entirely by AI through analysis of the body-worn and dash-mounted camera footage of the incident.

If you think this sounds like something out of a sci-fi movie, you would be wrong — all this AI technology presently exists. Welcome to the dawn of AI in policing, and the need for your association to understand how to protect its members’ rights.

What Is AI?

AI, or artificial intelligence, is a branch of computer science concerned with building computers and machines that can learn, reason and act in such a way that would normally require human intelligence or involve analyzing data on a scale far exceeding what humans can process. AI technology encompasses computer science, hardware and software engineering, data analytics, linguistics, neuroscience and even psychology to simulate human thinking and create machines that learn from data and improve their performance over time without explicit human programming. These machines employ “vision” to interpret and make decisions based on visual information such as recognizing faces and images or interpreting written text. They also use robotics to perform tasks autonomously without human direction. Finally, today’s AI machines benefit from understanding, interpreting and generating human language, which enables them to interact with anyone, not just those familiar with computer coding.

How AI Is Being Deployed in Policing

AI is currently being used by law enforcement agencies in numerous ways, including many articulated in the hypothetical scenario above. The following are some examples of how AI is being used by law enforcement agencies:

• Predictive policing: AI algorithms are currently being used to analyze historical crime data to identify patterns and trends, enabling police agencies to forecast where crimes will occur to allocate personnel, design patrol routes and prioritize response efforts.
• Suspect identification: AI can assist in identifying suspects through facial recognition, by comparing images captured from a crime scene to social media and other sources.
• Detecting threats: By mining massive amounts of data from various sources, such as social media, AI can detect and assess threats in an efficient manner, allowing law enforcement to proactively address them.
• Audio and video redaction: AI technology can instantaneously redact sensitive information from audio and video files to assist in producing information pursuant to public disclosure laws.
• Suspect location and tracking: AI can assist in locating suspects and/or tracking their movements through networked CCTV cameras and license plate readers.
• Hiring of personnel: AI-powered tools can analyze resumes and applications to identify candidates who meet specific qualifications, and combined with predictive analytics, can evaluate the likelihood of a candidate’s success based on historical data from previous hires. This includes analyzing factors like prior job performance, psychological assessments and training outcomes to predict future behavior and effectiveness.
• Body camera analysis: AI can comb through massive amounts of BWC footage, looking for patterns, language and uses of force to proactively allow departments to “flag” troubling behavior. One company, Truleo, is marketing this type of software to agencies as a “virtual sergeant” that can determine which personnel are likely to become “problem officers,” purportedly to provide proactive remedial training.
• Automated report writing: AI can transform BWC footage to written reports, relieving officers from preliminary responsibility for report writing. Axon’s Draft One is one of the early applications now available to agencies.

AI Issues and Police Unions

One of the biggest concerns nationally that unions have with AI is job security. The recent protracted motion picture industry strikes centered around job protection, and that issue should similarly concern police unions. If an officer spends 40% of his or her time on report writing, as some organizations estimate, and AI can effectively cut that time in half or less, then it is not inconceivable that there could be a reduction in the number of officer positions needed. Unions may seek to ensure that any efficiencies achieved through technology will result in a redeployment of personnel, rather than a reduction of personnel. In other words, if report writing time is cut, then unions should seek assurances that officers will be directed to other needs, like increasing staffing of patrol or bolstering special units to address acute needs.

As the workload changes, labor associations should be mindful of the impacts and effects of such changes. For example, a patrol officer might currently spend only seven hours of a 12-hour shift seated in a patrol unit, in part because of the need to write time-consuming reports. If report writing is significantly curtailed by use of AI technology, the time in a patrol unit might increase daily to nine or 10 hours. Unions should be mindful of the physical effects on their members from being seated in a car for extended periods and might seek reasonable limits to protect workers’ health. Also, as AI demands an increased reliance on tech-savvy personnel, there may be an opportunity to negotiate special pay for those who possess the skills needed to work with AI technology.

Moreover, as AI technology becomes more prolific, union leaders should zealously advocate for strong policies to protect against the unwarranted or unlawful dissemination of their members’ personal information. This can be done, in part, by monitoring the employer’s purchase of AI software and preemptively seeking to understand the algorithms, data sources and reporting capabilities of the AI tools to ensure that members’ interests are not harmed by faulty processes or policies.

Finally, as AI tools that measure performance become more widespread, there are likely to be impacts and effects on discipline and promotional opportunities. These factors should provide the union a seat at the bargaining table to negotiate inclusion of provisions in a collective bargaining agreement that: (1) provide the union the right to vet software before it is deployed; (2) assure periodic assessments of any AI tools that are used; (3) require an employer to timely notify the union when errors or other “glitches” are discovered within AI systems; (4) prohibit personnel decisions from being made solely based on AI algorithms; and (5) provide the union access to audit tools of what AI software is being used, the reports or other information those AI tools create and who has access to that information.

The time is now for law enforcement union leaders to become educated about AI. Understanding both its benefits and its limits is critical to effectively advocate for guidelines to ensure the safe, efficient and appropriate use of this emerging and powerful technology.